learn_rag_from_scratch1.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "\n",
    "# simple rag locally"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# document/text processing and embedding creation \n",
    "# import pdf doc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# process text for embedding, split into chunks of sentences.\n",
    "# embed text chunks with embedding model \n",
    "# save embeddings to file for later (embeddings will store on file for many years or until you lose your hard drive)\n",
    "\n",
    "import os \n",
    "import requests "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "File maiti2020.pdf exists.\n"
     ]
    }
   ],
   "source": [
    "# get pdf doc path \n",
    "pdf_path = \"maiti2020.pdf\"\n",
    "\n",
    "# download pdf\n",
    "if not os.path.exists(pdf_path):\n",
    "    print(\"[INFO] File doesn't exist, downloading...\")\n",
    "\n",
    "    # enter the url of the pdf \n",
    "    url = \"https://sci-hub.se/https://ieeexplore.ieee.org/document/9293712\"\n",
    "\n",
    "    # the local filename to save the downloaded file\n",
    "    filename = pdf_path\n",
    "\n",
    "    # send a GET request to the URL\n",
    "    response = requests.get(url)\n",
    "\n",
    "    # check if the request was successful\n",
    "    if response.status_code == 200:\n",
    "        # opem the file and save it \n",
    "        with open(filename, \"wb\") as file:\n",
    "            file.write(response.content)\n",
    "        print(f\"[INFO] The file has been downloaded and saved as {filename}\")\n",
    "    else:\n",
    "        print(f\"[INFO] Failed to download the file. Status code: {response.status_code}\")\n",
    "\n",
    "else:\n",
    "    print(f\"File {pdf_path} exists.\") \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: PyMuPDF in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (1.24.9)\n",
      "Requirement already satisfied: PyMuPDFb==1.24.9 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from PyMuPDF) (1.24.9)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "%pip install PyMuPDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: tqdm in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (4.66.4)\n",
      "Requirement already satisfied: colorama in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from tqdm) (0.4.3)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "%pip install tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import fitz\n",
    "from tqdm.auto import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d02159d4c7a94dbfb67315a8e81c9bb6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "0it [00:00, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "[{'page_number': 0,\n",
       "  'page_char_count': 5703,\n",
       "  'page_word_count': 876,\n",
       "  'page_sentence_count_raw': 35,\n",
       "  'page_token_count': 1425.75,\n",
       "  'text': 'Indian Stock Market Prediction using Deep Learning Ayan Maiti Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India ayan.182ma008@nitk.edu.in Pushparaj Shetty D Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India prajshetty@nitk.edu.in Abstract—In this paper, we predict the stock prices of ﬁve companies listed on India’s National Stock Exchange (NSE) using two models- the Long Short Term Memory (LSTM) model and the Generative Adversarial Network (GAN) model with LSTM as the generator and a simple dense neural network as the discriminant. Both models take the online published historical stock-price data as input and produce the prediction of the closing price for the next trading day. To emulate the thought process of a real trader, our implementation applies the technique of rolling segmentation for the partition of training and testing dataset to examine the effect of different interval partitions on the prediction performance. Index Terms—LSTM, GAN, generator, discriminant, neural networks, technical indicators, rolling segmentation I. INTRODUCTION The prediction of stock prices and the direction of the price movement is a widely studied topic in many ﬁelds including trading, ﬁnance, statistics, and computer science. Typically, the professional stock market traders use funda- mental and/or technical analysis of stocks to analyze the market elements/indices to make investment decisions [1]. Fundamental analysis is the conventional method involving the study of company fundamentals such as sales, revenue and expenses, net proﬁt, debts, annual growth rates, and so on, whereas technical analysis, is exclusively based on the study of historical price movements. Practitioners of technical analysis refer to the price charts to study price patterns and use price data in different calculations to forecast future price movements. The technical analysis paradigm assumes, there exists an inherent correlation between the performance of the company and its stock price and therefore, it is used to determine the timing of buying and selling the shares of the company. Time-series prediction problems like the prediction of stock prices are a difﬁcult type of predictive modeling prob- lem. Such problems usually involve the presence of sequence dependence among the input variables [2]. A robust type of neural network equipped to handle the sequence dependence is the Recurrent Neural Networks (RNNs). The Long Short- Term Memory network or LSTM network is a type of RNN and is used extensively in time-series problems because of its capability to handle dependency between both long and short distance sequences [3]. Generative adversarial network (GAN) has been successfully applied to a wide range of ﬁelds, such as image in painting, semantic segmentation, and video prediction. GAN consists of a discriminative network D, which learns to distinguish if a given data instance is real or not, and a generative network G, which learns to confuse D by generating high-quality data [4]. This approach has also been used for the prediction of the high-frequency stock prices of the Shanghai and Shenzhen stock exchanges of the Chinese stock market [5]. However, it has not been used to predict the daily stock prices of the Indian stock market as per our knowledge. Our contributions in this paper are as follows: 1. Im- plemented two models for daily frequency stock price prediction- LSTM and GAN 2. Performed rolling segmen- tation on the training set and testing set of the raw data to investigate the effect of the model parameter update cycle on stock forecast performance. We found that LSTM outperformed GAN signiﬁcantly with an error of 0.074 compared to 0.32 obtained by GAN. It implies that the price predicted by LSTM differs from the original price by 7.4% whereas for GAN it is 32%. II. RELATED WORKS & BACKGROUND In the recent times, deep learning is emerging as a promis- ing area of research and has been applied for classiﬁcation and prediction tasks in various domains like text classiﬁca- tion, image processing, speech recognition, and many others. It has also been applied in the prediction of time-series since it is suitable to address chaotic data, randomness, and non- linearity [1]. For example, Chong et al. [6] have utilized the capability of deep neural networks to extract abstract features from data by using a deep feature learning-based stock market prediction model, which extract information from the stock return time series without relying on prior knowledge of the predictors and tested it on high-frequency data from the Korean stock market. Lately, LSTM network, which is suitable for learning temporal patterns, is extensively utilized for various tasks of time-series analyses [7]. LSTM is preferred over the conventional RNN as it overcomes the problem of vanishing (or exploding) gradients and as it can effectively learn long- term dependencies through memory cells and gates. Thus, many studies on ﬁnancial time-series modeling are conducted using LSTMs. However, Zhou et al [5] have adopted GAN to predict stock prices on highly volatile stocks in the China Stock Exchange market. GAN has not been applied much to ﬁnancial markets across the globe and certainly not on the Indian Stock Market to the best of our knowledge. In this work, we implement both 2020 IEEE REGION 10 CONFERENCE (TENCON) Osaka, Japan, November 16-19, 2020 978-1-7281-8455-5/20/$31.00 ©2020 IEEE 1215 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'},\n",
       " {'page_number': 1,\n",
       "  'page_char_count': 5203,\n",
       "  'page_word_count': 882,\n",
       "  'page_sentence_count_raw': 48,\n",
       "  'page_token_count': 1300.75,\n",
       "  'text': 'LSTM and GAN models to predict the stock prices on the Indian market and compare their performance. A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time. It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3]. The working of LSTM can be summarised by the following set of equations. zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate. The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively. The σ and tanh functions are deﬁned in 2 and 3 respectively. σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [4], where image patches are generated from random noise using two networks trained simultaneously. It has shown very good results when applied to various tasks particularly image synthesizing. GAN belong to the family of generative models. GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution. The model can either generate samples from a distribution or represent an actual distribution. Traditionally, training models require to have a dataset with labels for all the data. However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information. The generative adversarial network comprises of two models, a Generator and a Dis- criminator. The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution. Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake. The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples. For an overview of the network see Fig. 1. 1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution. The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training. The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples. 2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution. The discriminator then produces the probability of the sample belonging to the real data distribution Pdata. The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator. 3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4. min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions. We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5). The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima. On other hand a local optima is relatively easier to ﬁnd but does not perform well. Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively. L is the cost function, η is the learning rate. The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate. It is the most important hyperparam- eter to tune to achieve good performance on the problem. In this tutorial, you will discover the learning III. METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE). To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020. There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day. The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'}]"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "def text_formatter(text: str) -> str:\n",
    "    # performs minor formatting on the text in doc.\n",
    "    cleaned_text = text.replace(\"\\n\", \" \").strip()\n",
    "    #potentially more text formatting functions can go here\n",
    "    return cleaned_text\n",
    "\n",
    "def open_and_read_pdf(pdf_path: str) -> list[dict]:\n",
    "    doc = fitz.open(pdf_path)\n",
    "    pages_and_texts = []\n",
    "    for page_number, page in tqdm(enumerate(doc)):\n",
    "        text = page.get_text()\n",
    "        text = text_formatter(text=text)\n",
    "        pages_and_texts.append({\"page_number\": page_number - 0,\n",
    "                                \"page_char_count\": len(text),\n",
    "                                \"page_word_count\": len(text.split(\" \")),\n",
    "                                \"page_sentence_count_raw\": len(text.split(\". \")),\n",
    "                                \"page_token_count\": len(text) / 4, # 1 token = ~ 4 characters\n",
    "                                \"text\": text\n",
    "                                })\n",
    "    return pages_and_texts\n",
    "\n",
    "pages_and_texts = open_and_read_pdf(pdf_path=pdf_path)\n",
    "pages_and_texts[:2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'page_number': 0,\n",
       "  'page_char_count': 5703,\n",
       "  'page_word_count': 876,\n",
       "  'page_sentence_count_raw': 35,\n",
       "  'page_token_count': 1425.75,\n",
       "  'text': 'Indian Stock Market Prediction using Deep Learning Ayan Maiti Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India ayan.182ma008@nitk.edu.in Pushparaj Shetty D Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India prajshetty@nitk.edu.in Abstract—In this paper, we predict the stock prices of ﬁve companies listed on India’s National Stock Exchange (NSE) using two models- the Long Short Term Memory (LSTM) model and the Generative Adversarial Network (GAN) model with LSTM as the generator and a simple dense neural network as the discriminant. Both models take the online published historical stock-price data as input and produce the prediction of the closing price for the next trading day. To emulate the thought process of a real trader, our implementation applies the technique of rolling segmentation for the partition of training and testing dataset to examine the effect of different interval partitions on the prediction performance. Index Terms—LSTM, GAN, generator, discriminant, neural networks, technical indicators, rolling segmentation I. INTRODUCTION The prediction of stock prices and the direction of the price movement is a widely studied topic in many ﬁelds including trading, ﬁnance, statistics, and computer science. Typically, the professional stock market traders use funda- mental and/or technical analysis of stocks to analyze the market elements/indices to make investment decisions [1]. Fundamental analysis is the conventional method involving the study of company fundamentals such as sales, revenue and expenses, net proﬁt, debts, annual growth rates, and so on, whereas technical analysis, is exclusively based on the study of historical price movements. Practitioners of technical analysis refer to the price charts to study price patterns and use price data in different calculations to forecast future price movements. The technical analysis paradigm assumes, there exists an inherent correlation between the performance of the company and its stock price and therefore, it is used to determine the timing of buying and selling the shares of the company. Time-series prediction problems like the prediction of stock prices are a difﬁcult type of predictive modeling prob- lem. Such problems usually involve the presence of sequence dependence among the input variables [2]. A robust type of neural network equipped to handle the sequence dependence is the Recurrent Neural Networks (RNNs). The Long Short- Term Memory network or LSTM network is a type of RNN and is used extensively in time-series problems because of its capability to handle dependency between both long and short distance sequences [3]. Generative adversarial network (GAN) has been successfully applied to a wide range of ﬁelds, such as image in painting, semantic segmentation, and video prediction. GAN consists of a discriminative network D, which learns to distinguish if a given data instance is real or not, and a generative network G, which learns to confuse D by generating high-quality data [4]. This approach has also been used for the prediction of the high-frequency stock prices of the Shanghai and Shenzhen stock exchanges of the Chinese stock market [5]. However, it has not been used to predict the daily stock prices of the Indian stock market as per our knowledge. Our contributions in this paper are as follows: 1. Im- plemented two models for daily frequency stock price prediction- LSTM and GAN 2. Performed rolling segmen- tation on the training set and testing set of the raw data to investigate the effect of the model parameter update cycle on stock forecast performance. We found that LSTM outperformed GAN signiﬁcantly with an error of 0.074 compared to 0.32 obtained by GAN. It implies that the price predicted by LSTM differs from the original price by 7.4% whereas for GAN it is 32%. II. RELATED WORKS & BACKGROUND In the recent times, deep learning is emerging as a promis- ing area of research and has been applied for classiﬁcation and prediction tasks in various domains like text classiﬁca- tion, image processing, speech recognition, and many others. It has also been applied in the prediction of time-series since it is suitable to address chaotic data, randomness, and non- linearity [1]. For example, Chong et al. [6] have utilized the capability of deep neural networks to extract abstract features from data by using a deep feature learning-based stock market prediction model, which extract information from the stock return time series without relying on prior knowledge of the predictors and tested it on high-frequency data from the Korean stock market. Lately, LSTM network, which is suitable for learning temporal patterns, is extensively utilized for various tasks of time-series analyses [7]. LSTM is preferred over the conventional RNN as it overcomes the problem of vanishing (or exploding) gradients and as it can effectively learn long- term dependencies through memory cells and gates. Thus, many studies on ﬁnancial time-series modeling are conducted using LSTMs. However, Zhou et al [5] have adopted GAN to predict stock prices on highly volatile stocks in the China Stock Exchange market. GAN has not been applied much to ﬁnancial markets across the globe and certainly not on the Indian Stock Market to the best of our knowledge. In this work, we implement both 2020 IEEE REGION 10 CONFERENCE (TENCON) Osaka, Japan, November 16-19, 2020 978-1-7281-8455-5/20/$31.00 ©2020 IEEE 1215 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'},\n",
       " {'page_number': 2,\n",
       "  'page_char_count': 4107,\n",
       "  'page_word_count': 738,\n",
       "  'page_sentence_count_raw': 39,\n",
       "  'page_token_count': 1026.75,\n",
       "  'text': 'Fig. 1. A Generative Adversarial Network ’Closing Price’ and ’Trading Volume’. Also, we selected 9 additional technical indicators as feature subsets after a rigorous literature review. Many investors in the stock market study these technical indicators derived from the past stock-price data and use it as the signal for future market trends [1] [5]. So, the input data X at each day (say, XT ) consists the total of 13 basic indicators as listed in Table I. TABLE I BASIC INDICATORS FOR PREDICTION Technical Indicators Opening Price Maximum Price Minimum Price Trading Volume Moving Averages Bollinger Bands Exponential Moving Averages PPSR On Balance Volume Price Rate of Change Average Directional Movement Index Relative Strength Index MACD B. Data Partition Traditionally, data is split for training and testing the models using ﬁxed partitioning method. Nevertheless, the trading pattern of the stock market changes regularly, for example, investors at times favour stocks with high volatility whereas sometimes choose to invest in technology stocks. Accordingly, we should update the model parameters fre- quently to adjust to the change of market pattern. For making our experiments closer to real transactions, we carry out rolling segmentation on the training set and testing set of the experimental data. This technique of data partitioning was used in [5]. As Figure 2 shows, at the start, we select the ﬁrst M days as the training set and the next N days as the testing set. For the next round, we roll the time window ahead for N days, that is, choosing the (N + 1)th day to the (M + N)th day as the training set and the (M + N + 1)th day to the (M + 2N)th day as the testing set. We repeat this until all the data has been experimented. In other words, this N can be regarded as the model update cycle, and M is the size of the corresponding training data. C. Lookback Days Let Xi denote the set of basic technical indicators and Yi represent the closing price of one stock for a 1-day interval at time i (i = 1, 2, ..., T), where T is the maximum lag of time (lookback days). Given the historical basic indicators data X (X = X1, X2, ..., XT ) and the past closing price Y (Y = Y1, Y2, ..., YT ), we try to predict the closing price YT +1 for the next 1-day time interval. For this project, we have chosen T as 90 and 210 for the LSTM model and 45 for both LSTM and GAN models based on computational time and limitations. We have tested the LSTM model with T=210 for smaller values of M and N (refer section III-B) and with T=90 for all values of M and N. GAN, model was tested only on T=45 for all values of M and N. Fig. 2. Rolling segmentation on dataset. The green bar shows the entire dataset, the blue bar shows the training set for one round experiment, and the yellow bar is the corresponding testing set. Image Source- [5] D. LSTM Model Architecture We have used a single layer of LSTM internal units followed by a Dropout layer. The activation function in LSTM is tanh and the loss function used is Mean Squared Error. The values for the hyper-parameters are mentioned in Table II. E. GAN Model Architecture As discussed earlier a GAN is a network of 2 models- generator G, and discriminator D. Here, we used the loss functions deﬁned and used by Zhou et al. [5]. a) Generator: We have used a layer of LSTM followed by a dropout layer as our generator. To make the discrimina- tive model D as “confused” as possible, so that D does not distinguish between actual and closing price, the generative model G should have an adversarial loss which should be minimized such that D will not discriminate the prediction correctly. Consequently, we label actual closing price ¯ Y as class 1 and predicted closing price ˆ Y as class 0, and deﬁne the adversarial loss for G as 6 LG adv( ˆ Y ) = Lsce(D( ˆ Y ), 1) (6) where, Lsce is the sigmoid cross-entropy loss, deﬁned as 7 Lsce(A, B) = − X i Bi log(sigmoid(Ai)) + (1 −Bi) log(1 −sigmoid(Ai)) (7) 1217 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'},\n",
       " {'page_number': 5,\n",
       "  'page_char_count': 2804,\n",
       "  'page_word_count': 446,\n",
       "  'page_sentence_count_raw': 70,\n",
       "  'page_token_count': 701.0,\n",
       "  'text': 'V. CONCLUSION The results of the model update cycles on the predictive performance show that M=1200 and N=300 produce the best results of RMSRE=0.073884, obtained at look-back of 90 days. In other words, the predicted price had an average error difference of 7.3884% with respect to the actual price. However, the corresponding DPA was mere 0.505893, implying the model is not able to predict the direction of the price movement. We also found that increase in the number of look-back days did not have signiﬁcant improvement of results on LSTM model. As we suspect that the models have not achieved optimum training, we present some ideas below to reﬁne and continue our work forward: 1) The GAN model can be experimented on higher num- ber of look-back days and experiment by changing the hyper-parameters to improve the performance. 2) Experiment by incorporating the loss function for di- rection accuracy used in GAN model into LSTM to check for any improvement in DPA. Also, new loss function can be designed to achieve better results. 3) We have used stock price data of daily frequency. Higher frequency data of every minute stock price can be obtained to train and test the performance of the models implemented in this project. ACKNOWLEDGMENT The authors would like to thank National Institute of Technology Karnataka, India, for the support during this work. REFERENCES [1] I. K. Nti, A. F. Adekoya, and B. A. Weyori, “A systematic review of fundamental and technical analysis of stock market predictions,” Artiﬁcial Intelligence Review, pp. 1–51, 2019. [2] R. J. Frank, N. Davey, and S. P. Hunt, “Time series prediction and neural networks,” Journal of intelligent and robotic systems, vol. 31, no. 1-3, pp. 91–103, 2001. [3] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997. [4] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” in Advances in neural information processing systems, 2014, pp. 2672– 2680. [5] X. Zhou, Z. Pan, G. Hu, S. Tang, and C. Zhao, “Stock market prediction on high-frequency data using generative adversarial nets,” Mathematical Problems in Engineering, vol. 2018, 2018. [6] E. Chong, C. Han, and F. C. Park, “Deep learning networks for stock market analysis and prediction: Methodology, data representations, and case studies,” Expert Systems with Applications, vol. 83, pp. 187–205, 2017. [7] D. H. D. Nguyen, L. P. Tran, and V. Nguyen, “Predicting stock prices using dynamic lstm models,” in International Conference on Applied Informatics. Springer, 2019, pp. 199–212. 1220 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'}]"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import random\n",
    "random.sample(pages_and_texts, k=3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>page_char_count</th>\n",
       "      <th>page_word_count</th>\n",
       "      <th>page_sentence_count_raw</th>\n",
       "      <th>page_token_count</th>\n",
       "      <th>text</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>5703</td>\n",
       "      <td>876</td>\n",
       "      <td>35</td>\n",
       "      <td>1425.75</td>\n",
       "      <td>Indian Stock Market Prediction using Deep Lear...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>5203</td>\n",
       "      <td>882</td>\n",
       "      <td>48</td>\n",
       "      <td>1300.75</td>\n",
       "      <td>LSTM and GAN models to predict the stock price...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>4107</td>\n",
       "      <td>738</td>\n",
       "      <td>39</td>\n",
       "      <td>1026.75</td>\n",
       "      <td>Fig. 1. A Generative Adversarial Network ’Clos...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>4327</td>\n",
       "      <td>836</td>\n",
       "      <td>40</td>\n",
       "      <td>1081.75</td>\n",
       "      <td>TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>1779</td>\n",
       "      <td>282</td>\n",
       "      <td>15</td>\n",
       "      <td>444.75</td>\n",
       "      <td>TABLE III HYPER-PARAMETER VALUES FOR GENERATOR...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   page_number  page_char_count  page_word_count  page_sentence_count_raw  \\\n",
       "0            0             5703              876                       35   \n",
       "1            1             5203              882                       48   \n",
       "2            2             4107              738                       39   \n",
       "3            3             4327              836                       40   \n",
       "4            4             1779              282                       15   \n",
       "\n",
       "   page_token_count                                               text  \n",
       "0           1425.75  Indian Stock Market Prediction using Deep Lear...  \n",
       "1           1300.75  LSTM and GAN models to predict the stock price...  \n",
       "2           1026.75  Fig. 1. A Generative Adversarial Network ’Clos...  \n",
       "3           1081.75  TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL...  \n",
       "4            444.75  TABLE III HYPER-PARAMETER VALUES FOR GENERATOR...  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd \n",
    "\n",
    "df = pd.DataFrame(pages_and_texts)\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>page_char_count</th>\n",
       "      <th>page_word_count</th>\n",
       "      <th>page_sentence_count_raw</th>\n",
       "      <th>page_token_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.50</td>\n",
       "      <td>3987.17</td>\n",
       "      <td>676.67</td>\n",
       "      <td>41.17</td>\n",
       "      <td>996.79</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.87</td>\n",
       "      <td>1471.60</td>\n",
       "      <td>252.99</td>\n",
       "      <td>17.93</td>\n",
       "      <td>367.90</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.00</td>\n",
       "      <td>1779.00</td>\n",
       "      <td>282.00</td>\n",
       "      <td>15.00</td>\n",
       "      <td>444.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1.25</td>\n",
       "      <td>3129.75</td>\n",
       "      <td>519.00</td>\n",
       "      <td>36.00</td>\n",
       "      <td>782.44</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>2.50</td>\n",
       "      <td>4217.00</td>\n",
       "      <td>787.00</td>\n",
       "      <td>39.50</td>\n",
       "      <td>1054.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.75</td>\n",
       "      <td>4984.00</td>\n",
       "      <td>866.00</td>\n",
       "      <td>46.00</td>\n",
       "      <td>1246.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00</td>\n",
       "      <td>5703.00</td>\n",
       "      <td>882.00</td>\n",
       "      <td>70.00</td>\n",
       "      <td>1425.75</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       page_number  page_char_count  page_word_count  page_sentence_count_raw  \\\n",
       "count         6.00             6.00             6.00                     6.00   \n",
       "mean          2.50          3987.17           676.67                    41.17   \n",
       "std           1.87          1471.60           252.99                    17.93   \n",
       "min           0.00          1779.00           282.00                    15.00   \n",
       "25%           1.25          3129.75           519.00                    36.00   \n",
       "50%           2.50          4217.00           787.00                    39.50   \n",
       "75%           3.75          4984.00           866.00                    46.00   \n",
       "max           5.00          5703.00           882.00                    70.00   \n",
       "\n",
       "       page_token_count  \n",
       "count              6.00  \n",
       "mean             996.79  \n",
       "std              367.90  \n",
       "min              444.75  \n",
       "25%              782.44  \n",
       "50%             1054.25  \n",
       "75%             1246.00  \n",
       "max             1425.75  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# why would we care about token count?\n",
    "# token count is important because:\n",
    "# 1. embedding models don't deal with infinte tokens\n",
    "# 2. llms don't deal with infinite tokens."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# further text processing (splitting pages into sentences.)\n",
    "# two ways to do this\n",
    "# 1. we've done this with splitting on \". \".\n",
    "# 2. we can do this with a nlp library such as spaCy and nltk"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[This is a sentecnce., This is another sentence., I like elephants.]"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from spacy.lang.en import English \n",
    "nlp = English()\n",
    "\n",
    "# add a sentencizer pipeline\n",
    "nlp.add_pipe(\"sentencizer\")\n",
    "\n",
    "# create document instance as an example \n",
    "doc = nlp(\"This is a sentecnce. This is another sentence. I like elephants.\")\n",
    "assert len(list(doc.sents)) == 3\n",
    "\n",
    "# print out our sentences split \n",
    "list(doc.sents)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'page_number': 5,\n",
       " 'page_char_count': 2804,\n",
       " 'page_word_count': 446,\n",
       " 'page_sentence_count_raw': 70,\n",
       " 'page_token_count': 701.0,\n",
       " 'text': 'V. CONCLUSION The results of the model update cycles on the predictive performance show that M=1200 and N=300 produce the best results of RMSRE=0.073884, obtained at look-back of 90 days. In other words, the predicted price had an average error difference of 7.3884% with respect to the actual price. However, the corresponding DPA was mere 0.505893, implying the model is not able to predict the direction of the price movement. We also found that increase in the number of look-back days did not have signiﬁcant improvement of results on LSTM model. As we suspect that the models have not achieved optimum training, we present some ideas below to reﬁne and continue our work forward: 1) The GAN model can be experimented on higher num- ber of look-back days and experiment by changing the hyper-parameters to improve the performance. 2) Experiment by incorporating the loss function for di- rection accuracy used in GAN model into LSTM to check for any improvement in DPA. Also, new loss function can be designed to achieve better results. 3) We have used stock price data of daily frequency. Higher frequency data of every minute stock price can be obtained to train and test the performance of the models implemented in this project. ACKNOWLEDGMENT The authors would like to thank National Institute of Technology Karnataka, India, for the support during this work. REFERENCES [1] I. K. Nti, A. F. Adekoya, and B. A. Weyori, “A systematic review of fundamental and technical analysis of stock market predictions,” Artiﬁcial Intelligence Review, pp. 1–51, 2019. [2] R. J. Frank, N. Davey, and S. P. Hunt, “Time series prediction and neural networks,” Journal of intelligent and robotic systems, vol. 31, no. 1-3, pp. 91–103, 2001. [3] S. Hochreiter and J. Schmidhuber, “Long short-term memory,” Neural computation, vol. 9, no. 8, pp. 1735–1780, 1997. [4] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, and Y. Bengio, “Generative adversarial nets,” in Advances in neural information processing systems, 2014, pp. 2672– 2680. [5] X. Zhou, Z. Pan, G. Hu, S. Tang, and C. Zhao, “Stock market prediction on high-frequency data using generative adversarial nets,” Mathematical Problems in Engineering, vol. 2018, 2018. [6] E. Chong, C. Han, and F. C. Park, “Deep learning networks for stock market analysis and prediction: Methodology, data representations, and case studies,” Expert Systems with Applications, vol. 83, pp. 187–205, 2017. [7] D. H. D. Nguyen, L. P. Tran, and V. Nguyen, “Predicting stock prices using dynamic lstm models,” in International Conference on Applied Informatics. Springer, 2019, pp. 199–212. 1220 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.'}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pages_and_texts[5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d19d8f24bfb34b769a36a72dad83d074",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/6 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for item in tqdm(pages_and_texts):\n",
    "    item[\"sentences\"] = list(nlp(item[\"text\"]).sents)\n",
    "\n",
    "    # make sure all sentences are strings (the default type is a spaCy datatype)\n",
    "    item[\"sentences\"] = [str(sentence) for sentence in item['sentences']]\n",
    "\n",
    "    # count the sentences \n",
    "    item[\"page_sentence_count_spacy\"] = len(item[\"sentences\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'page_number': 1,\n",
       "  'page_char_count': 5203,\n",
       "  'page_word_count': 882,\n",
       "  'page_sentence_count_raw': 48,\n",
       "  'page_token_count': 1300.75,\n",
       "  'text': 'LSTM and GAN models to predict the stock prices on the Indian market and compare their performance. A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time. It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3]. The working of LSTM can be summarised by the following set of equations. zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate. The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively. The σ and tanh functions are deﬁned in 2 and 3 respectively. σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [4], where image patches are generated from random noise using two networks trained simultaneously. It has shown very good results when applied to various tasks particularly image synthesizing. GAN belong to the family of generative models. GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution. The model can either generate samples from a distribution or represent an actual distribution. Traditionally, training models require to have a dataset with labels for all the data. However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information. The generative adversarial network comprises of two models, a Generator and a Dis- criminator. The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution. Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake. The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples. For an overview of the network see Fig. 1. 1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution. The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training. The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples. 2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution. The discriminator then produces the probability of the sample belonging to the real data distribution Pdata. The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator. 3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4. min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions. We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5). The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima. On other hand a local optima is relatively easier to ﬁnd but does not perform well. Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively. L is the cost function, η is the learning rate. The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate. It is the most important hyperparam- eter to tune to achieve good performance on the problem. In this tutorial, you will discover the learning III. METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE). To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020. There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day. The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.',\n",
       "  'sentences': ['LSTM and GAN models to predict the stock prices on the Indian market and compare their performance.',\n",
       "   'A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time.',\n",
       "   'It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3].',\n",
       "   'The working of LSTM can be summarised by the following set of equations.',\n",
       "   'zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate.',\n",
       "   'The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively.',\n",
       "   'The σ and tanh functions are deﬁned in 2 and 3 respectively.',\n",
       "   'σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [',\n",
       "   '4], where image patches are generated from random noise using two networks trained simultaneously.',\n",
       "   'It has shown very good results when applied to various tasks particularly image synthesizing.',\n",
       "   'GAN belong to the family of generative models.',\n",
       "   'GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution.',\n",
       "   'The model can either generate samples from a distribution or represent an actual distribution.',\n",
       "   'Traditionally, training models require to have a dataset with labels for all the data.',\n",
       "   'However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information.',\n",
       "   'The generative adversarial network comprises of two models, a Generator and a Dis- criminator.',\n",
       "   'The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution.',\n",
       "   'Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake.',\n",
       "   'The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples.',\n",
       "   'For an overview of the network see Fig.',\n",
       "   '1.',\n",
       "   '1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution.',\n",
       "   'The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training.',\n",
       "   'The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples.',\n",
       "   '2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution.',\n",
       "   'The discriminator then produces the probability of the sample belonging to the real data distribution Pdata.',\n",
       "   'The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator.',\n",
       "   '3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4.',\n",
       "   'min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions.',\n",
       "   'We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5).',\n",
       "   'The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima.',\n",
       "   'On other hand a local optima is relatively easier to ﬁnd but does not perform well.',\n",
       "   'Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively.',\n",
       "   'L is the cost function, η is the learning rate.',\n",
       "   'The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate.',\n",
       "   'It is the most important hyperparam- eter to tune to achieve good performance on the problem.',\n",
       "   'In this tutorial, you will discover the learning III.',\n",
       "   'METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE).',\n",
       "   'To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020.',\n",
       "   'There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day.',\n",
       "   'The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University.',\n",
       "   'Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.',\n",
       "   ' Restrictions apply.'],\n",
       "  'page_sentence_count_spacy': 43}]"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random.sample(pages_and_texts, k=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>page_char_count</th>\n",
       "      <th>page_word_count</th>\n",
       "      <th>page_sentence_count_raw</th>\n",
       "      <th>page_token_count</th>\n",
       "      <th>page_sentence_count_spacy</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.50</td>\n",
       "      <td>3987.17</td>\n",
       "      <td>676.67</td>\n",
       "      <td>41.17</td>\n",
       "      <td>996.79</td>\n",
       "      <td>31.83</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.87</td>\n",
       "      <td>1471.60</td>\n",
       "      <td>252.99</td>\n",
       "      <td>17.93</td>\n",
       "      <td>367.90</td>\n",
       "      <td>9.22</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.00</td>\n",
       "      <td>1779.00</td>\n",
       "      <td>282.00</td>\n",
       "      <td>15.00</td>\n",
       "      <td>444.75</td>\n",
       "      <td>15.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1.25</td>\n",
       "      <td>3129.75</td>\n",
       "      <td>519.00</td>\n",
       "      <td>36.00</td>\n",
       "      <td>782.44</td>\n",
       "      <td>31.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>2.50</td>\n",
       "      <td>4217.00</td>\n",
       "      <td>787.00</td>\n",
       "      <td>39.50</td>\n",
       "      <td>1054.25</td>\n",
       "      <td>33.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.75</td>\n",
       "      <td>4984.00</td>\n",
       "      <td>866.00</td>\n",
       "      <td>46.00</td>\n",
       "      <td>1246.00</td>\n",
       "      <td>34.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00</td>\n",
       "      <td>5703.00</td>\n",
       "      <td>882.00</td>\n",
       "      <td>70.00</td>\n",
       "      <td>1425.75</td>\n",
       "      <td>43.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       page_number  page_char_count  page_word_count  page_sentence_count_raw  \\\n",
       "count         6.00             6.00             6.00                     6.00   \n",
       "mean          2.50          3987.17           676.67                    41.17   \n",
       "std           1.87          1471.60           252.99                    17.93   \n",
       "min           0.00          1779.00           282.00                    15.00   \n",
       "25%           1.25          3129.75           519.00                    36.00   \n",
       "50%           2.50          4217.00           787.00                    39.50   \n",
       "75%           3.75          4984.00           866.00                    46.00   \n",
       "max           5.00          5703.00           882.00                    70.00   \n",
       "\n",
       "       page_token_count  page_sentence_count_spacy  \n",
       "count              6.00                       6.00  \n",
       "mean             996.79                      31.83  \n",
       "std              367.90                       9.22  \n",
       "min              444.75                      15.00  \n",
       "25%              782.44                      31.50  \n",
       "50%             1054.25                      33.50  \n",
       "75%             1246.00                      34.75  \n",
       "max             1425.75                      43.00  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(pages_and_texts)\n",
    "df.describe().round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# chunking our sentences together\n",
    "\n",
    "# why? \n",
    "# easier to manage similar sized chunks of text.\n",
    "# dont overload the embedding models capacity for tokens\n",
    "# our llm context window (the amount of tokens an llm can take in) may be limited and requires computing power so we want to make sure were using it as well as possible."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6136d934654e43f6af08c4c38ba955db",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/6 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# define split size to turn groups of sentences into chunks.\n",
    "num_sentence_chunk_size = 10\n",
    "\n",
    "# create a function that recursively splits a list into desired sizes.\n",
    "def split_list(input_list: list,\n",
    "               slice_size: int) -> list[list[str]]:\n",
    "    \"\"\"\n",
    "    splits the input_list into sublists of size slice_size (or as close as possible)\n",
    "    for example, a list of 17 sentences would be split into two lists of [[10], [7]]\n",
    "    \"\"\"\n",
    "\n",
    "    return [input_list[i:i + slice_size] for i in range (0, len(input_list), slice_size)]\n",
    "\n",
    "    # the return type returns a list of lists which contain ten sentences each, the for loop is run through the range of zero to \n",
    "    # length of the input list (total no. of sentences in a page) and then we set a slice size also, which makes sure \n",
    "    # to pick the next sentence i after the slice_size index (basically splits the input_list). \n",
    "    # input_list[i:i+slice_size] returns a sublist from the input_list, which contains the 'i' (current sentence)\n",
    "    # until i+slice_size (10 sentences after i).\n",
    "    # and such chunks \n",
    "\n",
    "# loop through pages and texts and split sentences into chunks\n",
    "for item in tqdm(pages_and_texts):\n",
    "    item[\"sentence_chunks\"] = split_list(input_list=item[\"sentences\"], slice_size=num_sentence_chunk_size)\n",
    "    item[\"num_chunks\"] = len(item[\"sentence_chunks\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'page_number': 1,\n",
       "  'page_char_count': 5203,\n",
       "  'page_word_count': 882,\n",
       "  'page_sentence_count_raw': 48,\n",
       "  'page_token_count': 1300.75,\n",
       "  'text': 'LSTM and GAN models to predict the stock prices on the Indian market and compare their performance. A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time. It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3]. The working of LSTM can be summarised by the following set of equations. zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate. The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively. The σ and tanh functions are deﬁned in 2 and 3 respectively. σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [4], where image patches are generated from random noise using two networks trained simultaneously. It has shown very good results when applied to various tasks particularly image synthesizing. GAN belong to the family of generative models. GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution. The model can either generate samples from a distribution or represent an actual distribution. Traditionally, training models require to have a dataset with labels for all the data. However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information. The generative adversarial network comprises of two models, a Generator and a Dis- criminator. The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution. Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake. The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples. For an overview of the network see Fig. 1. 1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution. The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training. The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples. 2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution. The discriminator then produces the probability of the sample belonging to the real data distribution Pdata. The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator. 3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4. min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions. We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5). The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima. On other hand a local optima is relatively easier to ﬁnd but does not perform well. Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively. L is the cost function, η is the learning rate. The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate. It is the most important hyperparam- eter to tune to achieve good performance on the problem. In this tutorial, you will discover the learning III. METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE). To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020. There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day. The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.  Restrictions apply.',\n",
       "  'sentences': ['LSTM and GAN models to predict the stock prices on the Indian market and compare their performance.',\n",
       "   'A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time.',\n",
       "   'It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3].',\n",
       "   'The working of LSTM can be summarised by the following set of equations.',\n",
       "   'zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate.',\n",
       "   'The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively.',\n",
       "   'The σ and tanh functions are deﬁned in 2 and 3 respectively.',\n",
       "   'σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [',\n",
       "   '4], where image patches are generated from random noise using two networks trained simultaneously.',\n",
       "   'It has shown very good results when applied to various tasks particularly image synthesizing.',\n",
       "   'GAN belong to the family of generative models.',\n",
       "   'GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution.',\n",
       "   'The model can either generate samples from a distribution or represent an actual distribution.',\n",
       "   'Traditionally, training models require to have a dataset with labels for all the data.',\n",
       "   'However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information.',\n",
       "   'The generative adversarial network comprises of two models, a Generator and a Dis- criminator.',\n",
       "   'The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution.',\n",
       "   'Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake.',\n",
       "   'The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples.',\n",
       "   'For an overview of the network see Fig.',\n",
       "   '1.',\n",
       "   '1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution.',\n",
       "   'The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training.',\n",
       "   'The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples.',\n",
       "   '2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution.',\n",
       "   'The discriminator then produces the probability of the sample belonging to the real data distribution Pdata.',\n",
       "   'The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator.',\n",
       "   '3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4.',\n",
       "   'min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions.',\n",
       "   'We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5).',\n",
       "   'The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima.',\n",
       "   'On other hand a local optima is relatively easier to ﬁnd but does not perform well.',\n",
       "   'Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively.',\n",
       "   'L is the cost function, η is the learning rate.',\n",
       "   'The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate.',\n",
       "   'It is the most important hyperparam- eter to tune to achieve good performance on the problem.',\n",
       "   'In this tutorial, you will discover the learning III.',\n",
       "   'METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE).',\n",
       "   'To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020.',\n",
       "   'There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day.',\n",
       "   'The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University.',\n",
       "   'Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.',\n",
       "   ' Restrictions apply.'],\n",
       "  'page_sentence_count_spacy': 43,\n",
       "  'sentence_chunks': [['LSTM and GAN models to predict the stock prices on the Indian market and compare their performance.',\n",
       "    'A. LSTM LSTM can be trained to learn dependencies ranging over very long time intervals of time.',\n",
       "    'It beats the vanishing gradients problem faced by a general RNN by substituting the ordinary neuron by a complex architecture called the LSTM unit or block [3].',\n",
       "    'The working of LSTM can be summarised by the following set of equations.',\n",
       "    'zt = tanh(W zxt + Rzht−1 + bz) it = σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo) st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate and ot denotes the output gate.',\n",
       "    'The forget gate, memory cell, and hidden state are denoted by ft, st, and ht, respectively.',\n",
       "    'The σ and tanh functions are deﬁned in 2 and 3 respectively.',\n",
       "    'σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B. GAN Generative adversarial networks was introduced by Good- fellow et al. [',\n",
       "    '4], where image patches are generated from random noise using two networks trained simultaneously.',\n",
       "    'It has shown very good results when applied to various tasks particularly image synthesizing.'],\n",
       "   ['GAN belong to the family of generative models.',\n",
       "    'GAN models learn the given set of distribution (training data) and produce data similar to the original data distribution.',\n",
       "    'The model can either generate samples from a distribution or represent an actual distribution.',\n",
       "    'Traditionally, training models require to have a dataset with labels for all the data.',\n",
       "    'However, generative models can be trained with a dataset holding a mixture of both unlabeled and labeled data, thus eliminating the constraint caused by the unavailability of information.',\n",
       "    'The generative adversarial network comprises of two models, a Generator and a Dis- criminator.',\n",
       "    'The purpose of the generator is to take noise as input, which is usually a uniform distribution, and generate samples that look like they were drawn from a distribution representing our target distribution.',\n",
       "    'Samples are then drawn from a distribution representing our target distribution (real data), and the samples produced by our Generator (fake data) are fed into the Discriminator which then predicts which ones are real and which ones are fake.',\n",
       "    'The output produced by the discriminator is then propagated to the generator to update, improve, and generate more realistic samples.',\n",
       "    'For an overview of the network see Fig.'],\n",
       "   ['1.',\n",
       "    '1) Generator: We can deﬁne the Generator function as a function G that is differentiable with respect to its parameters θG and input data z. z is considered as the noise drawn from a distribution Pz, e.g. Gaussian distribution.',\n",
       "    'The generator will then map z to samples G(z) that form the generator’s distribution Pg which can be expected to be the same as the true data distribution Pdata after training.',\n",
       "    'The mapped samples G(z) are then put into the discriminator and the generator is then trained to deceive the discriminator by making it give high probabilities to the generated samples.',\n",
       "    '2) Discriminator: Similarly, we can deﬁne the discrimi- nator function D differentiable with respect to its parameters θD and input x and G(z), where x is real data samples drawn from the Pdata distribution.',\n",
       "    'The discriminator then produces the probability of the sample belonging to the real data distribution Pdata.',\n",
       "    'The discriminator will then be trained continuously, in a supervised manner, so that it gives high probability to real data samples and a low probability to the samples generated by the Generator.',\n",
       "    '3) Cost Function: Mathematically, the generator and discriminator cost functions are related by 4.',\n",
       "    'min G max D (EX∼Pdata(x)[logD(x)] + EZ∼Pdata(z)[log(1 −D(G(z)))]) (4) where EX∼Pdata(x) and EZ∼Pdata(z) are the expected values of the Pdata and Pz distributions.',\n",
       "    'We can apply Stochastic Gradient Descent (SGD) to solve this optimization problem (see 5).'],\n",
       "   ['The optimization is difﬁcult and the space of solutions (sets of weights) should contain global optima.',\n",
       "    'On other hand a local optima is relatively easier to ﬁnd but does not perform well.',\n",
       "    'Wt+1 = Wt −η ∂L ∂W (5) where W is the set of weights and Wt+1 and Wt are set of weights at iterations t + 1 and t respectively.',\n",
       "    'L is the cost function, η is the learning rate.',\n",
       "    'The amount by which the set of weights in the model change to reach the space of solutions, or the step size, is called the learning rate.',\n",
       "    'It is the most important hyperparam- eter to tune to achieve good performance on the problem.',\n",
       "    'In this tutorial, you will discover the learning III.',\n",
       "    'METHODOLOGY A. Dataset We obtained historical data of 5 stocks from Yahoo Finance (https://in.ﬁnance.yahoo.com/) that are being actively traded on India’s National Stock Exchange (NSE).',\n",
       "    'To maximize the available data-points, the stocks were selected such that their historical data of range from January 1, 1996 to May 15, 2020.',\n",
       "    'There are 253 trading days in a year for the NSE, so on an average we get 5500 data-points for every stock at an interval of 1 day.'],\n",
       "   ['The historical price data contains 5 features- ’Open Price’, ’Maximum Price’, ’Minimum Price’, 1216 Authorized licensed use limited to: Carleton University.',\n",
       "    'Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.',\n",
       "    ' Restrictions apply.']],\n",
       "  'num_chunks': 5}]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# sample an example from the group (note: many samples have only 1 chunk as they have <=10 sentences total.)\n",
    "random.sample(pages_and_texts, k=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>page_char_count</th>\n",
       "      <th>page_word_count</th>\n",
       "      <th>page_sentence_count_raw</th>\n",
       "      <th>page_token_count</th>\n",
       "      <th>page_sentence_count_spacy</th>\n",
       "      <th>num_chunks</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "      <td>6.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.50</td>\n",
       "      <td>3987.17</td>\n",
       "      <td>676.67</td>\n",
       "      <td>41.17</td>\n",
       "      <td>996.79</td>\n",
       "      <td>31.83</td>\n",
       "      <td>3.83</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.87</td>\n",
       "      <td>1471.60</td>\n",
       "      <td>252.99</td>\n",
       "      <td>17.93</td>\n",
       "      <td>367.90</td>\n",
       "      <td>9.22</td>\n",
       "      <td>0.98</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.00</td>\n",
       "      <td>1779.00</td>\n",
       "      <td>282.00</td>\n",
       "      <td>15.00</td>\n",
       "      <td>444.75</td>\n",
       "      <td>15.00</td>\n",
       "      <td>2.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1.25</td>\n",
       "      <td>3129.75</td>\n",
       "      <td>519.00</td>\n",
       "      <td>36.00</td>\n",
       "      <td>782.44</td>\n",
       "      <td>31.50</td>\n",
       "      <td>4.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>2.50</td>\n",
       "      <td>4217.00</td>\n",
       "      <td>787.00</td>\n",
       "      <td>39.50</td>\n",
       "      <td>1054.25</td>\n",
       "      <td>33.50</td>\n",
       "      <td>4.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.75</td>\n",
       "      <td>4984.00</td>\n",
       "      <td>866.00</td>\n",
       "      <td>46.00</td>\n",
       "      <td>1246.00</td>\n",
       "      <td>34.75</td>\n",
       "      <td>4.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00</td>\n",
       "      <td>5703.00</td>\n",
       "      <td>882.00</td>\n",
       "      <td>70.00</td>\n",
       "      <td>1425.75</td>\n",
       "      <td>43.00</td>\n",
       "      <td>5.00</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       page_number  page_char_count  page_word_count  page_sentence_count_raw  \\\n",
       "count         6.00             6.00             6.00                     6.00   \n",
       "mean          2.50          3987.17           676.67                    41.17   \n",
       "std           1.87          1471.60           252.99                    17.93   \n",
       "min           0.00          1779.00           282.00                    15.00   \n",
       "25%           1.25          3129.75           519.00                    36.00   \n",
       "50%           2.50          4217.00           787.00                    39.50   \n",
       "75%           3.75          4984.00           866.00                    46.00   \n",
       "max           5.00          5703.00           882.00                    70.00   \n",
       "\n",
       "       page_token_count  page_sentence_count_spacy  num_chunks  \n",
       "count              6.00                       6.00        6.00  \n",
       "mean             996.79                      31.83        3.83  \n",
       "std              367.90                       9.22        0.98  \n",
       "min              444.75                      15.00        2.00  \n",
       "25%              782.44                      31.50        4.00  \n",
       "50%             1054.25                      33.50        4.00  \n",
       "75%             1246.00                      34.75        4.00  \n",
       "max             1425.75                      43.00        5.00  "
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# create a dataframe to get stats\n",
    "df = pd.DataFrame(pages_and_texts)\n",
    "df.describe().round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# splitting each chunk into its own item \n",
    "\n",
    "# we would like to embed each chunk of sentences into its own numerical representation\n",
    "# that will give us good level of granularity.\n",
    "# meaning, we can dive specifically into the text sample that was used in our model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a0a12fdc4cb24ad5bd6aa6f29b3239fe",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/6 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "23"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import re\n",
    "\n",
    "# split each chunk into its own item\n",
    "\n",
    "pages_and_chunks = []\n",
    "\n",
    "for item in tqdm(pages_and_texts):\n",
    "    for sentence_chunk in item[\"sentence_chunks\"]:\n",
    "        chunk_dict = {}\n",
    "        chunk_dict[\"page_number\"] = item[\"page_number\"]\n",
    "\n",
    "        # join the sentences together into a paragraph-like structure, aka a chunk (so they are a single string)\n",
    "        joined_sentence_chunk = \"\".join(sentence_chunk).replace('  ', \" \").strip()\n",
    "        joined_sentence_chunk = re.sub(r'\\.([A-Z])', r'. \\1', joined_sentence_chunk) # \".A\" -> \". A\" for any full-stop/capital letter combo \n",
    "        chunk_dict[\"sentence_chunk\"] = joined_sentence_chunk\n",
    "\n",
    "        # get stats about the chunk\n",
    "        chunk_dict[\"chunk_char_count\"] = len(joined_sentence_chunk)\n",
    "        chunk_dict[\"chunk_word_count\"] = len([word for word in joined_sentence_chunk.split(' ')])\n",
    "        chunk_dict[\"chunk_token_count\"] = len(joined_sentence_chunk) / 4 # 1 token ~= 4 characters\n",
    "\n",
    "        pages_and_chunks.append(chunk_dict)\n",
    "\n",
    "len(pages_and_chunks)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'page_number': 4,\n",
       "  'sentence_chunk': 'N=300 N=300 M=400 M=700 LSTM 0.108374 0.081476 TABLE X SUMMARY OF DPA WITH DIFFERENT (M,N). THESE FIGURES ARE AVERAGE VALUE OF 5 STOCKS ON BASIS OF LAST 210 DAYS. N=300 N=300 M=400 M=700 LSTM 0.502477 0.500712 1219 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore. Restrictions apply.',\n",
       "  'chunk_char_count': 351,\n",
       "  'chunk_word_count': 56,\n",
       "  'chunk_token_count': 87.75}]"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "random.sample(pages_and_chunks, k=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we have broken our entire pdf into chunks of 10 sentences or less.\n",
    "# this means we could reference a chunk of text and know its source."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>sentence_chunk</th>\n",
       "      <th>chunk_char_count</th>\n",
       "      <th>chunk_word_count</th>\n",
       "      <th>chunk_token_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Indian Stock Market Prediction using Deep Lear...</td>\n",
       "      <td>2430</td>\n",
       "      <td>349</td>\n",
       "      <td>607.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>A robust type of neural network equipped to ha...</td>\n",
       "      <td>1398</td>\n",
       "      <td>231</td>\n",
       "      <td>349.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>It implies that the price predicted by LSTM di...</td>\n",
       "      <td>1442</td>\n",
       "      <td>229</td>\n",
       "      <td>360.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>GAN has not been applied much to ﬁnancial mark...</td>\n",
       "      <td>429</td>\n",
       "      <td>66</td>\n",
       "      <td>107.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>LSTM and GAN models to predict the stock price...</td>\n",
       "      <td>1131</td>\n",
       "      <td>214</td>\n",
       "      <td>282.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1</td>\n",
       "      <td>GAN belong to the family of generative models....</td>\n",
       "      <td>1258</td>\n",
       "      <td>197</td>\n",
       "      <td>314.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>1</td>\n",
       "      <td>1.1) Generator: We can deﬁne the Generator fun...</td>\n",
       "      <td>1456</td>\n",
       "      <td>230</td>\n",
       "      <td>364.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>1</td>\n",
       "      <td>The optimization is difﬁcult and the space of ...</td>\n",
       "      <td>1111</td>\n",
       "      <td>201</td>\n",
       "      <td>277.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>1</td>\n",
       "      <td>The historical price data contains 5 features-...</td>\n",
       "      <td>236</td>\n",
       "      <td>33</td>\n",
       "      <td>59.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>2</td>\n",
       "      <td>Fig.1. A Generative Adversarial Network ’Closi...</td>\n",
       "      <td>1368</td>\n",
       "      <td>210</td>\n",
       "      <td>342.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>2</td>\n",
       "      <td>As Figure 2 shows, at the start, we select the...</td>\n",
       "      <td>1304</td>\n",
       "      <td>270</td>\n",
       "      <td>326.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>2</td>\n",
       "      <td>The green bar shows the entire dataset, the bl...</td>\n",
       "      <td>1409</td>\n",
       "      <td>252</td>\n",
       "      <td>352.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>2</td>\n",
       "      <td>Restrictions apply.</td>\n",
       "      <td>19</td>\n",
       "      <td>2</td>\n",
       "      <td>4.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>3</td>\n",
       "      <td>TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL...</td>\n",
       "      <td>1993</td>\n",
       "      <td>394</td>\n",
       "      <td>498.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>3</td>\n",
       "      <td>2 (12) RMSRE is used as a yardstick for the pr...</td>\n",
       "      <td>1016</td>\n",
       "      <td>194</td>\n",
       "      <td>254.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>3</td>\n",
       "      <td>Tables V and VI show the average values of RMS...</td>\n",
       "      <td>871</td>\n",
       "      <td>166</td>\n",
       "      <td>217.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>3</td>\n",
       "      <td>C. 210 days Look-back We could test the model ...</td>\n",
       "      <td>439</td>\n",
       "      <td>77</td>\n",
       "      <td>109.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>4</td>\n",
       "      <td>TABLE III HYPER-PARAMETER VALUES FOR GENERATOR...</td>\n",
       "      <td>1426</td>\n",
       "      <td>225</td>\n",
       "      <td>356.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>4</td>\n",
       "      <td>N=300 N=300 M=400 M=700 LSTM 0.108374 0.081476...</td>\n",
       "      <td>351</td>\n",
       "      <td>56</td>\n",
       "      <td>87.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>5</td>\n",
       "      <td>V. CONCLUSION The results of the model update ...</td>\n",
       "      <td>1367</td>\n",
       "      <td>224</td>\n",
       "      <td>341.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>5</td>\n",
       "      <td>REFERENCES [1] I. K. Nti, A. F. Adekoya, and B...</td>\n",
       "      <td>478</td>\n",
       "      <td>71</td>\n",
       "      <td>119.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>5</td>\n",
       "      <td>4] I. Goodfellow, J. Pouget-Abadie, M. Mirza, ...</td>\n",
       "      <td>801</td>\n",
       "      <td>117</td>\n",
       "      <td>200.25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>5</td>\n",
       "      <td>1220 Authorized licensed use limited to: Carle...</td>\n",
       "      <td>141</td>\n",
       "      <td>20</td>\n",
       "      <td>35.25</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    page_number                                     sentence_chunk  \\\n",
       "0             0  Indian Stock Market Prediction using Deep Lear...   \n",
       "1             0  A robust type of neural network equipped to ha...   \n",
       "2             0  It implies that the price predicted by LSTM di...   \n",
       "3             0  GAN has not been applied much to ﬁnancial mark...   \n",
       "4             1  LSTM and GAN models to predict the stock price...   \n",
       "5             1  GAN belong to the family of generative models....   \n",
       "6             1  1.1) Generator: We can deﬁne the Generator fun...   \n",
       "7             1  The optimization is difﬁcult and the space of ...   \n",
       "8             1  The historical price data contains 5 features-...   \n",
       "9             2  Fig.1. A Generative Adversarial Network ’Closi...   \n",
       "10            2  As Figure 2 shows, at the start, we select the...   \n",
       "11            2  The green bar shows the entire dataset, the bl...   \n",
       "12            2                                Restrictions apply.   \n",
       "13            3  TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL...   \n",
       "14            3  2 (12) RMSRE is used as a yardstick for the pr...   \n",
       "15            3  Tables V and VI show the average values of RMS...   \n",
       "16            3  C. 210 days Look-back We could test the model ...   \n",
       "17            4  TABLE III HYPER-PARAMETER VALUES FOR GENERATOR...   \n",
       "18            4  N=300 N=300 M=400 M=700 LSTM 0.108374 0.081476...   \n",
       "19            5  V. CONCLUSION The results of the model update ...   \n",
       "20            5  REFERENCES [1] I. K. Nti, A. F. Adekoya, and B...   \n",
       "21            5  4] I. Goodfellow, J. Pouget-Abadie, M. Mirza, ...   \n",
       "22            5  1220 Authorized licensed use limited to: Carle...   \n",
       "\n",
       "    chunk_char_count  chunk_word_count  chunk_token_count  \n",
       "0               2430               349             607.50  \n",
       "1               1398               231             349.50  \n",
       "2               1442               229             360.50  \n",
       "3                429                66             107.25  \n",
       "4               1131               214             282.75  \n",
       "5               1258               197             314.50  \n",
       "6               1456               230             364.00  \n",
       "7               1111               201             277.75  \n",
       "8                236                33              59.00  \n",
       "9               1368               210             342.00  \n",
       "10              1304               270             326.00  \n",
       "11              1409               252             352.25  \n",
       "12                19                 2               4.75  \n",
       "13              1993               394             498.25  \n",
       "14              1016               194             254.00  \n",
       "15               871               166             217.75  \n",
       "16               439                77             109.75  \n",
       "17              1426               225             356.50  \n",
       "18               351                56              87.75  \n",
       "19              1367               224             341.75  \n",
       "20               478                71             119.50  \n",
       "21               801               117             200.25  \n",
       "22               141                20              35.25  "
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# get stats about our chunks\n",
    "df = pd.DataFrame(pages_and_chunks)\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>chunk_char_count</th>\n",
       "      <th>chunk_word_count</th>\n",
       "      <th>chunk_token_count</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>23.00</td>\n",
       "      <td>23.00</td>\n",
       "      <td>23.00</td>\n",
       "      <td>23.00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>2.30</td>\n",
       "      <td>1038.00</td>\n",
       "      <td>175.13</td>\n",
       "      <td>259.50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>1.74</td>\n",
       "      <td>604.59</td>\n",
       "      <td>103.45</td>\n",
       "      <td>151.15</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>0.00</td>\n",
       "      <td>19.00</td>\n",
       "      <td>2.00</td>\n",
       "      <td>4.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>1.00</td>\n",
       "      <td>458.50</td>\n",
       "      <td>74.00</td>\n",
       "      <td>114.62</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>2.00</td>\n",
       "      <td>1131.00</td>\n",
       "      <td>201.00</td>\n",
       "      <td>282.75</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>3.50</td>\n",
       "      <td>1403.50</td>\n",
       "      <td>229.50</td>\n",
       "      <td>350.88</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>5.00</td>\n",
       "      <td>2430.00</td>\n",
       "      <td>394.00</td>\n",
       "      <td>607.50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       page_number  chunk_char_count  chunk_word_count  chunk_token_count\n",
       "count        23.00             23.00             23.00              23.00\n",
       "mean          2.30           1038.00            175.13             259.50\n",
       "std           1.74            604.59            103.45             151.15\n",
       "min           0.00             19.00              2.00               4.75\n",
       "25%           1.00            458.50             74.00             114.62\n",
       "50%           2.00           1131.00            201.00             282.75\n",
       "75%           3.50           1403.50            229.50             350.88\n",
       "max           5.00           2430.00            394.00             607.50"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.describe().round(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "# some of our chunks have low token count\n",
    "# lets check for samples with less than 30 tokens (about the length of a sentence) and see if they are worth keeping"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Chunk token count: 35.25 | Text: 1220 Authorized licensed use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore. Restrictions apply.\n",
      "Chunk token count: 4.75 | Text: Restrictions apply.\n"
     ]
    }
   ],
   "source": [
    "# to show random chunks with under 30 tokens in length \n",
    "min_token_length = 40\n",
    "filtered_df = df[df[\"chunk_token_count\"] <= min_token_length]\n",
    "\n",
    "# Adjust the sample size based on the available rows\n",
    "sample_size = min(5, len(filtered_df))\n",
    "\n",
    "for row in filtered_df.sample(sample_size).iterrows():\n",
    "    print(f'Chunk token count: {row[1][\"chunk_token_count\"]} | Text: {row[1][\"sentence_chunk\"]}')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# looks like these are headers and footers of different pages.\n",
    "# they don't seem to offer too much information.\n",
    "# let's filter our dataframe/list of dictionaries to only include chunks with over 40 tokens in length."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[{'page_number': 0,\n",
       "  'sentence_chunk': 'Indian Stock Market Prediction using Deep Learning Ayan Maiti Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India ayan.182ma008@nitk.edu.in Pushparaj Shetty D Department of Mathematical and Computational Sciences National Institute of Technology Karnataka Surathkal, India prajshetty@nitk.edu.in Abstract—In this paper, we predict the stock prices of ﬁve companies listed on India’s National Stock Exchange (NSE) using two models- the Long Short Term Memory (LSTM) model and the Generative Adversarial Network (GAN) model with LSTM as the generator and a simple dense neural network as the discriminant. Both models take the online published historical stock-price data as input and produce the prediction of the closing price for the next trading day. To emulate the thought process of a real trader, our implementation applies the technique of rolling segmentation for the partition of training and testing dataset to examine the effect of different interval partitions on the prediction performance. Index Terms—LSTM, GAN, generator, discriminant, neural networks, technical indicators, rolling segmentation I. INTRODUCTION The prediction of stock prices and the direction of the price movement is a widely studied topic in many ﬁelds including trading, ﬁnance, statistics, and computer science. Typically, the professional stock market traders use funda- mental and/or technical analysis of stocks to analyze the market elements/indices to make investment decisions [1]. Fundamental analysis is the conventional method involving the study of company fundamentals such as sales, revenue and expenses, net proﬁt, debts, annual growth rates, and so on, whereas technical analysis, is exclusively based on the study of historical price movements. Practitioners of technical analysis refer to the price charts to study price patterns and use price data in different calculations to forecast future price movements. The technical analysis paradigm assumes, there exists an inherent correlation between the performance of the company and its stock price and therefore, it is used to determine the timing of buying and selling the shares of the company. Time-series prediction problems like the prediction of stock prices are a difﬁcult type of predictive modeling prob- lem. Such problems usually involve the presence of sequence dependence among the input variables [2].',\n",
       "  'chunk_char_count': 2430,\n",
       "  'chunk_word_count': 349,\n",
       "  'chunk_token_count': 607.5},\n",
       " {'page_number': 0,\n",
       "  'sentence_chunk': 'A robust type of neural network equipped to handle the sequence dependence is the Recurrent Neural Networks (RNNs). The Long Short- Term Memory network or LSTM network is a type of RNN and is used extensively in time-series problems because of its capability to handle dependency between both long and short distance sequences [3]. Generative adversarial network (GAN) has been successfully applied to a wide range of ﬁelds, such as image in painting, semantic segmentation, and video prediction. GAN consists of a discriminative network D, which learns to distinguish if a given data instance is real or not, and a generative network G, which learns to confuse D by generating high-quality data [4]. This approach has also been used for the prediction of the high-frequency stock prices of the Shanghai and Shenzhen stock exchanges of the Chinese stock market [5]. However, it has not been used to predict the daily stock prices of the Indian stock market as per our knowledge. Our contributions in this paper are as follows: 1. Im- plemented two models for daily frequency stock price prediction- LSTM and GAN 2. Performed rolling segmen- tation on the training set and testing set of the raw data to investigate the effect of the model parameter update cycle on stock forecast performance. We found that LSTM outperformed GAN signiﬁcantly with an error of 0.074 compared to 0.32 obtained by GAN.',\n",
       "  'chunk_char_count': 1398,\n",
       "  'chunk_word_count': 231,\n",
       "  'chunk_token_count': 349.5}]"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pages_and_chunks_over_min_token_len = df[df[\"chunk_token_count\"] > min_token_length].to_dict(orient=\"records\")\n",
    "pages_and_chunks_over_min_token_len[:2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# smaller chunks filtered\n",
    "# time to embed our chunks of text"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [],
   "source": [
    "# embedding our text chunks\n",
    "# * while humans understand text, machines understand numbers best\n",
    "# * an embedding is a broad concept.\n",
    "# * but one of my favourite and simple definitions is \"a useful numerical representation\"\n",
    "# * the most powerful thing about modern embeddings is that they are learned representations.\n",
    "# * meaning: rather than directly mapping words/tokens/characters to numbers directly (eg: {\"a\": 0, \"b\": 1, \"c\": 3...}),\n",
    "# * the numerical representation of tokens is learned by going through large corpuses of text and figuring out how different \n",
    "# tokens relate to each other.\n",
    "# * ideally, embeddings of text will mean that similar meaning texts have similar numerical representation.\n",
    "# note: most modern nlp models deal with 'tokens' which can be considered as multiple different sizes and combinations of words\n",
    "# and characters rather than always whole words or single characters. for example, the string \"hello world!\" gets mapped to the token values {15339: b'hello', 1917: b' world', 0: b'!'} \n",
    "# using Byte pair encoding (or BPE vio OpenAI's tiktoken library). Google has a tokeniization library called SentencePiece."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# sentence-transformers library contains many pre-trained embedding models.\n",
    "# all-mpnet-base-v2 model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: sentence-transformers in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (3.0.1)\n",
      "Requirement already satisfied: transformers<5.0.0,>=4.34.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (4.44.0)\n",
      "Requirement already satisfied: tqdm in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (4.66.4)\n",
      "Requirement already satisfied: torch>=1.11.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (2.4.0+cu118)\n",
      "Requirement already satisfied: numpy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (1.26.4)\n",
      "Requirement already satisfied: scikit-learn in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (1.5.0)\n",
      "Requirement already satisfied: scipy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (1.14.0)\n",
      "Requirement already satisfied: huggingface-hub>=0.15.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (0.24.5)\n",
      "Requirement already satisfied: Pillow in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sentence-transformers) (10.3.0)\n",
      "Requirement already satisfied: filelock in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (3.15.4)\n",
      "Requirement already satisfied: fsspec>=2023.5.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (2024.5.0)\n",
      "Requirement already satisfied: packaging>=20.9 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (24.1)\n",
      "Requirement already satisfied: pyyaml>=5.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (6.0.1)\n",
      "Requirement already satisfied: requests in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (2.32.3)\n",
      "Requirement already satisfied: typing-extensions>=3.7.4.3 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.15.1->sentence-transformers) (4.12.2)\n",
      "Requirement already satisfied: sympy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch>=1.11.0->sentence-transformers) (1.13.1)\n",
      "Requirement already satisfied: networkx in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch>=1.11.0->sentence-transformers) (3.3)\n",
      "Requirement already satisfied: jinja2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch>=1.11.0->sentence-transformers) (3.1.4)\n",
      "Requirement already satisfied: colorama in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from tqdm->sentence-transformers) (0.4.3)\n",
      "Requirement already satisfied: regex!=2019.12.17 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from transformers<5.0.0,>=4.34.0->sentence-transformers) (2024.7.24)\n",
      "Requirement already satisfied: safetensors>=0.4.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from transformers<5.0.0,>=4.34.0->sentence-transformers) (0.4.4)\n",
      "Requirement already satisfied: tokenizers<0.20,>=0.19 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from transformers<5.0.0,>=4.34.0->sentence-transformers) (0.19.1)\n",
      "Requirement already satisfied: joblib>=1.2.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from scikit-learn->sentence-transformers) (1.4.2)\n",
      "Requirement already satisfied: threadpoolctl>=3.1.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from scikit-learn->sentence-transformers) (3.5.0)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from jinja2->torch>=1.11.0->sentence-transformers) (2.1.5)\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers) (3.3.2)\n",
      "Requirement already satisfied: idna<4,>=2.5 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers) (3.7)\n",
      "Requirement already satisfied: urllib3<3,>=1.21.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers) (1.26.20)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.15.1->sentence-transformers) (2024.6.2)\n",
      "Requirement already satisfied: mpmath<1.4,>=1.1.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sympy->torch>=1.11.0->sentence-transformers) (1.3.0)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "%pip install sentence-transformers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\threadpoolctl.py:1214: RuntimeWarning: \n",
      "Found Intel OpenMP ('libiomp') and LLVM OpenMP ('libomp') loaded at\n",
      "the same time. Both libraries are known to be incompatible and this\n",
      "can cause random crashes or deadlocks on Linux when loaded in the\n",
      "same Python program.\n",
      "Using threadpoolctl may cause crashes or deadlocks. For more\n",
      "information and possible workarounds, please see\n",
      "    https://github.com/joblib/threadpoolctl/blob/master/multiple_openmp.md\n",
      "\n",
      "  warnings.warn(msg, RuntimeWarning)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WARNING:tensorflow:From c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\tf_keras\\src\\losses.py:2976: The name tf.losses.sparse_softmax_cross_entropy is deprecated. Please use tf.compat.v1.losses.sparse_softmax_cross_entropy instead.\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\transformers\\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "from sentence_transformers import SentenceTransformer\n",
    "embedding_model = SentenceTransformer(model_name_or_path=\"all-mpnet-base-v2\", device=\"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentence: The Sentences Transformers library provides an easy and open-source way to create embeddings.\n",
      "Embedding: [-2.07981225e-02  3.03164274e-02 -2.01218035e-02  6.86483458e-02\n",
      " -2.55255830e-02 -8.47690552e-03 -2.07211357e-04 -6.32376820e-02\n",
      "  2.81605870e-02 -3.33353207e-02  3.02634258e-02  5.30721098e-02\n",
      " -5.03526367e-02  2.62288470e-02  3.33313793e-02 -4.51578237e-02\n",
      "  3.63043770e-02 -1.37112790e-03 -1.20171234e-02  1.14946710e-02\n",
      "  5.04510365e-02  4.70856838e-02  2.11913586e-02  5.14606833e-02\n",
      " -2.03746408e-02 -3.58889252e-02 -6.67861314e-04 -2.94393729e-02\n",
      "  4.95858900e-02 -1.05639752e-02 -1.52013600e-02 -1.31756405e-03\n",
      "  4.48196977e-02  1.56022664e-02  8.60380283e-07 -1.21387281e-03\n",
      " -2.37978473e-02 -9.09421477e-04  7.34481029e-03 -2.53935903e-03\n",
      "  5.23369797e-02 -4.68044244e-02  1.66214760e-02  4.71578725e-02\n",
      " -4.15600166e-02  9.01949592e-04  3.60278077e-02  3.42214927e-02\n",
      "  9.68227237e-02  5.94828725e-02 -1.64984781e-02 -3.51249948e-02\n",
      "  5.92518132e-03 -7.07973901e-04 -2.41031218e-02  3.49740684e-02\n",
      " -2.94746365e-02  6.04275987e-03 -9.80651937e-03  2.83219013e-02\n",
      " -1.85376257e-02  3.63212936e-02  1.30292866e-02 -3.71233225e-02\n",
      "  5.27257062e-02 -1.19706690e-02 -7.18082413e-02  1.24431578e-02\n",
      " -6.70568412e-03  7.42154419e-02  1.16357002e-02 -1.74533483e-02\n",
      " -1.82405710e-02 -1.88930482e-02  2.82415077e-02  1.32829621e-02\n",
      " -3.51910293e-02  8.87307804e-04  5.79572134e-02  3.22093479e-02\n",
      " -3.48579651e-03  4.13768403e-02  1.44357886e-02 -3.28044519e-02\n",
      " -9.79081169e-03 -3.16493325e-02  4.23870832e-02 -4.70847338e-02\n",
      " -2.08936594e-02 -1.91249475e-02 -1.22627066e-02  1.01605318e-02\n",
      "  3.91921736e-02 -2.61895843e-02  1.09028080e-02  1.35722766e-02\n",
      " -5.79266623e-02 -3.21499892e-02 -5.75727411e-03 -2.43516769e-02\n",
      "  5.23417369e-02  5.46120852e-03 -2.30996143e-02  2.57168920e-03\n",
      " -6.63346350e-02  3.54126208e-02 -1.03907697e-02  2.25410052e-02\n",
      " -1.84574183e-02 -2.42006574e-02 -4.78365384e-02 -4.79237828e-03\n",
      " -5.34137972e-02  3.01791690e-02 -1.56130241e-02 -5.51476218e-02\n",
      " -3.91875058e-02  5.92152588e-02 -3.47646885e-02  9.68120527e-03\n",
      "  2.13415809e-02  2.30417494e-02  1.91712212e-02  2.77378075e-02\n",
      " -7.73508521e-03  1.04445703e-02 -2.67719813e-02 -2.40198933e-02\n",
      " -1.92289259e-02  3.91503330e-03 -2.54714154e-02  3.61942388e-02\n",
      "  5.12866974e-02 -8.41696747e-03 -3.13829854e-02  1.47484578e-02\n",
      "  2.13939454e-02 -3.84901278e-02  2.01945659e-02  1.20765837e-02\n",
      " -3.12082679e-03  7.84027483e-03  3.30337090e-03 -4.94357944e-02\n",
      "  5.83886653e-02  3.26128071e-03  4.84482711e-03 -4.50681821e-02\n",
      "  2.45682597e-02  3.55428047e-02 -5.32506034e-02  9.21153054e-02\n",
      "  2.04395372e-02 -3.36951986e-02 -6.19803816e-02 -2.11038571e-02\n",
      "  7.82359987e-02  5.11908270e-02  5.93170375e-02 -1.25044011e-04\n",
      "  4.96349297e-02 -1.55723505e-02 -3.35671497e-03  1.82015859e-02\n",
      " -2.73443796e-02 -1.08772591e-02  1.41476262e-02  1.09877242e-02\n",
      "  4.32547927e-03  8.23311061e-02 -9.85404942e-04  7.58791193e-02\n",
      "  9.45002865e-03  2.37687863e-02  1.61929000e-02  6.24994077e-02\n",
      "  4.75922264e-02 -3.92629346e-03  9.07524452e-02  4.49874103e-02\n",
      " -3.47132124e-02  2.14078184e-02 -3.35604921e-02  4.93849255e-02\n",
      "  1.08669270e-02  2.63447091e-02 -3.26089524e-02  8.00303295e-02\n",
      "  9.29767732e-03  7.16571556e-03 -2.79172026e-02 -3.06820683e-02\n",
      "  4.01059398e-03 -4.93906029e-02 -3.13771772e-03  4.00538519e-02\n",
      " -3.97855230e-02  5.48014976e-02  1.35825694e-05 -8.38373825e-02\n",
      " -1.21547673e-02  3.40950340e-02  3.22403386e-03  6.11845925e-02\n",
      "  5.60066216e-02  9.62879229e-03  2.54617389e-02 -4.64168824e-02\n",
      " -3.98900062e-02  7.68132806e-02  2.28408314e-02 -2.26567835e-02\n",
      " -1.91193055e-02 -6.53027520e-02  4.56780642e-02 -4.43657488e-03\n",
      "  1.49631947e-02 -2.15078294e-02  2.74245255e-03  1.90358702e-02\n",
      "  5.91888577e-02 -2.47569978e-02  3.66145633e-02  5.63083626e-02\n",
      " -8.86441860e-03 -1.74324885e-02 -1.03288004e-03  2.47666612e-02\n",
      "  1.30763501e-02  5.04633747e-02 -5.28492453e-03  5.92397079e-02\n",
      "  6.29906803e-02 -4.36784290e-02 -4.97831851e-02  5.56296892e-02\n",
      " -2.44854260e-02 -8.26755613e-02  2.04911120e-02 -1.06446408e-01\n",
      "  6.64834445e-03  2.97304336e-02 -2.36440804e-02 -8.84610321e-03\n",
      "  2.45558773e-03 -3.35234590e-02  7.52212405e-02 -5.89879863e-02\n",
      " -3.67808156e-02  3.41542214e-02  5.41131347e-02 -1.74904633e-02\n",
      "  1.33920740e-02  4.71681468e-02  1.46117182e-02 -2.12311260e-02\n",
      " -6.55339062e-02  1.23857139e-02  2.76075620e-02 -8.02149717e-03\n",
      " -4.59636040e-02 -8.22450873e-03  9.16952454e-03 -1.56398322e-02\n",
      "  7.54615944e-03  1.58312020e-03 -3.03958859e-02 -5.10670729e-02\n",
      "  1.96313746e-02  1.26263024e-02 -1.51735963e-03  2.02891249e-02\n",
      "  1.37817832e-02  1.49110872e-02  2.50766873e-02 -3.62870395e-02\n",
      "  1.08084800e-02  2.74132518e-03  1.81510895e-02  5.39872460e-02\n",
      " -4.74541485e-02 -4.28731032e-02 -2.89914366e-02  2.13234536e-02\n",
      " -3.85161266e-02  6.31922409e-02 -5.77975772e-02  3.77895427e-03\n",
      " -2.54394282e-02 -1.77164722e-04  9.08248499e-03  1.59094781e-02\n",
      "  4.11798954e-02 -3.94368991e-02 -9.64434352e-03  1.30792353e-02\n",
      "  6.87962398e-02  4.32192832e-02  7.54126464e-04  6.77741915e-02\n",
      "  4.93705608e-02 -3.47819994e-03 -1.06054544e-02  6.72489591e-03\n",
      " -1.39062414e-02  4.88276593e-02 -1.05736414e-02  3.50220595e-03\n",
      "  2.90227309e-03  2.40044892e-02  1.20272590e-02 -2.09796336e-02\n",
      " -2.39112340e-02  3.26579437e-02 -1.01326371e-03 -5.92759578e-03\n",
      " -7.40535418e-03  3.63143324e-03 -2.26698592e-02 -2.21242309e-02\n",
      "  3.86996046e-02  1.72321536e-02  3.85920741e-02 -5.04710898e-02\n",
      " -3.42144556e-02 -4.00443301e-02 -3.57911140e-02 -4.62560579e-02\n",
      "  6.70232251e-02 -4.61646076e-03 -3.29684489e-03  2.08443590e-02\n",
      " -5.14256116e-03 -5.00850491e-02  2.22503841e-02  4.66933101e-02\n",
      "  1.36208748e-02  1.77531131e-02  4.28085402e-03 -2.79332213e-02\n",
      " -1.93421673e-02 -3.87860388e-02 -3.09555102e-02 -6.64134324e-02\n",
      " -1.13434102e-02  1.64267831e-02  1.77630130e-02 -2.28219968e-03\n",
      " -3.30088027e-02 -1.36267277e-03 -2.17934530e-02 -2.67509371e-02\n",
      " -1.26376059e-02  1.61859533e-03 -4.95672636e-02  7.85445645e-02\n",
      "  4.10963036e-02  9.65919159e-03 -1.14643313e-02  1.68858038e-03\n",
      "  5.37663698e-02  2.05541425e-03 -4.11200933e-02  1.46330092e-02\n",
      " -3.75564322e-02 -3.35689187e-02  5.19251451e-03 -6.33088946e-02\n",
      "  3.32963541e-02  8.76123831e-03  1.33862079e-03 -3.95751372e-03\n",
      " -1.61677320e-02  8.26746598e-02  4.75944541e-02 -3.43055241e-02\n",
      "  2.50881519e-02 -3.50976139e-02  3.68656963e-02  4.12650080e-03\n",
      "  4.16018218e-02 -1.35181755e-01 -4.76337411e-02 -1.20025743e-02\n",
      " -3.48892249e-02  3.25454548e-02 -2.93576508e-03 -4.85054497e-03\n",
      " -1.04223840e-01  2.78611071e-02  1.41570047e-02  3.94395813e-02\n",
      " -3.88806351e-02 -1.42464247e-02 -5.19984253e-02  8.92743003e-03\n",
      " -1.99770946e-02 -2.51725316e-02 -3.41300368e-02  1.93041097e-02\n",
      " -5.20207211e-02 -6.71999902e-02 -9.46360268e-03 -1.25584239e-03\n",
      " -5.66047989e-02  2.62097567e-02  9.91585199e-03  4.38286923e-02\n",
      "  2.26635626e-03 -3.11896093e-02 -6.25468194e-02 -3.87792885e-02\n",
      " -6.83938861e-02  4.93722335e-02  5.85507527e-02 -4.08729799e-02\n",
      " -1.98638011e-02 -2.12634839e-02  4.98037413e-02 -4.51748557e-02\n",
      " -2.37141401e-02  2.32675038e-02  1.00594826e-01  9.87114850e-03\n",
      " -1.38015356e-02 -5.21041490e-02  9.08213202e-03  1.72427855e-02\n",
      "  5.91432154e-02  2.62336992e-02 -7.04638939e-03 -1.50032481e-02\n",
      " -3.76662984e-03  6.28265133e-03 -5.23982942e-02 -4.96638939e-02\n",
      "  3.06611136e-02 -3.33647826e-03  2.34911032e-02 -8.58830139e-02\n",
      " -4.62448820e-02  5.59700727e-02  3.09069263e-04  2.01728791e-02\n",
      " -2.98059289e-03  1.76645238e-02  1.54669695e-02 -7.41718188e-02\n",
      "  7.34985946e-03 -1.05014481e-02  2.45247595e-02  1.36878621e-02\n",
      " -1.17803719e-02  4.51544449e-02  3.29039209e-02 -3.50392400e-03\n",
      " -2.71316059e-02 -5.27365059e-02 -4.60164584e-02  2.22849734e-02\n",
      "  2.62272060e-02  5.56147983e-03  1.45788630e-02 -2.97144428e-02\n",
      "  3.57042961e-02  2.22534314e-02  3.89617980e-02 -7.92635009e-02\n",
      " -9.01094824e-03  2.19012983e-02 -5.49062248e-03  8.69966298e-03\n",
      "  4.33031246e-02 -2.12631840e-02  1.13292420e-02 -6.33699149e-02\n",
      "  3.63723412e-02  2.67442800e-02 -6.64251596e-02  1.70499906e-02\n",
      " -2.79692356e-02  2.36356026e-03 -1.81953423e-02  1.52955204e-02\n",
      " -8.50425940e-03  1.16648218e-02 -9.75922197e-02 -2.92093456e-02\n",
      " -5.42547517e-02  3.61235179e-02  3.25116813e-02  8.26975890e-03\n",
      " -2.96541228e-04  1.11557022e-02 -3.85188609e-02  2.36161817e-02\n",
      "  9.85916704e-03  5.73998205e-02  4.86060865e-02 -1.37579739e-02\n",
      " -6.19213143e-03  1.11972783e-02 -3.37175056e-02 -1.10515635e-02\n",
      " -7.08333105e-02 -1.01816421e-02 -3.66010182e-02 -1.55561175e-02\n",
      " -2.13109832e-02 -1.02760009e-02 -4.35733832e-02  5.55186644e-02\n",
      " -3.76548283e-02  5.29251806e-02 -3.45223956e-02 -2.43011606e-03\n",
      "  7.25553781e-02  4.45065321e-03  4.71416265e-02 -9.43878107e-03\n",
      " -1.98978353e-02  5.71900085e-02  8.60541239e-02 -5.25058098e-02\n",
      " -1.39550185e-02  1.17373560e-02  1.33974375e-02 -4.73052897e-02\n",
      " -5.41674308e-02  4.62726206e-02 -2.58970521e-02  1.51415616e-02\n",
      "  3.38945128e-02 -3.78260086e-03 -5.76042794e-02 -1.60082411e-02\n",
      "  2.42738184e-02  3.37360427e-02 -1.96820591e-02 -2.53463667e-02\n",
      " -4.75616530e-02 -5.68756200e-02 -2.28193998e-02  3.83186489e-02\n",
      " -1.78330541e-02  1.35963671e-02  7.86027755e-04  9.74019244e-03\n",
      "  3.34298909e-02 -2.60134283e-02 -7.38575216e-03  3.56451757e-02\n",
      " -2.68532522e-02 -7.53624216e-02 -2.66983788e-02 -4.46457714e-33\n",
      " -3.31645980e-02  1.41703654e-02 -3.92909646e-02 -3.46318074e-02\n",
      " -5.88677730e-03 -1.18212411e-02  1.53951179e-02  1.18474467e-02\n",
      "  1.07757514e-02  3.62141430e-02  7.87954964e-03 -2.31844522e-02\n",
      "  1.07623041e-02  1.72345974e-02  9.54262621e-04  2.83640306e-02\n",
      "  2.37419996e-02 -1.48057574e-02  1.24194368e-03  3.52351787e-03\n",
      "  2.33735312e-02  5.58308251e-02  5.38328551e-02 -3.74079086e-02\n",
      " -2.11805608e-02  1.52748253e-04 -7.27779279e-03  5.50561072e-03\n",
      "  3.05824392e-02  4.54633534e-02 -3.35786417e-02  3.16142254e-02\n",
      " -2.56390171e-03  3.96354683e-02 -1.47573343e-02  5.67168072e-02\n",
      " -5.62787540e-02 -5.04602026e-03  3.56154181e-02 -2.76198853e-02\n",
      " -2.32292116e-02 -4.63292859e-02 -3.70919332e-02 -4.23189290e-02\n",
      "  3.70306075e-02  7.88710918e-03  3.85176614e-02  1.74770853e-03\n",
      "  5.62760746e-03  6.18104031e-03 -6.90269023e-02 -9.42973234e-03\n",
      " -7.74678448e-03  1.68349948e-02  1.22767128e-02  2.26406511e-02\n",
      "  1.21009462e-02  1.11743826e-02  1.21539682e-02 -1.16861854e-02\n",
      " -4.41612527e-02  2.30048075e-02  2.20671725e-02 -5.87504618e-02\n",
      " -3.96428518e-02  6.83134869e-02 -3.29948440e-02 -3.66774462e-02\n",
      " -3.53655443e-02  1.76184904e-02  6.95638778e-03  5.92693202e-02\n",
      "  4.12155576e-02  7.98109919e-02 -5.36559941e-03  1.14238542e-02\n",
      " -2.96388213e-02 -1.15410909e-02  2.22811550e-02  7.93186761e-03\n",
      "  2.60356162e-02  1.28211575e-02  1.71346348e-02 -6.90194033e-03\n",
      " -1.07602915e-02  1.35716666e-02 -9.90719767e-04 -6.16075583e-02\n",
      "  4.40513566e-02 -8.26611358e-04 -2.78340671e-02 -1.23619726e-02\n",
      "  1.34629672e-02 -3.85745950e-02  1.08695519e-03  2.18713433e-02\n",
      " -3.32398340e-02  1.84615236e-02 -5.10105211e-03  3.74664292e-02\n",
      " -3.67546454e-03 -2.19246037e-02 -4.96486388e-03 -9.59826447e-03\n",
      "  2.33591087e-02  1.04876561e-02  4.38721962e-02 -1.51424604e-02\n",
      " -6.30309582e-02  8.23254231e-03 -1.09130498e-02 -4.06409241e-02\n",
      " -6.21690303e-02  2.21327040e-02 -2.71434318e-02  4.05540094e-02\n",
      " -8.09456035e-03 -1.76409225e-03  3.01525965e-02 -5.42264432e-03\n",
      " -4.69822064e-02 -1.73767712e-02  4.11631130e-02  3.20635848e-02\n",
      " -2.22943965e-02 -1.58162042e-02 -4.50721458e-02  5.69487065e-02\n",
      "  4.71596308e-02 -5.78058660e-02  1.32474313e-02 -4.71290108e-03\n",
      "  1.66824520e-07  4.81090397e-02  5.03628366e-02  5.45263477e-02\n",
      "  2.07568686e-02 -1.19080720e-02 -6.37492212e-03  5.26370993e-03\n",
      "  7.21948892e-02 -2.21763570e-02  2.20102426e-02 -9.90464468e-04\n",
      " -1.37163969e-02  6.89210370e-03  2.46913023e-02 -1.39462069e-01\n",
      "  2.56975484e-03 -4.64827605e-02 -4.04967852e-02 -6.08556084e-02\n",
      " -1.53213246e-02  1.36129841e-01  9.45034325e-02  4.25741486e-02\n",
      "  4.67130505e-02 -2.30677556e-02 -1.20965596e-02  3.86673473e-02\n",
      "  2.11654883e-03 -2.51472890e-02 -1.15076387e-02 -3.46506611e-02\n",
      " -2.29533408e-02 -6.33843942e-03 -3.05175781e-02 -1.56236589e-02\n",
      "  1.39514012e-02  3.27458343e-04  2.00330676e-03  4.15112497e-03\n",
      " -2.22924966e-02 -3.62590253e-02 -2.36579385e-02 -1.87817495e-02\n",
      " -1.96288973e-02  4.52125333e-02 -8.12569410e-02 -2.14569680e-02\n",
      " -4.41543125e-02 -2.68476028e-02  2.01974288e-02  2.82991701e-03\n",
      " -1.95011310e-02 -3.45330834e-02  2.26913802e-02  3.78325917e-02\n",
      " -1.02544045e-02 -2.19749939e-03 -8.96744356e-02 -4.50030826e-02\n",
      "  8.09705257e-03 -2.05805860e-02 -2.02997755e-02 -2.09922250e-02\n",
      " -1.79405324e-02  5.81896789e-02 -7.63651775e-03  1.50847621e-02\n",
      "  1.78279755e-34  4.86179069e-02  4.22228687e-02  4.71596047e-02\n",
      "  5.89047372e-02  3.99784744e-02 -5.27071096e-02  1.56905856e-02\n",
      " -5.25094743e-04  1.13652200e-02 -6.56410605e-02 -2.20848713e-02]\n",
      "\n",
      "Sentence: Sentences can be embedded one by one or as a list of strings.\n",
      "Embedding: [ 4.31718156e-02 -5.38701043e-02 -3.78044657e-02  4.27235253e-02\n",
      " -2.35409252e-02  3.44861262e-02  2.89586708e-02  1.92819117e-03\n",
      "  2.41732821e-02 -3.17012295e-02  7.32855424e-02  1.25589743e-02\n",
      "  3.64621319e-02 -2.05251910e-02  2.81973872e-02 -6.87329769e-02\n",
      "  4.22230922e-02  9.31711169e-04  3.54035236e-02  1.41787436e-02\n",
      "  7.83996098e-03  2.31179260e-02 -4.84739756e-03  1.07173966e-02\n",
      "  4.39492334e-03  5.47800586e-03 -3.80338915e-02 -3.05490405e-03\n",
      "  5.72236814e-03 -6.78213388e-02 -4.88007627e-02 -1.45032080e-02\n",
      "  6.68007042e-03 -7.17479289e-02  1.64644905e-06  1.07564339e-02\n",
      " -3.60921957e-02 -2.37057637e-02 -5.22792228e-02  3.46110649e-02\n",
      " -5.42179262e-03  1.62611548e-02  1.96564831e-02  2.25395169e-02\n",
      " -2.25998904e-03  4.06343043e-02  8.17157701e-02  2.48179138e-02\n",
      "  5.31883873e-02  7.82714635e-02 -1.91813223e-02 -1.94087848e-02\n",
      " -2.62806490e-02 -2.44083870e-02  5.49405552e-02  1.90319177e-02\n",
      "  1.60811376e-02 -2.68895850e-02 -8.24686047e-03  7.33444020e-02\n",
      "  1.00123063e-02  2.93315742e-02  3.42870434e-03 -2.13270243e-02\n",
      " -1.62444520e-03 -5.56251314e-03 -7.64880404e-02 -5.85449971e-02\n",
      " -2.82272175e-02  7.51849590e-03  7.11225420e-02  1.95444049e-03\n",
      "  5.45924064e-03  3.22311744e-03  5.12800440e-02 -3.54106277e-02\n",
      " -5.03608063e-02  4.70519699e-02  5.15474612e-03  1.52287548e-02\n",
      " -1.06680496e-02  3.16298753e-02 -9.09036025e-03 -4.01326679e-02\n",
      " -4.35235687e-02 -1.94970146e-02  1.65605266e-02 -4.71168756e-02\n",
      " -3.92091610e-02 -3.07756532e-02 -2.94167083e-02 -4.20826711e-02\n",
      "  2.27072253e-03 -2.78329682e-02  1.69421360e-02  7.74501124e-03\n",
      " -5.23742028e-02 -4.50039431e-02  3.83605771e-02 -4.90786135e-02\n",
      "  5.06619252e-02  1.01615051e-02 -1.25022316e-02 -4.64555342e-03\n",
      " -1.54539859e-02  1.58862602e-02  1.18369320e-02 -3.59232537e-02\n",
      " -7.76226372e-02  3.43358591e-02 -2.14709882e-02 -6.86098561e-02\n",
      " -5.46235405e-02  7.83901438e-02 -3.00702676e-02 -3.37550156e-02\n",
      " -4.04999107e-02  4.80514877e-02  9.53903422e-03  2.31398940e-02\n",
      " -8.16116035e-02 -6.51694415e-03  1.54213002e-02  7.04257637e-02\n",
      " -1.25069022e-02 -2.48266403e-02 -1.71329416e-02  6.13118056e-03\n",
      "  5.44412769e-02 -1.40566248e-02 -6.24513486e-03  3.65787335e-02\n",
      "  7.36229867e-02 -6.05683075e-03 -3.61629948e-02 -1.42205507e-03\n",
      "  4.43165302e-02 -3.14511918e-03  3.18767503e-02 -1.30948192e-02\n",
      " -3.69525179e-02 -4.98031825e-03  1.30017346e-03 -2.05213595e-02\n",
      "  2.06277296e-02  5.93865849e-03 -3.07163293e-03 -3.97512428e-02\n",
      "  4.29889970e-02  6.49801716e-02 -6.76022023e-02  5.41655011e-02\n",
      "  1.52563700e-03 -3.72908823e-02 -4.02427129e-02 -2.28771456e-02\n",
      "  1.31769776e-01  4.87875566e-03  1.39470762e-02  4.92436327e-02\n",
      "  2.49219332e-02 -8.76096636e-03 -5.38767176e-03 -2.65595205e-02\n",
      " -1.19766528e-02 -2.32006721e-02 -2.67433878e-02  5.66911884e-03\n",
      "  2.21722033e-02  4.67293970e-02 -5.78487441e-02  8.22120085e-02\n",
      " -3.36835161e-03  8.09647217e-02  1.41423829e-02  1.02393024e-01\n",
      " -5.76836336e-03 -1.15877036e-02  4.90585342e-02  5.87828979e-02\n",
      "  6.50030300e-02  4.74622548e-02 -2.89463755e-02 -1.76580774e-03\n",
      "  3.32560986e-02  2.91198194e-02  6.03810847e-02  3.73481336e-04\n",
      "  1.06575862e-02 -5.96284978e-02 -7.28601217e-02  2.95080524e-02\n",
      "  9.54463892e-03 -2.71541923e-02 -5.63305616e-02  9.66733438e-04\n",
      " -4.77728471e-02  4.67577465e-02  4.87258099e-03 -6.57520443e-02\n",
      " -1.42248664e-02  3.99872698e-02 -1.09797828e-02  7.68942237e-02\n",
      " -4.00004685e-02  2.96826642e-02  2.81302799e-02 -5.55424541e-02\n",
      "  6.31277077e-03  5.00450656e-02  1.89883709e-02  5.38683645e-02\n",
      " -1.95981245e-02  1.08601283e-02  1.64150950e-02  1.44135924e-02\n",
      "  1.71448775e-02  2.17624661e-02 -4.98863868e-02  1.56105701e-02\n",
      "  4.83735465e-03  1.87053438e-02 -3.18538491e-03  2.66863313e-02\n",
      "  5.55552840e-02 -4.88005541e-02 -3.02929021e-02  2.52110660e-02\n",
      "  1.07264556e-02  1.88270137e-02 -1.50687946e-02  3.43831889e-02\n",
      "  4.15125042e-02  1.37788523e-02 -5.54848425e-02  1.43847577e-02\n",
      " -5.88139892e-02 -6.01675883e-02  2.69856472e-02 -5.46130054e-02\n",
      "  8.14631023e-03 -1.17758494e-02  1.57441534e-02  1.43901503e-03\n",
      " -2.64554638e-02 -4.48877104e-02  4.39732820e-02 -1.06161373e-04\n",
      " -2.25905683e-02  3.00296135e-02  1.97440665e-02  7.44072394e-03\n",
      " -1.93789694e-02  8.09792802e-03  4.34860699e-02 -1.08573739e-04\n",
      " -3.77225019e-02  2.67195851e-02 -4.63157520e-02 -1.53400761e-03\n",
      "  8.05301871e-03 -4.30901498e-02 -2.13848855e-02  1.20185381e-02\n",
      "  8.41404963e-03  2.48264614e-03 -3.09566353e-02 -9.05277729e-02\n",
      " -4.76694033e-02  1.22605944e-02 -1.36466604e-02 -2.63655055e-02\n",
      " -7.65552092e-03  8.72376096e-03  2.65724473e-02  8.40124791e-04\n",
      " -5.55928051e-03 -9.29543748e-03  3.19337696e-02  5.94646633e-02\n",
      "  1.83206182e-02 -7.56547600e-02 -5.59388846e-02 -1.20871123e-02\n",
      " -3.16260979e-02  3.62186730e-02  7.53606064e-03 -6.15653470e-02\n",
      " -2.30458435e-02 -3.51675530e-03  1.23332227e-02 -9.67646576e-03\n",
      "  4.96861152e-02 -8.42256919e-02  1.52396243e-02 -1.82445236e-02\n",
      "  7.70462081e-02  9.28717852e-02  4.03725728e-02  1.11732557e-01\n",
      " -1.03271045e-02 -2.54558716e-02  2.13154256e-02 -1.16188079e-03\n",
      "  2.82600801e-03  5.06966449e-02 -3.13697979e-02 -8.14269856e-03\n",
      "  1.38386898e-02  4.66890410e-02  5.09671122e-02  3.77154239e-02\n",
      " -2.94988696e-02  3.60632427e-02 -2.61162641e-03  2.72251316e-04\n",
      " -6.71806931e-02 -6.54026940e-02 -3.43590714e-02  1.91067904e-02\n",
      "  4.13293839e-02 -1.10970605e-02  4.51953188e-02 -5.93565032e-02\n",
      "  1.06963487e-02 -1.82229448e-02 -5.65814339e-02  1.20387692e-02\n",
      "  4.44775335e-02  1.87049713e-02  1.63809657e-02  5.51151484e-02\n",
      " -2.23331824e-02  2.12861616e-02 -1.20339962e-02  3.26752774e-02\n",
      "  1.47003895e-02 -8.16674810e-03  1.12904049e-02 -3.00620534e-02\n",
      " -2.34345384e-02 -2.68647112e-02 -1.28722424e-03 -7.67189860e-02\n",
      "  2.22605979e-03 -5.89481089e-03  2.63104085e-02  2.07130332e-03\n",
      " -6.91151470e-02 -1.43792685e-02  2.68787798e-02 -3.51540558e-02\n",
      " -2.69611645e-02  2.54715560e-03 -6.48881271e-02  3.18727829e-02\n",
      "  1.70126520e-02 -4.54004258e-02 -1.80616081e-02 -1.61115881e-02\n",
      "  5.70773259e-02 -2.78267823e-03 -6.45586476e-02  7.86598176e-02\n",
      "  2.29075179e-02  6.81849616e-03 -9.11737606e-03 -2.27726400e-02\n",
      " -4.76526730e-02  4.88431044e-02 -2.09891722e-02 -2.43694223e-02\n",
      " -5.01211593e-03  6.70254081e-02  6.91366522e-03  2.25843284e-02\n",
      "  2.51125861e-02 -6.92505576e-03  8.59403051e-03  2.38977522e-02\n",
      "  3.29738148e-02 -1.05310574e-01  1.22094126e-02 -1.22263655e-02\n",
      " -5.73769175e-02  1.84311159e-02  2.97157913e-02 -6.09428957e-02\n",
      " -6.55256435e-02  3.55712995e-02  5.64320292e-03  3.34640313e-03\n",
      " -3.59686241e-02 -8.83413944e-03 -6.97894841e-02  6.89779669e-02\n",
      " -4.88217222e-03  2.23995112e-02 -3.16053778e-02 -7.41180591e-03\n",
      "  3.19351666e-02 -5.18788397e-02  2.11601350e-02 -5.03340140e-02\n",
      "  9.10580624e-03  2.13354006e-02  1.66838989e-02  3.49019654e-02\n",
      " -6.38500378e-02 -6.75630989e-03 -1.27405087e-02 -4.63365950e-02\n",
      " -1.14779873e-02  2.08778288e-02  2.44821459e-02  3.66468797e-03\n",
      " -2.86099338e-03  2.29389612e-02  2.13746056e-02 -3.48901041e-02\n",
      " -3.00387777e-02  4.78870459e-02  5.83370775e-02 -9.70496889e-03\n",
      "  1.38233909e-02 -3.27485502e-02 -8.11468752e-04  9.54235811e-03\n",
      "  1.20402053e-02  1.97231155e-02 -4.74864268e-04 -1.39225731e-02\n",
      " -5.21069951e-02 -1.75592601e-02 -5.41698039e-02 -1.17970314e-02\n",
      " -1.71029810e-02 -3.50194946e-02  3.38660926e-02 -6.76587150e-02\n",
      " -2.27607153e-02  1.95607040e-02  5.50249405e-02  1.22028934e-02\n",
      " -1.75161730e-03  7.22439028e-03  1.16349179e-02 -1.61907896e-02\n",
      " -3.37754823e-02  3.22626606e-02 -2.03813817e-02 -2.33859085e-02\n",
      " -1.29991733e-02 -1.66799091e-02  1.03070512e-02 -1.46030691e-02\n",
      " -7.79069960e-02 -8.25812519e-02 -3.38809304e-02  3.81114446e-02\n",
      "  7.86011480e-03  2.41454951e-02 -2.75715683e-02  1.30867865e-02\n",
      " -7.88596552e-03  1.78650916e-02  5.37323244e-02 -3.01822592e-02\n",
      "  1.69455428e-02  1.19571229e-02  3.52556410e-04  4.90209088e-02\n",
      " -8.57210066e-03  1.71269954e-03  4.83875209e-03 -4.10080589e-02\n",
      " -4.68120091e-02 -2.32563657e-03 -5.16776703e-02  3.10030933e-02\n",
      "  1.60961654e-02 -1.00802835e-02 -3.72486911e-03 -3.53388265e-02\n",
      "  2.95961648e-02  2.89097428e-02 -7.59910941e-02 -5.02981320e-02\n",
      " -2.11783871e-02  3.20462063e-02 -3.84537876e-02  2.45103091e-02\n",
      " -2.04188507e-02  6.02109311e-03 -9.81934462e-03  3.74778546e-02\n",
      "  3.40838321e-02  1.28864199e-02  5.67342900e-02 -8.09703842e-02\n",
      " -8.93603731e-03  1.33353109e-02 -2.51565874e-02  2.58417218e-03\n",
      " -6.51802421e-02  1.34400846e-02 -2.04682536e-02  6.53375546e-03\n",
      "  4.56976704e-03  1.99271347e-02 -6.07339740e-02  1.40691623e-02\n",
      " -5.75333387e-02  9.79796425e-03  3.55393179e-02 -2.45283600e-02\n",
      " -4.73310472e-03 -2.77493242e-02  2.34282520e-02 -8.76456252e-05\n",
      "  7.30438298e-03  1.42028444e-02  4.92807515e-02 -3.16540226e-02\n",
      " -1.34902494e-02  3.08487974e-02  2.80401763e-02 -4.33069095e-02\n",
      " -4.42284904e-02  3.80738862e-02  9.46701548e-05 -4.34896238e-02\n",
      "  1.43868346e-02  2.44337623e-03 -4.84073050e-02  1.08956061e-02\n",
      " -9.87490825e-03  4.59295288e-02  3.96379381e-02 -2.60116793e-02\n",
      "  2.48133354e-02 -5.37149124e-02  5.62824830e-02  8.81359726e-03\n",
      "  5.25077134e-02 -1.47370491e-02 -1.74380299e-02  3.45084704e-02\n",
      "  3.75523269e-02 -4.70166877e-02 -1.94911230e-02  3.82631756e-02\n",
      " -5.67595661e-02 -1.78618007e-03  2.33404096e-02 -5.88216602e-33\n",
      " -4.87188138e-02 -2.76264977e-02 -3.38240489e-02  2.66187694e-02\n",
      " -3.39277312e-02 -8.49193707e-03 -1.91250406e-02  3.00252791e-02\n",
      "  3.40782031e-02  5.11157922e-02 -1.92479808e-02  2.85642296e-02\n",
      "  3.66040319e-02  1.68858394e-02  4.77259457e-02  1.23802070e-02\n",
      "  2.14844178e-02  4.93718602e-04  1.21273529e-02 -5.82144596e-02\n",
      "  1.62954219e-02 -7.14258850e-03  4.80092503e-02  2.51190532e-02\n",
      "  4.60097268e-02 -2.29840092e-02 -2.05696374e-02 -3.22233210e-03\n",
      "  4.00091484e-02  3.52310091e-02 -3.43153775e-02  2.75635067e-03\n",
      " -1.25138499e-02  1.97685380e-02  5.53492270e-03  1.03744574e-01\n",
      "  5.77612640e-03 -5.65427281e-02  4.19559181e-02 -3.78831141e-02\n",
      " -3.93443145e-02 -6.24309666e-02 -2.24392489e-03 -5.46548776e-02\n",
      "  4.56134975e-02 -5.69243263e-03  3.38916853e-02 -1.44447908e-02\n",
      "  2.72112153e-03  1.11191729e-02 -5.00660837e-02 -1.61127746e-02\n",
      "  1.72816694e-03  6.88877106e-02  1.16492417e-02  2.83170715e-02\n",
      "  6.97187614e-03  2.68372353e-02 -7.72093097e-03  2.16828659e-02\n",
      "  1.15182865e-02  8.72833058e-02 -6.27270481e-03 -6.44473583e-02\n",
      " -1.58233810e-02  4.03268076e-02 -1.69728398e-02 -1.61188897e-02\n",
      " -3.75577249e-02  7.02938437e-02 -3.30486335e-02  4.66324389e-02\n",
      "  1.18027702e-02  6.51075318e-02 -1.16979610e-02 -8.28346424e-03\n",
      " -5.46905436e-02 -2.00227425e-02  8.42751353e-04 -8.19524936e-03\n",
      "  2.08356641e-02  1.37454700e-02 -1.29930757e-03 -3.94575149e-02\n",
      " -2.00185161e-02 -1.53721785e-02  1.17272064e-02 -4.40110825e-02\n",
      "  5.39267398e-02 -2.33010575e-02 -2.24211365e-02 -3.65212956e-03\n",
      "  2.92213354e-02  7.56442873e-03 -2.90923808e-02  4.01517451e-02\n",
      " -2.00853441e-02 -1.79869309e-03 -1.26235634e-02  2.51077209e-02\n",
      " -4.69285809e-02 -3.08553688e-02 -3.63379571e-04  6.01790939e-03\n",
      "  3.97508927e-02  1.38547821e-02  2.49774344e-02  1.76975839e-02\n",
      " -9.31572840e-02 -9.83685162e-03  8.44927970e-03 -1.95391178e-02\n",
      " -3.26569192e-02  5.13736671e-03  5.80933318e-03  2.08536983e-02\n",
      " -5.97834494e-03  5.86811416e-02 -1.49496226e-02 -5.72965629e-02\n",
      " -5.98235894e-03  1.95206725e-03  2.72973604e-03  6.06998522e-03\n",
      " -2.00525466e-02 -1.31687829e-02 -4.06228602e-02  5.68997450e-02\n",
      "  4.44969535e-02 -1.24308122e-02  1.96967646e-02  3.80979627e-02\n",
      "  2.30237688e-07  1.10575678e-02  4.79512848e-02  6.18298799e-02\n",
      "  4.40279134e-02  6.17665844e-03  2.58296775e-03  3.38913947e-02\n",
      " -5.32940868e-03 -2.59284787e-02 -1.26145016e-02  2.46495474e-02\n",
      " -1.68769213e-03  1.17907964e-03  2.40443349e-02 -9.77308974e-02\n",
      "  1.97367668e-02 -5.52921109e-02 -6.17424622e-02 -4.87151593e-02\n",
      "  1.11083000e-03  1.18732050e-01  8.13258067e-02  3.32449377e-02\n",
      "  4.38326597e-02 -2.49559153e-02 -3.59627157e-02  1.66319218e-02\n",
      "  5.93769178e-03 -1.43971080e-02  4.46710642e-03 -6.01986460e-02\n",
      " -5.65911420e-02 -8.21544789e-03  5.83061902e-03 -1.69482213e-02\n",
      "  9.58630629e-03  1.46733271e-02  5.05845137e-02  3.06891538e-02\n",
      "  6.60468563e-02 -2.56551914e-02 -2.78858561e-02 -3.19173485e-02\n",
      " -3.39236706e-02  1.49902795e-02 -3.03336717e-02 -6.06491277e-03\n",
      " -4.81766695e-03  1.72137134e-02 -8.23369995e-03  1.55548304e-02\n",
      "  2.69106980e-02  5.44307753e-03 -1.06899096e-02 -7.82140065e-03\n",
      " -4.44506146e-02  2.55873855e-02 -5.74761070e-02 -2.05442756e-02\n",
      " -3.07849273e-02 -1.57855265e-02 -7.07538892e-03 -4.21312861e-02\n",
      "  3.79934385e-02  6.27764612e-02 -7.67794251e-03 -3.18353623e-02\n",
      "  1.99277769e-34  1.04834037e-02 -3.39326411e-02  3.93821709e-02\n",
      "  5.53065687e-02  9.42170154e-03  1.09728016e-02 -4.91940118e-02\n",
      "  2.95024216e-02 -8.85373726e-03 -5.96247911e-02 -2.37825420e-02]\n",
      "\n",
      "Sentence: Embeddings are one of the most powerful concepts in machine learning!\n",
      "Embedding: [-2.98611317e-02 -1.37522798e-02 -4.75401990e-02  2.72126608e-02\n",
      "  3.40054594e-02  3.16466317e-02  4.26963605e-02  3.29796574e-03\n",
      "  4.35717702e-02  2.53837164e-02  3.02528758e-02  3.21130641e-02\n",
      " -3.99913080e-02  1.28760673e-02  6.70220032e-02 -7.92899057e-02\n",
      "  4.68772016e-02  2.40266267e-02 -2.07996797e-02 -1.07433377e-02\n",
      " -1.19410539e-02 -5.39290346e-02  4.21055444e-02  2.23588645e-02\n",
      " -2.98949573e-02  8.35980382e-03  1.58385299e-02 -4.80235815e-02\n",
      "  1.88430236e-03 -1.67521406e-02 -2.15628594e-02 -3.88488062e-02\n",
      "  3.06274295e-02  4.20526229e-02  1.69483383e-06 -1.86928734e-02\n",
      " -1.24558434e-02  1.32128336e-02 -4.89039607e-02  1.34747047e-02\n",
      "  2.28873119e-02  8.81782454e-03  8.64927098e-03 -2.00949367e-02\n",
      " -3.15217860e-02 -2.53432915e-02  7.57319480e-02  3.62446606e-02\n",
      "  1.25289615e-02  3.09694726e-02  4.50758031e-03 -3.50041613e-02\n",
      " -4.42560646e-04 -9.76646878e-03  6.04545176e-02  4.03472781e-02\n",
      "  1.10734636e-02  6.56200061e-03 -5.84602589e-03  3.79779818e-03\n",
      " -4.46914732e-02  1.76404882e-02  2.45917067e-02 -3.60036292e-03\n",
      "  1.02473356e-01  3.73758934e-02  6.13310048e-03 -2.24676095e-02\n",
      "  1.46482233e-02  5.00536747e-02 -2.29907595e-02  1.12924874e-02\n",
      " -3.10552437e-02 -1.49508985e-02 -2.53128516e-03  3.20944376e-02\n",
      " -4.67056260e-02 -4.85887118e-02  2.98305433e-02  6.44215718e-02\n",
      " -3.12613398e-02  3.57406400e-02  4.16527167e-02 -5.52517064e-02\n",
      " -8.74629710e-03 -2.18630563e-02 -1.12744505e-02 -2.14436464e-02\n",
      " -1.32823670e-02 -2.04865839e-02 -1.00576170e-02  3.54763642e-02\n",
      " -7.47602852e-03 -3.70188467e-02  5.77893667e-02 -2.18169019e-02\n",
      "  4.36227862e-03  2.04381328e-02  3.36815007e-02 -4.92800958e-02\n",
      "  4.82793190e-02 -1.81002240e-03 -1.05118118e-02  4.13323045e-02\n",
      " -6.79833218e-02  1.75715964e-02 -4.43412997e-02  9.90840700e-03\n",
      " -3.81809883e-02  1.10827386e-02 -5.07279262e-02 -2.17450727e-02\n",
      " -1.03835883e-02  4.60332148e-02  1.55862588e-02 -4.21366654e-02\n",
      " -2.72146370e-02  3.22818086e-02 -4.24739309e-02  2.71206740e-02\n",
      " -7.41061866e-02  4.20107394e-02  2.02438086e-02  7.31810629e-02\n",
      " -8.97698291e-03 -2.31159814e-02 -3.93559448e-02 -1.46008581e-02\n",
      " -3.30910385e-02  1.12239458e-02  2.58562807e-03 -4.36858460e-03\n",
      "  1.85855273e-02  2.69934218e-02 -1.67215224e-02  3.69569696e-02\n",
      "  4.44489531e-02 -2.21724417e-02  6.72966987e-03  1.22936275e-02\n",
      "  1.71758104e-02 -2.36477703e-03  3.72264087e-02 -2.22870745e-02\n",
      "  2.94603389e-02 -2.33690999e-02  5.38470875e-03 -3.06581315e-02\n",
      " -2.38920338e-02 -2.63614431e-02 -2.01789495e-02  1.11245580e-01\n",
      " -1.99836604e-02 -3.54029648e-02  3.84143814e-02  2.53069103e-02\n",
      "  1.99551433e-02  5.53518869e-02 -1.99332405e-02 -2.16719205e-03\n",
      "  4.91092689e-02 -4.03530560e-02 -1.16977096e-02 -5.33113219e-02\n",
      "  8.29580799e-03 -5.08251451e-02 -2.65503637e-02 -1.53242461e-02\n",
      "  5.78805711e-03  2.46571959e-03 -3.44449244e-02 -1.85131549e-03\n",
      " -3.95730287e-02 -2.71690693e-02  4.93568778e-02  8.38369131e-02\n",
      "  5.43491952e-02  8.22261199e-02  1.23894261e-02 -4.79796855e-03\n",
      "  7.77373731e-04  2.98485663e-02 -1.85585059e-02  5.98795600e-02\n",
      " -6.82790065e-03  9.78133874e-04  2.85485666e-02 -7.64624495e-03\n",
      " -1.86619759e-02 -2.69287322e-02 -2.90332865e-02 -1.37871085e-02\n",
      " -2.57600797e-03 -2.20173262e-02 -1.70821901e-02 -3.81843746e-02\n",
      "  2.21505426e-02 -3.59234065e-02 -1.19439485e-02 -3.18308249e-02\n",
      " -4.80801798e-02  9.77639761e-03 -1.04861963e-03  4.15372252e-02\n",
      " -1.10974777e-02 -4.72829826e-02  1.90985203e-02 -5.31177223e-02\n",
      "  2.11326182e-02 -2.53063533e-03  5.61055616e-02 -1.33795096e-02\n",
      " -5.95859485e-03 -1.20308073e-02  4.63929661e-02 -2.81909034e-02\n",
      "  2.25355159e-02 -2.50469800e-03 -3.52453291e-02  2.55494993e-02\n",
      "  9.10393428e-03  3.25215166e-03  2.55969423e-03 -1.25624416e-02\n",
      " -3.51496153e-02 -4.28946465e-02 -2.32331711e-03  2.41021328e-02\n",
      " -5.16842585e-03  1.68739948e-02  5.52647049e-03  2.36792285e-02\n",
      "  5.65164387e-02 -3.47868651e-02 -6.34518191e-02 -7.45624863e-03\n",
      " -1.78447962e-02  5.35899214e-02  2.67291274e-02 -8.74199271e-02\n",
      "  1.04195466e-02 -4.14006528e-04 -3.04448698e-03  9.14251432e-03\n",
      "  2.91529987e-02 -5.81832230e-02  6.83463663e-02 -4.08618264e-02\n",
      " -9.09790024e-03 -3.40769999e-02  3.52410972e-02 -1.02628311e-02\n",
      " -5.72506862e-04 -2.73456285e-03  1.59635767e-02  4.49077226e-03\n",
      " -2.09051874e-02  3.02769877e-02  2.46119127e-02 -1.44067276e-02\n",
      "  1.73268188e-02  1.99036743e-03  4.23051193e-02 -2.39177216e-02\n",
      " -3.25547643e-02 -1.45939644e-02  3.95101011e-02 -6.04650229e-02\n",
      " -3.02065499e-02  1.67189036e-02 -2.26817373e-02 -2.61954926e-02\n",
      " -5.51320463e-02  1.44907851e-02 -1.99246574e-02  3.99756664e-03\n",
      "  3.12610008e-02 -4.90728244e-02 -9.49713227e-04  5.39496876e-02\n",
      " -9.10082832e-03 -2.69486755e-02 -3.63159217e-02 -1.38436900e-02\n",
      " -4.45622206e-02  5.49358614e-02  2.17595650e-03  2.23447802e-03\n",
      " -5.23020420e-03 -1.47894183e-02  3.60592082e-02  1.45262973e-02\n",
      "  8.39193817e-03 -6.10361546e-02 -7.89948646e-03 -2.98300222e-03\n",
      "  3.56564159e-03  8.33992213e-02 -2.61215065e-02  8.06722194e-02\n",
      "  3.63061344e-03  1.69974267e-02  2.58604623e-02  1.09440798e-03\n",
      " -4.57063131e-02  5.55679090e-02  2.00643819e-02  4.76660840e-02\n",
      " -4.91054617e-02 -1.86080616e-02  3.34404819e-02 -2.57309806e-02\n",
      " -3.16372886e-03  7.21444637e-02 -1.61519069e-02 -1.33933993e-02\n",
      " -6.06294498e-02 -2.82186829e-02 -8.91922135e-03 -2.71166023e-03\n",
      "  8.04916117e-03 -4.95210178e-02  7.89433718e-02  2.76428591e-02\n",
      " -5.42582385e-03 -3.06722778e-03 -4.11826298e-02  1.39172496e-02\n",
      "  3.04253511e-02  1.02856206e-02  1.06678959e-02 -5.56553900e-02\n",
      " -1.75083261e-02  2.03868672e-02  8.43313057e-03  3.82471010e-02\n",
      " -3.89100052e-02 -1.61303207e-02  3.18058617e-02 -7.32970312e-02\n",
      " -1.76502187e-02 -4.79873829e-02 -5.55041544e-02 -5.00749191e-03\n",
      "  4.46762424e-04  3.57333384e-02 -8.24701565e-04 -3.34323980e-02\n",
      " -3.32416892e-02 -2.46461108e-02  2.15332191e-02  3.90857039e-03\n",
      "  2.53471099e-02  6.02992205e-03 -7.81607255e-03  1.23765310e-02\n",
      " -1.71039440e-02  2.68103518e-02  2.83681974e-03 -1.27643989e-02\n",
      "  1.00510888e-01  1.03580998e-02 -3.55143361e-02  1.56616438e-02\n",
      " -9.85950008e-02  4.58441265e-02 -3.15230414e-02 -2.35780738e-02\n",
      " -2.78349929e-02 -7.75483641e-05 -2.82362942e-02 -1.92918275e-02\n",
      "  1.87389385e-02  5.71941249e-02  2.56911889e-02 -3.20029743e-02\n",
      "  1.99074410e-02 -3.15819867e-02 -4.02061678e-02  5.77630661e-02\n",
      "  1.72974262e-02 -5.37013598e-02 -1.25325536e-02 -1.45484209e-02\n",
      " -5.76174743e-02  1.09727914e-02 -2.04728059e-02  2.85540950e-02\n",
      " -5.04399948e-02  4.36991118e-02  1.75710917e-02 -1.02343578e-02\n",
      " -9.69771817e-02 -2.99994722e-02 -2.86679715e-02  2.24936493e-02\n",
      " -1.68121029e-02 -1.43673792e-02 -8.79591797e-03 -1.69044118e-02\n",
      "  2.41557769e-02 -6.53192475e-02 -4.10800129e-02 -2.34057270e-02\n",
      " -6.76065162e-02 -1.55690592e-02  3.62357907e-02  7.83159658e-02\n",
      " -4.97516915e-02 -7.08547533e-02 -5.01179956e-02 -8.56470841e-04\n",
      "  5.44895837e-03  4.34705149e-03  9.88053009e-02 -2.16416083e-02\n",
      " -1.87750850e-02  1.15069812e-02  2.63996348e-02  1.65235344e-02\n",
      " -2.24058032e-02 -4.31826375e-02  1.31803915e-01 -2.97034364e-02\n",
      "  2.65934020e-02 -1.38888033e-02 -1.67003870e-02  3.44145745e-02\n",
      " -8.94354563e-03  6.16001636e-02 -3.42302918e-02  2.46420270e-03\n",
      " -8.14096630e-03  5.80325276e-02  5.24208546e-02 -1.53281242e-02\n",
      "  4.01382521e-02  1.51406927e-02 -3.01461620e-03 -4.97020781e-02\n",
      " -4.24301904e-03  5.77289239e-02  3.17873470e-02  4.74009030e-02\n",
      "  2.95218267e-02 -1.50123071e-02 -2.47945357e-02 -7.11501911e-02\n",
      "  2.06848755e-02  3.11488286e-02 -5.86065371e-03  1.62786990e-02\n",
      " -3.93683128e-02  5.46505861e-02  3.26595753e-02 -1.87021382e-02\n",
      " -9.79863405e-02  4.33777180e-03 -5.58156818e-02 -1.34620992e-02\n",
      "  2.88454164e-02  1.58748273e-02 -3.32564376e-02  1.44417130e-03\n",
      " -5.51108494e-02  8.24674591e-02  2.38845740e-02 -2.04837658e-02\n",
      " -4.78584785e-03  3.78722474e-02 -4.87562791e-02  3.44647020e-02\n",
      "  1.10358289e-02  1.12450160e-02  1.33263050e-02 -3.46375071e-02\n",
      " -6.92220777e-02  7.30119320e-03 -6.57008542e-03  1.73203647e-02\n",
      "  5.23058651e-03  4.48132120e-02  3.89852896e-02 -1.99274272e-02\n",
      " -1.80920549e-02  3.25938389e-02 -2.02027392e-02  4.86310106e-04\n",
      " -8.88758805e-03 -1.91348437e-02  2.50686277e-02  4.74019460e-02\n",
      "  2.18650233e-03 -1.69988517e-02  3.62669975e-02  3.46247596e-03\n",
      "  4.21921723e-03  8.04171041e-02  3.10627390e-02 -1.04949484e-03\n",
      " -3.55466641e-02  4.34836857e-02 -3.06218527e-02 -3.03191822e-02\n",
      " -4.13175561e-02 -1.05258301e-02 -2.35241968e-02 -1.86771937e-02\n",
      "  4.42934455e-03  5.45056686e-02 -6.05017319e-02  2.48422083e-02\n",
      " -3.36967483e-02 -4.54169177e-02 -2.63173189e-02  6.98054116e-03\n",
      "  6.92871064e-02 -2.04491969e-02 -1.96812414e-02 -9.72558651e-03\n",
      " -1.21564716e-02  7.89338723e-03  1.84750929e-03 -6.93651065e-02\n",
      "  2.43357476e-02  4.00609635e-02  3.44013534e-02 -2.84281597e-02\n",
      " -1.09431623e-02  1.38742756e-02 -4.40585706e-03  1.19352713e-03\n",
      " -8.81165639e-02  1.15931230e-02 -2.56350618e-02  5.57525083e-02\n",
      "  1.26946181e-01  5.39566129e-02 -1.41437072e-02  1.27196005e-02\n",
      " -1.32235689e-02 -5.94484210e-02  2.86703967e-02  2.57285032e-02\n",
      " -8.33771937e-03  8.17305408e-04  5.93051873e-03  3.29110473e-02\n",
      "  4.12751660e-02 -5.77967055e-03 -1.71124451e-02  1.06227621e-02\n",
      " -2.19601821e-02 -4.97206748e-02  2.53765676e-02 -5.60257579e-33\n",
      " -1.35397101e-02 -3.77958380e-02 -2.67921132e-03 -3.69663932e-04\n",
      " -1.98267736e-02  5.47051430e-03  6.02685567e-03  1.93069372e-02\n",
      "  3.87977273e-03  2.97698714e-02 -1.88228618e-02  2.20037065e-03\n",
      "  8.17020051e-03  1.61965005e-02  3.17529812e-02 -6.83413958e-03\n",
      "  2.19251607e-02  4.37707175e-04  2.96859220e-02 -2.62557808e-02\n",
      "  6.49381289e-03  3.56024988e-02  1.58606865e-03 -4.76583727e-02\n",
      " -5.26238829e-02  3.78274433e-02  3.54342088e-02 -3.10347807e-02\n",
      "  7.96400011e-03  5.48469685e-02 -3.56443636e-02  9.10131540e-03\n",
      " -9.45980474e-03 -4.63287011e-02 -1.63906775e-02  6.32453188e-02\n",
      " -1.38587961e-02 -5.95724098e-02 -1.57990549e-02  2.01886706e-02\n",
      " -1.98293068e-02 -3.49211767e-02  2.27937903e-02 -5.91621324e-02\n",
      "  4.18854468e-02  1.20740815e-03  5.19158542e-02 -1.88436117e-02\n",
      " -3.12102009e-02  2.34933030e-02 -7.41029680e-02 -2.76625564e-04\n",
      " -1.51720028e-02  6.11713566e-02  1.25065088e-01 -1.28459297e-02\n",
      " -1.12671526e-02  1.51757977e-03 -8.09153318e-02  1.12689631e-02\n",
      " -1.97574068e-02  2.74268053e-02  9.40995198e-03 -9.58754774e-03\n",
      "  2.54850090e-02  6.81658536e-02 -1.83453653e-02 -1.00963995e-01\n",
      " -9.45249572e-03 -5.27017144e-03  1.98684316e-02  9.80847925e-02\n",
      "  3.15633900e-02  5.30422665e-02  3.75123210e-02 -6.64208531e-02\n",
      " -5.92880212e-02 -1.57074276e-02  1.76609624e-02 -5.81073016e-02\n",
      "  2.23230626e-02  1.29869003e-02 -3.30260471e-02  9.96903400e-04\n",
      " -9.87091009e-03 -3.12955342e-02  2.28521042e-03 -4.91250381e-02\n",
      "  1.47694303e-02 -1.83367021e-02 -4.16306369e-02  3.76896970e-02\n",
      "  3.35410200e-02 -7.97111914e-02  4.01299521e-02  1.59071647e-02\n",
      "  5.06078172e-03  4.28808108e-02  2.29760278e-02 -4.13351692e-02\n",
      " -3.10503859e-02 -5.26404902e-02 -4.95404713e-02 -2.94256713e-02\n",
      "  5.94924986e-02 -2.59802770e-02  3.02497428e-02  8.80404562e-03\n",
      " -4.84467037e-02 -2.00851522e-02  9.82179958e-03 -7.89194107e-02\n",
      "  4.52878326e-03 -9.34903789e-03  9.23305750e-03 -3.17361727e-02\n",
      "  2.10833419e-02  6.37119077e-03  3.36347707e-02  3.83614413e-02\n",
      " -4.55527827e-02  1.08119461e-03 -9.83326510e-03  7.70191802e-03\n",
      " -2.87617669e-02 -1.74959395e-02 -4.27809078e-03  2.81287208e-02\n",
      "  4.97339964e-02 -7.45570511e-02 -1.07009020e-02 -7.66060175e-03\n",
      "  2.33968819e-07  1.52483536e-02  8.39613974e-02  3.67242917e-02\n",
      " -3.69249023e-02  3.64752263e-02  4.26421985e-02 -4.39831009e-03\n",
      "  1.78134013e-02 -2.67076101e-02 -7.13903550e-03  5.59975393e-02\n",
      "  3.13966833e-02  2.13440275e-03  3.90371084e-02 -8.78527462e-02\n",
      " -2.21659709e-02 -2.47735735e-02 -1.18189799e-02 -7.89705571e-03\n",
      " -2.08857376e-02  4.30556349e-02  1.07643567e-01  4.40618359e-02\n",
      "  1.47962533e-02  2.44862884e-02 -3.86268124e-02  1.80743858e-02\n",
      " -1.47838739e-03  7.74166659e-02 -4.19566259e-02 -3.80528979e-02\n",
      "  3.61257233e-02  1.59031420e-03  1.95324589e-02 -2.00080480e-02\n",
      "  4.22538295e-02  3.06111053e-02 -3.53681808e-03  5.93103329e-03\n",
      " -2.23223381e-02 -2.07131263e-02 -3.62910307e-03  1.74653605e-02\n",
      " -4.08758782e-02  5.91595285e-02 -5.89100234e-02 -3.96753810e-02\n",
      " -3.33528556e-02  1.02161970e-02  6.97290059e-03  7.70389959e-02\n",
      " -1.86911654e-02 -1.82568301e-02 -2.42319293e-02 -3.40699730e-03\n",
      " -3.60555276e-02  4.33389693e-02 -3.48602608e-02  5.27769215e-02\n",
      "  2.89709512e-02 -4.98462208e-02 -1.94748938e-02  1.16397776e-02\n",
      " -3.04615125e-02  8.04637522e-02  6.56247884e-02 -2.84533426e-02\n",
      "  1.81615255e-34 -4.19165473e-03 -2.57882643e-02  5.17320670e-02\n",
      "  4.94420342e-02  1.32476436e-02 -4.21994478e-02 -1.12458421e-02\n",
      " -2.61519291e-02  5.51130101e-02  2.20025145e-02 -2.51169838e-02]\n",
      "\n",
      "Sentence: Learn to use embeddings well and you'll be well on your way to being an AI engineer.\n",
      "Embedding: [-2.20730770e-02  2.08950453e-02 -6.03005365e-02  8.43943749e-03\n",
      "  4.37650904e-02  1.55070275e-02  4.99908179e-02 -3.03232111e-02\n",
      "  4.94783856e-02  2.35512480e-02  3.29350494e-02  1.53877661e-02\n",
      " -6.68355301e-02  1.11002833e-01  6.92677498e-02 -2.31888313e-02\n",
      "  3.79102603e-02 -4.94148023e-03 -1.57800280e-02 -3.45476307e-02\n",
      " -2.65052747e-02 -2.47879792e-02 -1.86141878e-02  3.00361644e-02\n",
      " -2.81185824e-02 -8.75132531e-03 -3.30770528e-03 -2.06116010e-02\n",
      "  1.03315599e-02 -1.51483146e-02 -3.48330885e-02 -2.63248123e-02\n",
      "  2.06908472e-02  3.79108936e-02  1.81912844e-06 -2.44284305e-03\n",
      " -1.80562783e-03  5.61765116e-03 -2.79870331e-02  1.54702654e-02\n",
      "  3.06456462e-02  3.72601114e-02 -1.55611672e-02  2.54414603e-02\n",
      " -6.42072111e-02  3.16353180e-02  6.63442388e-02  3.80970761e-02\n",
      "  5.57844117e-02  5.31659797e-02 -9.69291013e-03 -3.61423567e-02\n",
      "  3.72434445e-02 -4.67829639e-03  5.14575541e-02  1.00058094e-02\n",
      "  4.90282755e-03  1.41563276e-02  4.95099910e-02  3.32950498e-03\n",
      " -3.21100503e-02  4.42387983e-02  3.27416360e-02 -7.90619757e-03\n",
      "  1.07809819e-01  7.32945129e-02  3.36702913e-02 -4.28345688e-02\n",
      "  1.05966143e-02  2.05654055e-02 -2.02670153e-02  1.04964161e-02\n",
      " -1.97616089e-02 -2.89434392e-05 -2.61862390e-02 -1.85173973e-02\n",
      " -3.44269723e-02 -4.08621244e-02  2.32571457e-02  2.14195568e-02\n",
      "  1.31320590e-02 -3.27211581e-02 -1.91425551e-02 -2.86572259e-02\n",
      " -1.16859032e-02  1.21910051e-02  1.05248326e-02 -3.39585282e-02\n",
      "  3.08909197e-03 -4.44888622e-02  2.65104845e-02  1.09536350e-02\n",
      "  2.51450278e-02 -6.48841541e-03  4.54364717e-03 -2.02785116e-02\n",
      " -1.03216199e-02  2.06590313e-02 -1.65314153e-02 -2.45611817e-02\n",
      "  5.47549352e-02  2.68261470e-02  2.95510497e-02  3.86754833e-02\n",
      " -7.76720569e-02  3.80055793e-02 -2.98364665e-02  7.96886161e-02\n",
      " -3.00943404e-02  7.57849962e-03 -6.89826906e-02 -2.92666946e-02\n",
      " -2.35579778e-02  3.48198041e-02  2.52938103e-02 -4.53816988e-02\n",
      " -1.57938730e-02  4.39031124e-02 -4.04336005e-02  8.32531881e-03\n",
      " -2.84665022e-02  4.94934171e-02  2.41276678e-02  3.02192271e-02\n",
      " -4.99590635e-02 -5.94533049e-02 -3.70175838e-02  1.30331032e-02\n",
      " -3.36468965e-02  3.45589556e-02 -1.44523308e-02  2.57640351e-02\n",
      "  4.61183256e-03  2.21551768e-02 -4.93460335e-03  9.66005400e-02\n",
      " -2.72454903e-03  5.65360126e-04 -3.24247815e-02  1.31681636e-02\n",
      "  4.41607460e-02 -7.03052338e-03  6.84261173e-02 -2.28166264e-02\n",
      " -2.81033595e-03 -4.23883386e-02 -1.33632179e-02 -5.96738309e-02\n",
      " -6.96125487e-03 -2.31901109e-02 -3.78851257e-02  9.80186686e-02\n",
      " -2.21728683e-02 -2.30062362e-02  3.22815143e-02  8.21803696e-03\n",
      " -7.04125408e-03  4.84079421e-02  4.23291065e-02 -2.59713852e-03\n",
      "  1.20505894e-04  1.67414024e-02  2.91197933e-02 -1.28740259e-02\n",
      " -2.41077878e-02 -3.29319611e-02 -3.50291003e-03 -3.19321938e-02\n",
      " -2.64171995e-02  2.30404325e-02  1.11637348e-02 -9.96002741e-03\n",
      " -1.75900850e-02 -2.00276845e-03  1.21594938e-02  4.67822738e-02\n",
      "  5.20882159e-02  7.21171647e-02  1.94978528e-02  1.15072541e-02\n",
      "  5.94166014e-03 -1.47834038e-02 -2.87724733e-02  6.72666579e-02\n",
      " -1.68758128e-02  5.25876321e-03 -3.39737758e-02  5.24596162e-02\n",
      " -2.59793270e-02 -4.41379584e-02  1.47452264e-03 -1.06599098e-02\n",
      " -1.51859205e-02 -1.55874272e-03  1.81505345e-02 -4.85411584e-02\n",
      "  3.67905921e-03 -6.59313723e-02 -1.49418358e-02 -3.23529430e-02\n",
      " -2.79949531e-02  1.71856899e-02 -7.87088182e-03  4.65692244e-02\n",
      "  1.47123430e-02 -7.40438476e-02 -6.52104095e-02 -5.22735231e-02\n",
      " -1.82343237e-02  5.20859137e-02  3.06304768e-02 -2.36036908e-02\n",
      "  2.42384989e-02 -1.83939543e-02 -4.83018439e-03 -2.13386621e-02\n",
      "  1.56583544e-02  9.87346098e-03 -4.25561294e-02  6.00463292e-03\n",
      " -3.14500160e-03  4.51509189e-03 -1.52781548e-03  1.13731511e-02\n",
      " -6.96354583e-02 -3.37258019e-02  1.33406790e-02  4.87290090e-03\n",
      " -7.81489722e-03  4.78049293e-02 -1.59712229e-02  3.14606056e-02\n",
      "  5.15920334e-02 -4.05122675e-02 -5.06460667e-02  9.99934506e-03\n",
      " -2.00729817e-02  4.21552770e-02  3.03182397e-02 -1.00431323e-01\n",
      " -4.12019677e-02  3.43989953e-02  3.29208747e-02  1.07411435e-03\n",
      "  3.70961428e-02 -6.94237128e-02  6.52394369e-02  8.31667334e-03\n",
      "  1.68036204e-02 -2.60705873e-02  8.14496726e-03 -1.48009798e-02\n",
      " -2.65671089e-02  3.29321362e-02 -7.37513416e-03  7.23976223e-03\n",
      " -2.69329250e-02  1.71754714e-02 -2.28083450e-02 -4.75334842e-03\n",
      "  2.88568903e-02  1.30763699e-04  5.44128641e-02 -1.43379122e-02\n",
      "  1.89891793e-02 -1.32732214e-02  4.01177444e-02 -7.29275644e-02\n",
      " -2.41210964e-02  3.16216946e-02 -1.68014653e-02  8.47541727e-03\n",
      " -5.23940809e-02 -1.43882902e-02 -1.46156605e-02  6.39906572e-03\n",
      "  2.15113126e-02 -5.18959835e-02 -4.30576093e-02  2.34075431e-02\n",
      "  2.30266433e-03 -2.48433873e-02 -4.38242853e-02 -2.16570888e-02\n",
      " -6.91595301e-02  1.76770054e-02  2.12067626e-02 -2.20293663e-02\n",
      " -1.06773563e-02  9.57430061e-03  2.21988466e-02  5.51470742e-02\n",
      "  1.03682717e-02 -8.14825520e-02 -7.94707611e-03 -1.85866170e-02\n",
      "  1.20494282e-02  7.51403421e-02 -1.41215799e-02  8.83924887e-02\n",
      "  3.12628336e-02  8.12257919e-03 -2.29444765e-02  3.96010354e-02\n",
      " -2.00167615e-02  9.16027650e-02 -2.06838921e-02  5.84990792e-02\n",
      " -4.32368927e-02 -4.74747550e-03 -9.51893814e-03  5.42949978e-03\n",
      "  1.19142031e-04  6.15377054e-02 -1.68790098e-03 -4.66321558e-02\n",
      " -2.01404970e-02  1.32406205e-02  9.70684830e-03  2.73846909e-02\n",
      "  3.06639411e-02  6.71581039e-03  8.71220604e-02 -1.97373237e-03\n",
      "  8.18295684e-03  5.44365076e-03 -5.76205999e-02  1.34821329e-02\n",
      "  5.06284507e-03 -2.10568961e-02  1.25936940e-02 -5.49777551e-03\n",
      " -1.44644817e-02 -2.92567872e-02  5.53298891e-02 -2.60099582e-02\n",
      " -2.82454793e-03 -2.30902322e-02  8.89711455e-03 -2.61565447e-02\n",
      "  9.08590562e-04 -6.16204068e-02 -7.56419152e-02 -1.05932532e-02\n",
      " -1.19563714e-02  6.71458393e-02 -1.96234267e-02 -5.00934161e-02\n",
      " -3.91229130e-02 -3.07008307e-02  7.18906522e-02  9.29243863e-03\n",
      " -6.34386251e-03  7.86956924e-04 -1.36484150e-02  2.87188105e-02\n",
      "  4.01224531e-02  1.28037278e-02  1.77381393e-02 -4.75981506e-03\n",
      "  5.47173731e-02  1.10807794e-03 -2.25790460e-02 -2.80286372e-03\n",
      " -1.13696091e-01  2.55904309e-02  4.00426943e-04 -4.39810529e-02\n",
      "  1.36318980e-02 -1.54137649e-02 -4.99014333e-02 -2.32889466e-02\n",
      " -1.62985502e-03  3.95835228e-02  1.89040247e-02 -3.02703008e-02\n",
      "  2.71438081e-02  1.05684157e-03 -4.21068333e-02  3.71961370e-02\n",
      "  3.54258306e-02 -6.98274672e-02 -2.20937636e-02 -4.01496030e-02\n",
      " -1.90164112e-02 -2.69835647e-02 -1.51212923e-02  3.33365500e-02\n",
      " -9.74889398e-02  1.73102487e-02  6.21016510e-03 -2.59423885e-03\n",
      " -1.10152908e-01 -6.10186085e-02 -1.36549976e-02 -1.35034686e-02\n",
      " -6.72574192e-02 -4.05122805e-03 -6.64987601e-03  3.86571139e-03\n",
      "  9.43475123e-03 -3.86637077e-02 -1.93593334e-02  1.34933963e-02\n",
      " -4.58106697e-02  6.06737286e-02  6.06379844e-02  4.85596210e-02\n",
      " -4.56088893e-02 -5.71936667e-02 -1.55094313e-02  3.40963714e-02\n",
      "  9.48137953e-04 -9.94345453e-03  2.84658018e-02 -3.29024345e-02\n",
      " -2.83211041e-02  3.19907628e-02  2.61299126e-02 -2.74055079e-02\n",
      " -1.36353048e-02  7.47715915e-03  1.19430251e-01 -4.45806980e-02\n",
      "  1.07672177e-02 -8.69424716e-02 -2.19551120e-02  1.83874536e-02\n",
      " -1.06521165e-02 -1.89243089e-02 -3.06514297e-02 -3.04701775e-02\n",
      " -3.22214216e-02  4.12151106e-02  8.95758905e-03 -2.73179207e-02\n",
      "  9.39997379e-03 -9.57167416e-04 -1.94010213e-02 -4.92622294e-02\n",
      " -9.18889046e-03  4.66894209e-02  5.41893691e-02  2.21609641e-02\n",
      " -2.86351983e-02  5.20295240e-02  2.47589573e-02 -7.14267641e-02\n",
      " -1.26206772e-02  7.35520758e-03  2.13785302e-02  2.93514021e-02\n",
      " -2.57651079e-02  5.20562232e-02 -2.74893269e-02 -3.10242623e-02\n",
      " -9.02880505e-02  6.10371493e-02 -5.22609800e-02  2.13111993e-02\n",
      "  4.41735797e-02  3.23370881e-02  1.75648369e-02 -2.39519682e-02\n",
      " -2.69709043e-02  5.11278957e-02  2.69065015e-02 -4.51933183e-02\n",
      "  2.52657337e-03  2.44933944e-02 -2.89541092e-02  2.79992670e-02\n",
      " -1.36022661e-02 -4.32368666e-02  1.85830183e-02  7.63815187e-05\n",
      "  2.43505416e-03 -3.73323751e-03 -1.72279775e-02  1.01292524e-02\n",
      "  1.98437311e-02 -2.60018688e-02 -3.40175000e-03  1.09125990e-02\n",
      " -4.16363552e-02  3.37032452e-02 -2.81635020e-02  1.79126430e-02\n",
      " -4.53095771e-02 -1.09819314e-02 -2.20827595e-03  1.99102350e-02\n",
      "  3.56372185e-02 -3.11799422e-02  3.78753133e-02 -1.41408853e-02\n",
      " -2.16907077e-02  2.73019671e-02  3.69814457e-03  6.35387823e-02\n",
      "  1.22669144e-02 -6.02271082e-03 -7.60751078e-03 -1.86564587e-02\n",
      " -5.64720621e-03 -2.20051059e-03 -1.31825255e-02  1.67724472e-02\n",
      " -3.77264507e-02  2.97894981e-02 -5.01569845e-02  4.89087887e-02\n",
      " -6.07445277e-02 -8.39419737e-02 -5.09000719e-02  1.81768332e-02\n",
      "  6.66732788e-02 -3.30040231e-03 -2.82402639e-03 -5.35405949e-02\n",
      "  3.90341394e-02  2.19852570e-02  3.13555002e-02 -3.36526595e-02\n",
      "  1.96913853e-02  1.67883448e-02  5.04003130e-02  3.08061624e-03\n",
      " -7.24784390e-04  4.42907140e-02 -4.12956253e-03  4.29329090e-02\n",
      " -6.62552565e-02  1.16060267e-03 -2.81716529e-02  1.56886056e-02\n",
      "  9.78133529e-02  5.53594455e-02 -1.39378849e-02  2.12307293e-02\n",
      " -1.30955633e-02 -6.82027489e-02 -8.09109246e-04  4.99291830e-02\n",
      " -2.69266032e-02 -2.97804568e-02  3.84461880e-02  1.97355188e-02\n",
      "  3.37088443e-02  1.65873226e-02  5.77319553e-03 -3.04898042e-02\n",
      " -1.52511774e-02 -3.56159136e-02 -8.69219843e-03 -5.42296723e-33\n",
      "  3.24374111e-03 -3.46329696e-02  3.58932689e-02  1.83770359e-02\n",
      " -2.17505246e-02 -3.26411650e-02  2.88397330e-03  1.50463805e-02\n",
      " -1.75260333e-03 -1.99418794e-02 -6.10355521e-03  2.23846957e-02\n",
      " -8.78934923e-04  2.48684902e-02  3.39736454e-02  2.75593121e-02\n",
      "  3.37792784e-02  3.98564897e-02  2.55545285e-02  1.83043107e-02\n",
      " -2.92878468e-02  5.18079568e-03  8.37757951e-04 -3.66560407e-02\n",
      " -3.46732736e-02  3.82686965e-02  5.50826173e-03 -4.35187593e-02\n",
      "  2.44077537e-02  3.54167968e-02 -2.13442221e-02  2.86623593e-02\n",
      " -2.65397568e-04  3.73409279e-02 -8.68165772e-03  3.04782088e-03\n",
      " -2.71682143e-02 -3.85087803e-02 -6.12388887e-02 -2.00842717e-03\n",
      " -1.22080659e-02 -8.67197812e-02  3.75363277e-03 -1.77707765e-02\n",
      "  8.32477212e-03 -1.69167910e-02  7.02404156e-02  3.32234055e-02\n",
      "  4.34313007e-02  1.47016719e-02 -1.25546858e-01  1.50866546e-02\n",
      " -5.43164276e-02 -1.79147825e-03  4.99600507e-02 -1.53786773e-02\n",
      "  3.32683474e-02 -3.07709239e-02 -1.83896907e-02  9.45789739e-03\n",
      " -4.60291132e-02 -2.03867909e-03  2.62428392e-02 -5.00789061e-02\n",
      "  2.01835595e-02  6.08983189e-02 -2.01180670e-02 -2.60053910e-02\n",
      "  1.05925277e-02 -3.31154354e-02  1.62595753e-02  7.77862146e-02\n",
      " -1.90730614e-03 -5.62884519e-03  1.43716130e-02 -4.06833701e-02\n",
      " -5.14970683e-02  1.66229860e-04 -3.33056273e-03  1.44689241e-02\n",
      "  4.24957077e-04  3.04452889e-02 -1.83636360e-02  1.51190010e-03\n",
      "  2.99861152e-02 -3.68002206e-02  8.35626386e-03 -3.31025571e-02\n",
      "  2.66911928e-02  5.47833275e-03 -1.80523973e-02  2.42576823e-02\n",
      "  5.72715187e-03 -5.93372844e-02  1.04358502e-01 -9.87931248e-03\n",
      " -1.36105726e-02  5.79998791e-02  2.50108391e-02  2.89337225e-02\n",
      " -3.20522115e-02 -3.40233147e-02 -3.41698825e-02 -2.76981499e-02\n",
      "  6.47003129e-02  1.50798038e-02 -1.61925349e-02  3.03265769e-02\n",
      " -2.67188177e-02 -3.67774256e-02 -2.27845367e-02 -5.36433645e-02\n",
      "  1.90500077e-02 -3.42503004e-02  1.32688498e-02 -5.41319279e-03\n",
      "  7.49743683e-03 -7.37026450e-04 -3.08569167e-02  3.82288210e-02\n",
      " -2.08311118e-02 -3.43155526e-02  5.60238399e-03  1.45000033e-02\n",
      " -3.76364067e-02 -5.11782542e-02 -3.51075791e-02  1.71867702e-02\n",
      "  1.50721679e-02 -9.62026268e-02 -1.53545206e-02  1.58376358e-02\n",
      "  2.42940928e-07 -5.88812865e-03  7.68795758e-02  5.86063564e-02\n",
      "  2.21232790e-02 -2.36690678e-02  5.25274165e-02  1.48661500e-02\n",
      "  7.34176859e-03 -4.98920912e-03  4.37413864e-02 -1.28331967e-02\n",
      "  3.37342508e-02 -1.10814739e-02 -1.33937830e-02 -7.80063272e-02\n",
      " -1.36330416e-02  1.94748528e-02  1.91757898e-03 -3.00252102e-02\n",
      "  1.02689613e-04  9.54533368e-02  1.19653925e-01  3.73371951e-02\n",
      "  4.25124634e-03  2.05129944e-02 -3.85414399e-02 -1.90614443e-02\n",
      "  5.88793345e-02  6.81264624e-02 -3.12595516e-02 -6.50440529e-02\n",
      "  2.48045176e-02  3.90052708e-04  7.54762292e-02 -3.46075632e-02\n",
      "  1.32948896e-02  4.14005555e-02  3.07568982e-02  5.50354458e-03\n",
      " -1.53082772e-03  2.75993440e-02  6.46032067e-03  1.05398372e-02\n",
      " -3.09298970e-02  4.60232869e-02 -3.64921056e-02 -1.39540313e-02\n",
      " -3.53720933e-02  7.97828892e-04  1.40632866e-02  1.80258714e-02\n",
      " -1.43368309e-02  2.19214824e-03 -3.96873578e-02 -1.17281973e-02\n",
      " -4.45220396e-02  8.05769674e-03 -4.04861271e-02  3.56148966e-02\n",
      "  5.12852445e-02 -6.64038733e-02 -5.32595143e-02  8.92900117e-03\n",
      "  1.56424697e-02  1.02110893e-01  8.10768548e-03 -4.03861562e-03\n",
      "  2.02352605e-34 -1.38294362e-02 -1.17622903e-02  1.51006151e-02\n",
      "  8.25896040e-02  2.39228457e-02 -1.10378191e-02  3.65656638e-03\n",
      " -7.44783645e-03  2.94555277e-02  3.52993887e-03 -6.10421523e-02]\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# create a list of sentences to turn into numbers\n",
    "sentences = [\n",
    "    \"The Sentences Transformers library provides an easy and open-source way to create embeddings.\",\n",
    "    \"Sentences can be embedded one by one or as a list of strings.\",\n",
    "    \"Embeddings are one of the most powerful concepts in machine learning!\",\n",
    "    \"Learn to use embeddings well and you'll be well on your way to being an AI engineer.\"\n",
    "]\n",
    "\n",
    "# sentences are encoded/embedded by calling model.encode()\n",
    "embeddings = embedding_model.encode(sentences)\n",
    "embeddings_dict = dict(zip(sentences, embeddings))\n",
    "\n",
    "# see the embeddings\n",
    "for sentence, embedding in embeddings_dict.items():\n",
    "    print(\"Sentence:\", sentence)\n",
    "    print(\"Embedding:\", embedding)\n",
    "    print(\"\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sentence: Yo! How cool are embeddings?\n",
      "Embedding: \n",
      "[-1.97447557e-02 -4.51087812e-03 -4.98482119e-03  6.55444711e-02\n",
      " -9.87676159e-03  2.72835623e-02  3.66426148e-02 -3.30224284e-03\n",
      "  8.50076973e-03  8.24948400e-03 -2.28497181e-02  4.02429961e-02\n",
      " -5.75200096e-02  6.33692592e-02  4.43207547e-02 -4.49507385e-02\n",
      "  1.25284195e-02 -2.52012350e-02 -3.55292223e-02  1.29559385e-02\n",
      "  8.67022946e-03 -1.92917287e-02  3.55628273e-03  1.89505890e-02\n",
      " -1.47128049e-02 -9.39845853e-03  7.64168892e-03  9.62188747e-03\n",
      " -5.98928845e-03 -3.90169173e-02 -5.47824614e-02 -5.67457359e-03\n",
      "  1.11645376e-02  4.08067219e-02  1.76319088e-06  9.15296562e-03\n",
      " -8.77261534e-03  2.39382591e-02 -2.32784078e-02  8.04999545e-02\n",
      "  3.19176763e-02  5.12596220e-03 -1.47708524e-02 -1.62524320e-02\n",
      " -6.03212640e-02 -4.35689725e-02  4.51211594e-02 -1.79053564e-02\n",
      "  2.63367072e-02 -3.47867236e-02 -8.89172871e-03 -5.47675341e-02\n",
      " -1.24372784e-02 -2.38606706e-02  8.33496451e-02  5.71242943e-02\n",
      "  1.13329012e-02 -1.49594918e-02  9.20379013e-02  2.72709411e-02\n",
      " -1.42186023e-02  1.91208869e-02  1.49963452e-02 -3.12199369e-02\n",
      "  8.99579599e-02  4.51188460e-02  2.58020442e-02 -5.51741198e-03\n",
      "  1.15910163e-02  4.72100377e-02 -1.51018659e-02  1.70818511e-02\n",
      " -7.22600333e-03  3.45763005e-02 -8.76463205e-03  5.22015952e-02\n",
      " -6.49008676e-02 -4.31378931e-02  6.36964515e-02  4.02881205e-02\n",
      " -1.99043229e-02  5.39062684e-03  1.28820436e-02 -4.81255725e-02\n",
      "  4.58801799e-02 -2.17095390e-02  1.89203415e-02 -3.46344672e-02\n",
      " -1.66458301e-02  7.65167689e-03 -2.26693377e-02 -1.96454134e-02\n",
      "  1.87632199e-02  1.01383384e-02  6.85413256e-02 -5.39849745e-03\n",
      " -3.38231237e-03  4.08413373e-02  4.98623773e-02 -1.16485506e-02\n",
      "  8.91739056e-02  4.02785838e-02 -3.64715816e-03  4.37757745e-02\n",
      " -2.96079498e-02 -5.53745497e-03 -2.00208034e-02 -2.01983526e-02\n",
      "  4.59849313e-02  2.29337215e-02 -5.37305102e-02 -3.19279470e-02\n",
      "  1.37546111e-03  6.25036582e-02 -2.18308959e-02 -6.43255413e-02\n",
      " -2.24791747e-02  3.31955180e-02 -3.12837511e-02  5.17936312e-02\n",
      " -2.84002163e-02  2.55067237e-02  3.36492956e-02  7.50667751e-02\n",
      " -4.46478464e-03 -4.87705208e-02 -7.35219121e-02 -5.46353385e-02\n",
      "  8.88533425e-03  2.95797326e-02 -9.95699223e-03 -6.32763747e-03\n",
      "  4.46259566e-02 -1.58484038e-02 -1.71330329e-02  3.36602665e-02\n",
      " -1.57676684e-03 -5.45971617e-02  2.91161425e-02 -2.80596353e-02\n",
      "  2.96793375e-02  5.12153767e-02  1.48766274e-02 -4.76488955e-02\n",
      "  1.26052229e-02  1.49844482e-03  1.33206593e-02 -2.82468554e-02\n",
      " -3.29254530e-02 -8.53566732e-03 -5.27607128e-02  7.29349926e-02\n",
      " -6.41821325e-02 -2.51785689e-03 -9.02636070e-03 -1.10470084e-03\n",
      "  1.57514773e-02  4.30823602e-02  1.12269586e-02 -3.54585014e-02\n",
      "  4.95163240e-02  1.21271154e-02  1.66341115e-03 -5.06928004e-03\n",
      " -1.11002382e-02 -8.66921898e-03 -3.26441042e-02 -3.98021936e-02\n",
      " -2.05971543e-02  1.09073790e-02 -6.62529543e-02  3.71706970e-02\n",
      " -3.74916308e-02 -3.24731842e-02  5.85899539e-02  8.48080441e-02\n",
      "  3.92413251e-02  3.15813757e-02  3.78386639e-02 -1.35472352e-02\n",
      "  5.95062748e-02  2.58905236e-02 -1.31899798e-02  6.30590320e-02\n",
      "  3.27137485e-02  6.92142872e-03 -1.42607111e-02  7.76677281e-02\n",
      " -1.16103422e-02 -3.66428234e-02 -2.83837561e-02  2.72280257e-02\n",
      "  2.49365233e-02 -4.22294997e-03 -3.63100916e-02 -2.04887893e-02\n",
      "  3.98860462e-02 -2.64727417e-02  4.41383990e-03 -5.19634858e-02\n",
      "  1.71419506e-05  4.81283776e-02  2.04450767e-02  9.84970182e-02\n",
      "  3.68267447e-02  1.53405312e-02  7.50893552e-04 -3.38639170e-02\n",
      " -2.69872788e-02  4.72445525e-02  4.56702486e-02 -3.49246562e-02\n",
      " -1.18770637e-02  3.45577323e-03 -6.32299995e-03 -4.78412546e-02\n",
      "  1.84098855e-02 -2.23157853e-02 -3.70728262e-02  5.87340072e-02\n",
      "  6.22730982e-03 -1.46715697e-02  7.29223490e-02  2.21958430e-03\n",
      " -6.53118715e-02  3.51679325e-02 -1.54901454e-02  6.01421520e-02\n",
      " -9.41006560e-03  2.81196442e-02 -1.12651801e-02  5.24157716e-04\n",
      "  1.01888590e-01 -5.69957234e-02 -3.52360010e-02 -5.20475022e-03\n",
      " -8.46639741e-03  1.39208678e-02  1.80780143e-02 -1.10493943e-01\n",
      "  5.13116643e-02 -4.36432995e-02  2.84142289e-02  9.55939107e-03\n",
      "  4.28096540e-02 -3.95832844e-02  5.25828451e-02  1.92814507e-02\n",
      "  3.44888819e-03 -1.76870655e-02  3.85699198e-02  6.92507392e-03\n",
      " -3.59440707e-02 -2.63613686e-02 -4.96698590e-03  4.24525887e-02\n",
      " -4.22463715e-02  3.45897838e-03  3.55866924e-02 -1.68201961e-02\n",
      "  3.54331806e-02  4.52796416e-03  6.46285107e-03 -2.17710678e-02\n",
      " -2.50033438e-02  1.41418697e-02  1.51257738e-02 -2.99570262e-02\n",
      " -3.94227579e-02  1.87821742e-02 -1.68786338e-03 -9.83518432e-04\n",
      " -3.26321498e-02  5.06566744e-03 -8.90460610e-03 -1.55095719e-02\n",
      "  1.87757872e-02 -4.52472977e-02 -1.72956958e-02  3.50973345e-02\n",
      "  1.33018382e-02  1.00633204e-02 -4.36593331e-02 -1.68619175e-02\n",
      " -1.91048216e-02  6.35489002e-02  8.08330625e-03 -1.02532869e-02\n",
      " -4.53627855e-03 -4.60835844e-02 -1.64704174e-02 -6.24690857e-03\n",
      "  2.58868448e-02 -6.39063790e-02 -7.82399531e-03 -2.36076023e-02\n",
      "  2.74617802e-02  5.80534972e-02 -3.40748541e-02  6.46012127e-02\n",
      "  2.00062972e-02 -2.14935467e-02  1.69361252e-02 -5.54061355e-03\n",
      " -2.40397044e-02  3.09443865e-02 -2.34438945e-03  3.02589964e-02\n",
      " -4.57217507e-02  2.00641807e-02 -2.57117692e-02 -1.13375194e-03\n",
      " -3.57524641e-02  6.92952871e-02  2.80846772e-03  3.58742997e-02\n",
      " -1.52722523e-02 -3.41523364e-02  1.80923082e-02  1.65399592e-02\n",
      "  1.31705729e-02 -6.36677584e-03  5.49303442e-02  8.47317465e-03\n",
      " -4.26077507e-02  2.17085015e-02 -4.89024222e-02  1.18895645e-04\n",
      "  5.16286232e-02  7.59225339e-04  1.59222875e-02 -1.61299501e-02\n",
      "  1.44981062e-02  2.19508708e-02  3.02651133e-02 -3.44152115e-02\n",
      " -4.80379984e-02 -3.71690653e-02  4.68194373e-02 -3.46219353e-02\n",
      "  4.74856148e-04 -4.34820354e-02 -1.80124883e-02 -6.44844621e-02\n",
      " -2.66967900e-02  3.63660641e-02 -3.76219526e-02 -1.64600462e-02\n",
      "  2.20248923e-02  2.16065251e-04  3.64510044e-02 -2.76135728e-02\n",
      " -5.87059744e-03  6.97318837e-03 -1.25865208e-03  2.12773439e-02\n",
      "  6.21468574e-03 -3.57214175e-02 -5.09367548e-02  2.85552610e-02\n",
      "  6.48821816e-02  3.45730968e-02 -2.57281382e-02  4.52558417e-03\n",
      " -4.63605821e-02  3.32225114e-02  1.69979956e-03 -1.29355220e-02\n",
      " -3.03734522e-02  1.23609686e-02 -3.17998871e-04  1.66232251e-02\n",
      "  1.04892328e-02  1.71540566e-02  1.88860241e-02 -3.62256467e-02\n",
      "  4.65737022e-02  2.17129588e-02  4.97537218e-02  3.03067677e-02\n",
      "  1.59921334e-03 -6.30024150e-02 -6.34589605e-03  1.07322223e-04\n",
      " -5.56746777e-03  2.37264726e-02 -8.38226639e-03  5.38897403e-02\n",
      " -9.08976272e-02 -1.37359835e-02  1.18454434e-02  3.37267341e-03\n",
      " -2.81855054e-02  1.56339083e-03  2.22415328e-02  6.21024556e-02\n",
      " -8.68121237e-02 -4.40635579e-03 -1.63995605e-02  1.69664826e-02\n",
      " -1.34548554e-02  3.00700450e-03 -2.14617942e-02 -2.09504068e-02\n",
      " -1.39878048e-02 -1.23850303e-02  5.39979450e-02  6.03615008e-02\n",
      "  2.52982546e-02 -1.29661411e-01 -1.08081609e-01 -4.15772945e-03\n",
      " -7.20267138e-03  2.75885463e-02  4.87621799e-02 -2.95971632e-02\n",
      " -4.32554372e-02  2.75214743e-02  1.35718351e-02  3.87700833e-02\n",
      "  2.42039729e-02 -2.70842165e-02  8.59166309e-02 -1.98402666e-02\n",
      " -2.26344261e-02 -6.24928065e-02 -1.56844649e-02  4.11566719e-02\n",
      "  1.66952517e-02  7.97291473e-02 -3.24838832e-02  1.46566634e-03\n",
      " -3.27905975e-02  6.44814819e-02  2.09739506e-02 -7.56984949e-02\n",
      " -1.31797104e-03  2.24044384e-03 -6.60840189e-03 -6.84252083e-02\n",
      " -5.36862481e-03  6.55135885e-02  5.45567973e-03  1.58993099e-02\n",
      " -2.18051989e-02  2.16415804e-03  4.72960481e-03 -7.25655034e-02\n",
      " -1.59350801e-02 -1.05622821e-02  2.70785931e-02  1.93769834e-03\n",
      " -4.45122793e-02  2.89782975e-02  2.43083779e-02 -1.73885599e-02\n",
      " -3.80669311e-02 -3.06799486e-02 -3.24777849e-02 -4.33423003e-04\n",
      "  2.55846251e-02  2.70478483e-02 -1.72480068e-04 -4.93668136e-04\n",
      " -7.12093860e-02  5.69768548e-02  6.61400855e-02 -3.87267731e-02\n",
      " -1.30683947e-02  1.00662997e-02 -2.17740256e-02  1.92212574e-02\n",
      "  7.66689423e-03 -3.86652350e-02 -1.66109465e-02 -3.41468900e-02\n",
      " -1.04186032e-02  1.75906476e-02 -1.23776672e-02 -1.68434158e-02\n",
      " -2.40113325e-02  4.70195524e-03  4.88457829e-03  4.73663807e-02\n",
      "  4.34111729e-02 -8.08340777e-03 -2.48272568e-02 -1.93977896e-02\n",
      " -3.16525251e-02 -1.56418141e-02 -7.79948710e-03  1.28887836e-02\n",
      "  2.61943452e-02  3.65825393e-03  5.79228699e-02  5.43151945e-02\n",
      " -5.05587235e-02  1.78923889e-03 -2.45471150e-02 -2.47596223e-02\n",
      "  3.60351661e-03  1.94153097e-02 -4.23822999e-02 -1.86906885e-02\n",
      "  2.32946873e-02 -3.17983218e-02 -2.60645133e-02 -5.40357223e-03\n",
      " -3.82070728e-02  2.21719649e-02  1.33360717e-02  5.58053814e-02\n",
      " -3.57716456e-02 -3.54791768e-02  1.27723515e-02  6.80177733e-02\n",
      "  5.37151620e-02  2.54151691e-02 -1.16639109e-02 -1.07512241e-02\n",
      " -9.74425394e-03  7.20505510e-03  9.21905041e-03 -4.73684855e-02\n",
      " -3.89397633e-03  3.11453696e-02  3.62332165e-02 -1.65903177e-02\n",
      " -3.63395289e-02  1.95635054e-02 -2.15059314e-02 -7.04774261e-03\n",
      "  9.13179666e-03 -4.05358188e-02 -3.67075726e-02  1.16995983e-01\n",
      "  1.17913917e-01  8.00503567e-02 -1.61983389e-02 -2.00732686e-02\n",
      " -5.54061234e-02 -7.22410530e-02  2.14557908e-02 -4.46469523e-04\n",
      " -1.42903039e-02  8.35547317e-03  3.34207267e-02  1.42891202e-02\n",
      "  4.62512597e-02 -3.53420377e-02  1.68673582e-02 -2.99746287e-03\n",
      " -5.44997007e-02 -4.80719693e-02  9.47594002e-04 -6.29721654e-33\n",
      " -2.30286829e-02 -2.51127556e-02 -5.35219312e-02 -2.09469888e-02\n",
      " -6.79714652e-03 -4.64016385e-02 -4.49627312e-03  1.65115166e-02\n",
      " -1.12677272e-02  1.33013595e-02 -1.72552485e-02 -1.96653306e-02\n",
      "  5.53233689e-03  1.02775749e-02  9.47301334e-04 -2.24022754e-02\n",
      "  5.30364960e-02  7.77773978e-03 -9.48467944e-03  2.25516129e-02\n",
      " -4.34490805e-03  2.25209147e-02  1.98086947e-02 -7.57428780e-02\n",
      " -4.36685141e-03  2.50830445e-02  2.59393808e-02 -3.07076219e-02\n",
      "  7.04765022e-02  8.63500759e-02 -7.75880367e-02  1.59991737e-02\n",
      " -5.04692085e-02  4.88355197e-02  3.74993822e-03 -6.12885808e-04\n",
      " -3.87277193e-02 -2.32236236e-02 -3.63983549e-02 -5.07025514e-03\n",
      "  1.10517833e-02 -3.26515734e-02  3.68387252e-02 -4.54949997e-02\n",
      " -1.18531089e-03  1.92692224e-03  2.18784083e-02  2.71093063e-02\n",
      " -4.06263247e-02  6.99159727e-02 -7.33281821e-02 -8.15295242e-03\n",
      " -1.42555973e-02  3.78030818e-03  1.20974340e-01 -6.68213218e-02\n",
      "  3.05052288e-02  1.24480268e-02 -4.59293984e-02  1.03872800e-02\n",
      " -3.97978537e-02 -1.33042140e-02 -1.59402639e-02 -4.29347008e-02\n",
      "  4.05275300e-02  7.07073882e-02 -4.50929105e-02 -3.62473577e-02\n",
      " -1.87589228e-02  1.60928387e-02  2.12657824e-02  6.70026243e-02\n",
      "  3.25864255e-02  1.51125509e-02  3.20371725e-02 -1.35436021e-02\n",
      "  2.31780857e-02 -1.13125453e-02  1.23796379e-02 -3.73516902e-02\n",
      "  1.55540649e-03  2.15824321e-02 -3.49441916e-02 -2.97690537e-02\n",
      "  2.32397728e-02 -1.25702145e-02 -1.09432861e-02 -8.87969360e-02\n",
      " -2.20184047e-02 -1.18422601e-02 -5.71083687e-02  3.91809531e-02\n",
      " -1.98827051e-02 -5.59270158e-02  7.60349398e-03  2.23641582e-02\n",
      " -1.86266713e-02  3.79805341e-02 -8.79637664e-04 -5.26199117e-02\n",
      "  2.05066288e-03  1.72815211e-02 -4.84029241e-02 -1.87741071e-02\n",
      "  5.12661380e-08  2.06594784e-02  3.03574465e-02  5.71857858e-03\n",
      " -5.33938259e-02 -2.46520769e-02  1.80341508e-02 -3.39978300e-02\n",
      "  3.46761190e-05 -6.40034601e-02  2.50049401e-02 -2.04111971e-02\n",
      " -2.72041257e-03 -3.55961509e-02  2.71922797e-02  6.48939759e-02\n",
      "  9.83438338e-04 -4.38491106e-02 -4.45297472e-02 -7.44884415e-03\n",
      "  1.15205124e-02 -2.91251205e-03 -2.15495247e-02  2.84243678e-03\n",
      "  4.29399312e-02 -6.09040558e-02 -7.86323845e-03 -3.90125276e-03\n",
      "  2.47718305e-07  7.75813882e-04  7.47736916e-02  1.99315301e-03\n",
      " -6.07218733e-03  3.69211026e-02  2.78420616e-02 -5.64900041e-02\n",
      "  1.61059219e-02 -9.50814970e-03 -2.60848599e-03 -2.45738104e-02\n",
      "  1.91390552e-02  5.08195646e-02  2.61258464e-02 -1.03838079e-01\n",
      " -3.05815171e-02 -3.53344083e-02 -4.07037064e-02 -2.19842177e-02\n",
      " -2.24091802e-02  5.05568050e-02  7.22607076e-02  5.54789379e-02\n",
      "  4.89434749e-02  3.37947858e-03 -6.84759989e-02  7.10320659e-03\n",
      "  3.15669156e-03  4.78091426e-02 -7.19795674e-02 -3.30301337e-02\n",
      "  3.19159776e-02  1.76431250e-03 -4.62790467e-02 -1.96379535e-02\n",
      "  1.67493597e-02  4.73604053e-02 -2.09441483e-02  7.10240752e-03\n",
      "  4.53146398e-02 -4.76523302e-02 -4.74880077e-02  1.00799426e-02\n",
      " -8.39594901e-02  3.36930119e-02 -3.72189730e-02  1.19432611e-02\n",
      " -3.16896848e-02 -1.29722990e-03 -1.55541338e-02  1.81728378e-02\n",
      " -1.49368607e-02 -1.70671716e-02 -4.19716574e-02  3.94657254e-03\n",
      " -2.24301796e-02  2.07292363e-02 -4.61415574e-02  8.50214716e-03\n",
      " -2.56865118e-02 -1.65338106e-02 -1.51846260e-02 -1.00041823e-02\n",
      "  2.19642017e-02  2.61103678e-02  7.31358677e-02 -1.83709692e-02\n",
      "  1.93979481e-34 -7.27935508e-03  5.96493576e-03  4.44310606e-02\n",
      "  4.14822698e-02  1.12917526e-02 -1.93217583e-02  4.41878587e-02\n",
      " -8.93788785e-03  3.61120291e-02 -5.52126169e-02 -2.89572682e-02]\n",
      "Embedding size: (768,)\n"
     ]
    }
   ],
   "source": [
    "single_sentence = \"Yo! How cool are embeddings?\"\n",
    "single_embedding = embedding_model.encode(single_sentence)\n",
    "print(f\"Sentence: {single_sentence}\")\n",
    "print(f\"Embedding: \\n{single_embedding}\")\n",
    "print(f\"Embedding size: {single_embedding.shape}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [],
   "source": [
    "# no matter the size of the text input to our all-mpnet-base-v2 model, it will\n",
    "# be turned into an embedding size of (768,). This value is fixed. So whether a sentence is 1 token\n",
    "# long or 1000 tokens long, it will be truncated/padded with zeros to size 384 and the\n",
    "# turned into an embedding vector of size (768,). Ofcourse, \n",
    "# other embedding models may have different input/ouput shapes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "# how about we add an embedding field to each of our chunk items?\n",
    "# lets start by trying to create embeddings on the cpu, we will time it with the %%time magic to see how long it takes."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "#%%time\n",
    "\n",
    "#embedding_model.to(\"cpu\")\n",
    "\n",
    "#for item in tqdm(pages_and_chunks_over_min_token_len):\n",
    "    #item[\"embedding\"] = embedding_model.encode(item[\"sentence_chunk\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8d1af303677a4a2c90aa8a48969a5164",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/21 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: total: 594 ms\n",
      "Wall time: 2.39 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "embedding_model.to(\"cuda\")\n",
    "\n",
    "for item in tqdm(pages_and_chunks_over_min_token_len):\n",
    "    item[\"embedding\"] = embedding_model.encode(item[\"sentence_chunk\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "print(torch.cuda.is_available())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 46,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu118\n",
      "Requirement already satisfied: torch in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (2.4.0+cu118)\n",
      "Requirement already satisfied: torchvision in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (0.19.0+cu118)\n",
      "Requirement already satisfied: torchaudio in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (2.4.0+cu118)\n",
      "Requirement already satisfied: filelock in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (3.15.4)\n",
      "Requirement already satisfied: typing-extensions>=4.8.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (4.12.2)\n",
      "Requirement already satisfied: sympy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (1.13.1)\n",
      "Requirement already satisfied: networkx in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (3.3)\n",
      "Requirement already satisfied: jinja2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (3.1.4)\n",
      "Requirement already satisfied: fsspec in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch) (2024.5.0)\n",
      "Requirement already satisfied: numpy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torchvision) (1.26.4)\n",
      "Requirement already satisfied: pillow!=8.3.*,>=5.3.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torchvision) (10.3.0)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from jinja2->torch) (2.1.5)\n",
      "Requirement already satisfied: mpmath<1.4,>=1.1.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sympy->torch) (1.3.0)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "%pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu118"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# what if we could go faster? \n",
    "# you see many modern models can handle batched predictions.\n",
    "# this means computing on multiple samples at once.\n",
    "# those are the types of operations where a gpu flourishes!\n",
    "# we can perform batched operations by turning our target text samples\n",
    "# into a single list and then passing that list to our embedding model."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "# turn text chunks into a single list\n",
    "text_chunks = [item[\"sentence_chunk\"] for item in pages_and_chunks_over_min_token_len]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CPU times: total: 422 ms\n",
      "Wall time: 1.37 s\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "tensor([[-0.0583,  0.0158, -0.0479,  ..., -0.0029, -0.0370, -0.0626],\n",
       "        [-0.0050,  0.0587, -0.0420,  ...,  0.0059, -0.0437, -0.0510],\n",
       "        [-0.0163,  0.0235, -0.0360,  ..., -0.0087, -0.0449, -0.0686],\n",
       "        ...,\n",
       "        [-0.0150,  0.0370, -0.0283,  ...,  0.0021, -0.0828, -0.0711],\n",
       "        [-0.0227, -0.0059, -0.0357,  ...,  0.0301, -0.0409, -0.0521],\n",
       "        [-0.0209,  0.0293, -0.0500,  ...,  0.0295, -0.0333, -0.0745]],\n",
       "       device='cuda:0')"
      ]
     },
     "execution_count": 49,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "%%time\n",
    "\n",
    "# embed all texts in batches\n",
    "text_chunk_embeddings = embedding_model.encode(text_chunks, batch_size=32, convert_to_tensor=True)\n",
    "text_chunk_embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# save embeddings to file\n",
    "\n",
    "# since creating embeddings can be a timely process, lets turn our pages_and_chunks_over_min_token_len list of dictionaries into a dataFrame and save it.\n",
    "text_chunks_and_embeddings_df = pd.DataFrame(pages_and_chunks_over_min_token_len)\n",
    "embeddings_df_save_path = \"text_chunks_and_embeddings_df.csv\"\n",
    "text_chunks_and_embeddings_df.to_csv(embeddings_df_save_path, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>sentence_chunk</th>\n",
       "      <th>chunk_char_count</th>\n",
       "      <th>chunk_word_count</th>\n",
       "      <th>chunk_token_count</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Indian Stock Market Prediction using Deep Lear...</td>\n",
       "      <td>2430</td>\n",
       "      <td>349</td>\n",
       "      <td>607.50</td>\n",
       "      <td>[-5.83338365e-02  1.58389956e-02 -4.78904732e-...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>A robust type of neural network equipped to ha...</td>\n",
       "      <td>1398</td>\n",
       "      <td>231</td>\n",
       "      <td>349.50</td>\n",
       "      <td>[-5.00883535e-03  5.87146804e-02 -4.19578291e-...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>It implies that the price predicted by LSTM di...</td>\n",
       "      <td>1442</td>\n",
       "      <td>229</td>\n",
       "      <td>360.50</td>\n",
       "      <td>[-1.62794497e-02  2.35260874e-02 -3.59748304e-...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>GAN has not been applied much to ﬁnancial mark...</td>\n",
       "      <td>429</td>\n",
       "      <td>66</td>\n",
       "      <td>107.25</td>\n",
       "      <td>[-4.02408987e-02  9.56517980e-02 -3.72772142e-...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>LSTM and GAN models to predict the stock price...</td>\n",
       "      <td>1131</td>\n",
       "      <td>214</td>\n",
       "      <td>282.75</td>\n",
       "      <td>[-1.27546098e-02  4.36626449e-02 -3.94201539e-...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   page_number                                     sentence_chunk  \\\n",
       "0            0  Indian Stock Market Prediction using Deep Lear...   \n",
       "1            0  A robust type of neural network equipped to ha...   \n",
       "2            0  It implies that the price predicted by LSTM di...   \n",
       "3            0  GAN has not been applied much to ﬁnancial mark...   \n",
       "4            1  LSTM and GAN models to predict the stock price...   \n",
       "\n",
       "   chunk_char_count  chunk_word_count  chunk_token_count  \\\n",
       "0              2430               349             607.50   \n",
       "1              1398               231             349.50   \n",
       "2              1442               229             360.50   \n",
       "3               429                66             107.25   \n",
       "4              1131               214             282.75   \n",
       "\n",
       "                                           embedding  \n",
       "0  [-5.83338365e-02  1.58389956e-02 -4.78904732e-...  \n",
       "1  [-5.00883535e-03  5.87146804e-02 -4.19578291e-...  \n",
       "2  [-1.62794497e-02  2.35260874e-02 -3.59748304e-...  \n",
       "3  [-4.02408987e-02  9.56517980e-02 -3.72772142e-...  \n",
       "4  [-1.27546098e-02  4.36626449e-02 -3.94201539e-...  "
      ]
     },
     "execution_count": 51,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import saved file and view\n",
    "text_chunks_and_embeddings_df_load = pd.read_csv(embeddings_df_save_path)\n",
    "text_chunks_and_embeddings_df_load.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# document ingestion and embedding creation extensions\n",
    "# one major extension to the workflow above would to functionize it.\n",
    "# or turn it into a script.\n",
    "# take all the functionality we have created and package it into a single process (eg: go from document to embeddings file)\n",
    "# so you could input a document on one end and have embeddings come out the other end. \n",
    "# the hardest part of this is knowing what kind of preprocesing your text may need\n",
    "# before its turned into embeddings.\n",
    "# cleaner text generally means better results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'cuda'"
      ]
     },
     "execution_count": 53,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# similarity search/ semantic search/ vector search\n",
    "\n",
    "# import embeddings we created and prepare them for use by turning\n",
    "# them into a tensor.\n",
    "\n",
    "import random\n",
    "\n",
    "import torch\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "\n",
    "device = \"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
    "device"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([21, 768])"
      ]
     },
     "execution_count": 54,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# import texts and embedding df\n",
    "text_chunks_and_embeddings_df = pd.read_csv(\"text_chunks_and_embeddings_df.csv\")\n",
    "\n",
    "# convert embedding column back to np.array (it got converted to string when it got saved to csv)\n",
    "text_chunks_and_embeddings_df[\"embedding\"] = text_chunks_and_embeddings_df[\"embedding\"].apply(lambda x: np.fromstring(x.strip(\"[]\"), sep=\" \"))\n",
    "\n",
    "# convert texts and embedding df to list of dicts\n",
    "pages_and_chunks = text_chunks_and_embeddings_df.to_dict(orient=\"records\")\n",
    "\n",
    "# convert embeddings to torch tensor and send to device (note: numpy arrays are float64, torch tensors are float32 by default)\n",
    "embeddings = torch.tensor(np.array(text_chunks_and_embeddings_df[\"embedding\"].tolist()), dtype=torch.float32).to(device)\n",
    "embeddings.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>page_number</th>\n",
       "      <th>sentence_chunk</th>\n",
       "      <th>chunk_char_count</th>\n",
       "      <th>chunk_word_count</th>\n",
       "      <th>chunk_token_count</th>\n",
       "      <th>embedding</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>Indian Stock Market Prediction using Deep Lear...</td>\n",
       "      <td>2430</td>\n",
       "      <td>349</td>\n",
       "      <td>607.50</td>\n",
       "      <td>[-0.0583338365, 0.0158389956, -0.0478904732, -...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0</td>\n",
       "      <td>A robust type of neural network equipped to ha...</td>\n",
       "      <td>1398</td>\n",
       "      <td>231</td>\n",
       "      <td>349.50</td>\n",
       "      <td>[-0.00500883535, 0.0587146804, -0.0419578291, ...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0</td>\n",
       "      <td>It implies that the price predicted by LSTM di...</td>\n",
       "      <td>1442</td>\n",
       "      <td>229</td>\n",
       "      <td>360.50</td>\n",
       "      <td>[-0.0162794497, 0.0235260874, -0.0359748304, 0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0</td>\n",
       "      <td>GAN has not been applied much to ﬁnancial mark...</td>\n",
       "      <td>429</td>\n",
       "      <td>66</td>\n",
       "      <td>107.25</td>\n",
       "      <td>[-0.0402408987, 0.095651798, -0.0372772142, -0...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1</td>\n",
       "      <td>LSTM and GAN models to predict the stock price...</td>\n",
       "      <td>1131</td>\n",
       "      <td>214</td>\n",
       "      <td>282.75</td>\n",
       "      <td>[-0.0127546098, 0.0436626449, -0.0394201539, -...</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   page_number                                     sentence_chunk  \\\n",
       "0            0  Indian Stock Market Prediction using Deep Lear...   \n",
       "1            0  A robust type of neural network equipped to ha...   \n",
       "2            0  It implies that the price predicted by LSTM di...   \n",
       "3            0  GAN has not been applied much to ﬁnancial mark...   \n",
       "4            1  LSTM and GAN models to predict the stock price...   \n",
       "\n",
       "   chunk_char_count  chunk_word_count  chunk_token_count  \\\n",
       "0              2430               349             607.50   \n",
       "1              1398               231             349.50   \n",
       "2              1442               229             360.50   \n",
       "3               429                66             107.25   \n",
       "4              1131               214             282.75   \n",
       "\n",
       "                                           embedding  \n",
       "0  [-0.0583338365, 0.0158389956, -0.0478904732, -...  \n",
       "1  [-0.00500883535, 0.0587146804, -0.0419578291, ...  \n",
       "2  [-0.0162794497, 0.0235260874, -0.0359748304, 0...  \n",
       "3  [-0.0402408987, 0.095651798, -0.0372772142, -0...  \n",
       "4  [-0.0127546098, 0.0436626449, -0.0394201539, -...  "
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "text_chunks_and_embeddings_df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([-5.8334e-02,  1.5839e-02, -4.7890e-02, -2.5234e-02, -1.9885e-02,\n",
       "        -1.4428e-02,  2.4676e-02,  1.3173e-02,  8.2875e-03,  8.0184e-03,\n",
       "         3.1326e-02, -2.0393e-02, -4.4198e-02,  7.7428e-02,  1.8015e-02,\n",
       "        -3.9796e-02,  3.0903e-03, -2.2387e-02,  2.4959e-02, -4.2340e-02,\n",
       "        -1.2606e-02, -4.3285e-02,  4.0382e-02, -1.5585e-02, -1.2439e-02,\n",
       "        -1.6579e-02, -4.5618e-03, -7.3461e-03, -4.2320e-03, -4.8279e-02,\n",
       "        -1.9721e-03, -5.0935e-02,  3.8436e-02,  4.4508e-02,  2.1387e-06,\n",
       "        -1.2472e-02,  1.1335e-02, -1.4419e-03, -1.1654e-02, -1.4956e-02,\n",
       "        -8.8225e-03,  1.2399e-03,  3.9052e-02,  1.5295e-02, -7.9618e-02,\n",
       "        -2.3329e-02,  1.0824e-02,  7.9838e-02,  3.2771e-02,  9.2618e-02,\n",
       "         1.4225e-03, -7.4517e-02,  4.4374e-02, -5.7742e-02, -8.0691e-03,\n",
       "        -1.0166e-01,  9.1773e-03,  2.3898e-02, -4.3723e-02, -3.9482e-02,\n",
       "        -2.7414e-02,  5.1684e-02,  6.4842e-02,  1.6136e-02,  2.6487e-02,\n",
       "         1.4429e-02,  4.4059e-02,  3.0365e-02,  2.6624e-03,  9.8591e-03,\n",
       "        -4.0306e-02,  8.5023e-03,  2.7525e-02,  1.2801e-03, -4.2965e-02,\n",
       "         1.3810e-03, -5.3294e-02,  1.7597e-02, -3.3801e-02,  4.7601e-02,\n",
       "         4.3650e-03, -6.4924e-03,  9.9265e-03, -2.0348e-02, -3.8258e-03,\n",
       "         7.3670e-03, -2.4014e-02, -4.8479e-03,  4.0676e-02, -1.3365e-02,\n",
       "        -1.3513e-02, -2.5627e-02, -2.8099e-02,  1.8774e-02, -2.0093e-02,\n",
       "        -9.2774e-04,  3.1383e-02, -1.7295e-02, -2.3610e-02,  5.9621e-04,\n",
       "        -3.3763e-02, -1.1947e-02,  2.3719e-02,  1.4906e-02, -8.5736e-02,\n",
       "         3.4060e-02, -2.6096e-02,  1.9967e-02,  3.5196e-02,  8.0418e-03,\n",
       "         9.0350e-03, -6.6188e-03,  1.3824e-03,  7.8310e-02,  2.4558e-02,\n",
       "         7.8474e-03, -1.1036e-01, -1.7292e-02, -2.9085e-02,  1.5249e-02,\n",
       "         3.7627e-02,  1.6551e-02, -6.4112e-03, -4.8695e-03, -1.4578e-02,\n",
       "        -5.7873e-02, -7.9807e-02,  4.2515e-02,  1.3288e-02, -3.5839e-02,\n",
       "         2.6738e-02,  2.9748e-03, -5.8538e-02,  3.0645e-02, -4.2246e-03,\n",
       "         6.0076e-02,  1.5270e-02,  3.4592e-02, -1.0383e-01, -4.9512e-02,\n",
       "         1.3557e-02, -2.6772e-02, -1.7141e-03, -1.7650e-02, -3.0501e-02,\n",
       "        -7.0201e-02, -2.5667e-02,  1.2944e-02, -1.1018e-02, -3.4126e-02,\n",
       "         2.1249e-03,  5.8349e-02,  2.8727e-02, -3.5832e-02,  5.3085e-02,\n",
       "        -3.1387e-03,  4.0877e-02,  1.2194e-02,  3.1678e-02,  5.5830e-02,\n",
       "         4.9057e-02, -1.4449e-02,  1.0503e-03,  9.4699e-03, -7.6162e-03,\n",
       "        -8.7893e-02, -5.9841e-02, -6.3438e-03, -2.9539e-02, -2.7646e-02,\n",
       "        -1.1880e-02, -2.3435e-02,  5.2476e-03, -4.3896e-02,  2.9890e-02,\n",
       "         1.9343e-02,  7.4896e-02,  8.3429e-02,  8.0651e-03, -2.3046e-02,\n",
       "         1.2301e-02, -2.8644e-02,  1.1205e-02, -3.1340e-03,  1.1367e-03,\n",
       "         4.5306e-04, -1.8989e-02, -5.6971e-03, -5.2616e-02, -1.7533e-02,\n",
       "         2.5728e-02, -4.1790e-02, -1.5685e-02, -3.7099e-02,  1.3551e-02,\n",
       "        -7.4928e-02, -3.9998e-02, -2.0721e-02,  2.8053e-02,  5.4757e-02,\n",
       "        -2.1590e-03, -2.2107e-02,  5.8553e-02, -3.2943e-02, -3.7033e-03,\n",
       "        -2.5049e-02,  2.7755e-02, -9.1840e-03, -4.8352e-02, -5.7240e-02,\n",
       "         1.2120e-02,  2.4693e-02, -6.8474e-02, -4.2526e-02, -1.5902e-02,\n",
       "         1.4455e-02, -1.4401e-02,  3.3366e-02, -6.6342e-02,  4.5828e-02,\n",
       "         1.5743e-02, -2.6403e-02,  8.7371e-03,  9.3685e-03,  1.2748e-03,\n",
       "        -1.0859e-02,  2.7780e-02, -3.0591e-03,  3.1945e-02,  3.3945e-02,\n",
       "         3.7823e-02, -6.0387e-02,  1.5004e-02, -5.0486e-03, -4.3735e-02,\n",
       "         2.6760e-03, -8.4687e-02,  2.8069e-02, -7.6618e-02, -3.4907e-02,\n",
       "         5.9787e-02, -4.0059e-03, -2.0177e-02, -5.7670e-02,  4.9818e-02,\n",
       "        -2.7807e-02, -6.0427e-02, -2.8621e-02, -4.9569e-02,  6.3796e-03,\n",
       "         3.6073e-02, -1.9230e-03,  2.8956e-02,  3.9507e-02,  1.1562e-02,\n",
       "        -9.2006e-03,  3.8497e-02,  5.5719e-02, -5.2746e-02,  2.5489e-02,\n",
       "         9.1830e-04, -3.3125e-02, -6.4394e-03,  1.8132e-02, -4.2360e-02,\n",
       "         7.6229e-03,  1.8253e-02, -2.4526e-02, -3.5108e-02, -1.1202e-02,\n",
       "         3.2810e-02, -7.0872e-04,  1.1232e-02, -6.3074e-03, -2.0818e-02,\n",
       "        -2.3369e-02,  4.8393e-02,  1.7219e-02, -6.8821e-02,  6.2147e-02,\n",
       "         1.5834e-03, -2.7817e-02,  2.8094e-03, -2.7779e-02, -1.2831e-02,\n",
       "         4.0119e-02,  5.5981e-02, -2.7536e-02, -1.0083e-02, -4.0097e-02,\n",
       "         5.0938e-03,  1.6810e-02, -1.8660e-02, -1.5549e-02, -5.7234e-02,\n",
       "        -1.1924e-02,  8.4680e-06,  9.7806e-02,  5.2042e-02, -3.4646e-03,\n",
       "        -4.9764e-02,  2.1589e-02, -1.4004e-02,  9.5303e-03, -5.4140e-03,\n",
       "         2.9708e-02,  6.8202e-03, -3.5916e-02,  5.5314e-03,  6.0485e-02,\n",
       "         1.8166e-02,  5.5175e-03, -1.9665e-02,  4.6688e-03, -1.4786e-02,\n",
       "         2.6393e-03, -1.8120e-02,  4.8758e-02, -2.4204e-02,  2.8547e-02,\n",
       "         2.3482e-02, -8.2442e-02,  1.2447e-02,  2.5771e-02, -4.2022e-02,\n",
       "         1.1867e-02, -4.6639e-02, -4.0284e-02,  6.2093e-03, -7.1375e-03,\n",
       "        -2.4432e-02, -2.1924e-02, -4.7193e-02, -3.0140e-02,  2.4575e-02,\n",
       "        -2.2478e-02, -3.4409e-02,  1.7797e-02, -4.0742e-03,  8.3209e-03,\n",
       "         2.2172e-02,  6.4549e-02, -3.5143e-02, -2.1610e-02,  2.3512e-02,\n",
       "         1.9219e-02,  4.8466e-02, -1.1789e-02,  4.3972e-02,  6.8457e-02,\n",
       "        -3.0875e-02,  5.8105e-04, -7.6143e-03, -1.6089e-02,  3.0555e-02,\n",
       "         4.1235e-02,  3.8529e-02, -5.6695e-02, -1.8334e-02,  6.9373e-03,\n",
       "         9.1386e-02, -2.8058e-02,  5.9367e-02,  6.0963e-02, -4.5610e-02,\n",
       "         9.1820e-03, -3.2657e-02,  8.8289e-03,  1.1120e-02,  5.8065e-02,\n",
       "        -3.5718e-02, -1.5781e-02,  5.6421e-03,  5.2080e-02, -2.1525e-02,\n",
       "        -1.3322e-02, -2.2831e-03,  4.4100e-02, -2.8747e-02, -3.0473e-02,\n",
       "         1.6320e-02, -5.5533e-02,  2.7084e-02, -1.3614e-02,  3.5573e-02,\n",
       "         6.1459e-03, -5.4025e-02, -4.7340e-03, -7.9317e-02,  4.9036e-02,\n",
       "        -2.7292e-02,  4.4272e-03, -6.7228e-03, -3.2559e-02, -8.4721e-03,\n",
       "        -4.2761e-02,  1.5747e-03,  4.8496e-04, -1.6948e-02,  1.9338e-02,\n",
       "         6.3761e-02, -1.2308e-01,  1.1866e-02, -4.3284e-03, -4.4528e-02,\n",
       "        -6.6656e-02,  1.7606e-02,  8.2715e-04, -7.4087e-02, -7.1834e-03,\n",
       "        -4.3667e-02, -2.0174e-02,  4.4180e-02,  4.3177e-03,  7.9422e-02,\n",
       "         3.7669e-03, -2.0739e-02,  5.3244e-02,  8.1581e-02,  3.2902e-02,\n",
       "         6.9898e-02,  2.0701e-02,  2.7043e-03, -7.8566e-03, -1.0069e-01,\n",
       "         1.0563e-02,  1.8231e-02,  1.2664e-03, -2.8132e-03,  9.6432e-04,\n",
       "         1.5503e-02, -6.7187e-02, -1.8554e-02,  8.0123e-02, -5.7241e-02,\n",
       "        -4.2726e-04,  2.3875e-02,  1.2416e-02,  7.8321e-03, -6.1970e-03,\n",
       "        -3.1164e-02, -1.3565e-02,  6.4770e-02,  3.2926e-02,  4.4997e-02,\n",
       "        -1.9744e-03,  1.3868e-02, -3.9377e-02, -4.5085e-02,  2.2614e-02,\n",
       "         2.9566e-02, -2.6370e-02,  2.6774e-02,  4.2217e-02, -2.6599e-02,\n",
       "        -4.3727e-02, -1.9062e-02, -5.8217e-02, -2.2352e-02,  3.7360e-02,\n",
       "         1.5810e-02,  1.2547e-02, -1.3403e-02, -4.1472e-02,  1.1579e-02,\n",
       "         7.2252e-02,  2.1878e-02,  2.4138e-02, -5.4807e-02,  1.6635e-03,\n",
       "         4.5286e-03, -7.1496e-03,  1.5890e-02,  3.5809e-03,  5.5086e-02,\n",
       "        -2.7972e-02, -4.8548e-02, -1.9691e-02,  1.3820e-02, -6.5902e-02,\n",
       "         1.1917e-02,  5.7879e-02,  5.1642e-02, -3.3112e-02, -2.4511e-03,\n",
       "         1.2612e-02,  7.2279e-02,  8.9911e-03,  3.3691e-03, -2.8205e-02,\n",
       "         9.1694e-02,  1.4260e-02, -2.7705e-02,  5.3718e-02,  7.6666e-02,\n",
       "        -8.4142e-02, -5.9159e-02,  3.8511e-02, -1.1102e-02,  9.9410e-02,\n",
       "        -4.3536e-02,  1.4120e-02, -8.6309e-03, -6.8790e-02,  4.4997e-02,\n",
       "         2.4502e-02,  1.1417e-02,  4.6377e-02, -1.0118e-02,  2.8872e-02,\n",
       "         4.4533e-03, -3.8201e-02,  2.7539e-03, -5.5479e-02, -1.7999e-02,\n",
       "         2.3244e-02,  1.0746e-03, -2.7730e-02, -5.6682e-02, -3.2176e-02,\n",
       "         7.6727e-02,  2.8480e-02,  3.7722e-02, -2.4181e-02,  8.8300e-03,\n",
       "         3.9961e-02, -5.4238e-03, -7.1468e-02, -1.5210e-02,  4.1910e-02,\n",
       "         3.7744e-02,  1.1658e-02, -1.6827e-02,  1.1048e-03, -2.7730e-02,\n",
       "         1.9477e-02,  6.1008e-02, -2.8059e-02,  3.3822e-02, -5.2404e-03,\n",
       "         3.2664e-03, -3.5091e-03,  5.2878e-02, -4.0961e-02,  7.9178e-02,\n",
       "        -3.3991e-03, -4.3267e-03,  5.7074e-02, -1.3208e-02,  8.9476e-02,\n",
       "         5.0340e-02,  8.9118e-03,  1.3699e-02,  1.2841e-02, -8.8883e-02,\n",
       "        -5.3563e-33,  1.7663e-02, -1.5404e-02,  2.7557e-03, -6.5133e-03,\n",
       "        -2.0902e-02,  1.7173e-02, -3.3840e-02, -4.9700e-02,  7.0341e-03,\n",
       "        -1.5366e-02,  4.5448e-03,  3.9022e-02,  1.5852e-03,  3.6615e-02,\n",
       "         3.6732e-02, -6.8714e-02,  2.1316e-02,  3.5064e-03,  9.0545e-03,\n",
       "        -1.0591e-02, -2.0533e-03,  6.4551e-02,  4.7186e-02,  3.5525e-02,\n",
       "        -3.8568e-02, -2.0477e-02, -3.0283e-02, -4.0950e-02,  1.3166e-02,\n",
       "         4.4984e-04,  9.5687e-03, -4.2803e-03,  2.1211e-02, -7.3160e-02,\n",
       "         2.3672e-02,  1.8204e-02,  1.1076e-03, -1.4717e-02,  1.1373e-02,\n",
       "         4.2131e-02, -2.5084e-02,  5.5809e-02,  8.1000e-02, -1.5468e-03,\n",
       "         1.4319e-03,  2.3442e-02,  5.1768e-02,  5.5819e-02,  1.3033e-02,\n",
       "        -5.4238e-02, -2.4662e-02,  3.5454e-03,  2.5112e-03,  3.8916e-02,\n",
       "         2.7309e-02,  5.9183e-02,  4.0622e-02, -1.4064e-02, -5.8079e-02,\n",
       "         1.1355e-02,  9.9685e-03,  1.7939e-02, -5.7024e-02,  4.4820e-02,\n",
       "         1.9256e-02,  2.3260e-02,  7.4276e-03, -1.3818e-02,  1.5770e-02,\n",
       "         2.8271e-02,  6.7421e-03,  2.2270e-02, -7.9791e-03, -2.7856e-02,\n",
       "         7.1387e-02, -7.5109e-02, -3.2987e-02, -2.6278e-03,  3.2379e-03,\n",
       "         5.5321e-02,  4.6678e-03,  2.2388e-02,  8.9887e-04, -1.5974e-02,\n",
       "         5.0952e-02, -6.2975e-02,  2.2302e-02, -1.5972e-02,  2.7953e-02,\n",
       "        -4.2083e-03, -1.3679e-02,  4.9719e-02, -2.9585e-02, -2.6301e-02,\n",
       "        -8.7375e-04, -2.1104e-02,  5.4193e-02,  4.0392e-02, -4.0194e-03,\n",
       "        -5.8300e-02,  1.1644e-02, -5.0610e-02, -1.3771e-02, -2.4708e-02,\n",
       "         4.0451e-04, -3.3793e-02,  1.3452e-02,  8.6288e-03,  8.7992e-03,\n",
       "        -4.4108e-02,  1.7865e-02,  1.2484e-02,  2.5338e-02, -7.9079e-03,\n",
       "         2.4510e-02, -3.1162e-02, -3.3805e-02,  3.2119e-02, -7.1936e-03,\n",
       "         3.9270e-02, -1.4077e-02, -1.9608e-02,  3.0249e-02,  2.6918e-03,\n",
       "         1.6348e-02,  1.8695e-02,  6.6909e-04,  1.0319e-02, -3.7051e-02,\n",
       "        -7.0366e-02, -1.0007e-02,  5.0852e-02,  2.8906e-07, -3.8874e-03,\n",
       "         4.8768e-02,  2.8317e-02,  1.8493e-02,  3.2855e-02, -1.5249e-02,\n",
       "        -1.7964e-02, -2.4225e-02, -1.1050e-02, -1.7250e-02,  7.6559e-02,\n",
       "        -5.6025e-03,  1.9027e-02,  3.0956e-02, -3.7723e-02, -3.9381e-02,\n",
       "        -1.8042e-02,  7.6257e-02, -4.2299e-02,  4.0781e-02, -4.7865e-02,\n",
       "         3.0273e-02,  1.7258e-02, -2.4764e-02,  3.4612e-02, -4.6776e-02,\n",
       "         1.7685e-03, -1.4867e-02,  2.8038e-02,  1.3520e-02,  3.7251e-02,\n",
       "         4.3461e-02, -1.6697e-02, -4.1953e-03,  4.9939e-03,  3.0681e-02,\n",
       "         2.8040e-02, -2.4118e-02,  5.9867e-03,  4.0798e-02, -1.8108e-02,\n",
       "         1.0897e-02,  2.0495e-02, -4.6972e-02,  3.8357e-02, -5.4855e-02,\n",
       "        -5.1507e-02, -2.3968e-04, -5.6918e-02,  1.6926e-02,  2.3437e-02,\n",
       "         1.3010e-02,  2.1695e-02,  1.2499e-02, -1.6646e-02, -5.3870e-02,\n",
       "         4.0679e-02, -3.1636e-02, -1.7440e-02,  5.3094e-02, -6.5308e-02,\n",
       "         2.7760e-02, -2.6846e-02, -6.3330e-03,  4.3792e-02, -6.7927e-03,\n",
       "        -1.1324e-02,  2.4370e-34,  5.3651e-02, -1.6185e-02, -3.2694e-02,\n",
       "         1.9976e-02,  3.6043e-02,  2.5052e-03,  1.9488e-02,  2.2122e-02,\n",
       "        -2.8980e-03, -3.7016e-02, -6.2576e-02], device='cuda:0')"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "embeddings[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[-0.0583,  0.0158, -0.0479,  ..., -0.0029, -0.0370, -0.0626],\n",
       "        [-0.0050,  0.0587, -0.0420,  ...,  0.0059, -0.0437, -0.0510],\n",
       "        [-0.0163,  0.0235, -0.0360,  ..., -0.0087, -0.0449, -0.0686],\n",
       "        ...,\n",
       "        [-0.0150,  0.0370, -0.0283,  ...,  0.0021, -0.0828, -0.0711],\n",
       "        [-0.0227, -0.0059, -0.0357,  ...,  0.0301, -0.0409, -0.0521],\n",
       "        [-0.0209,  0.0293, -0.0500,  ...,  0.0295, -0.0333, -0.0745]],\n",
       "       device='cuda:0')"
      ]
     },
     "execution_count": 57,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "embeddings"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\transformers\\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "# make another instance of our embedding model\n",
    "\n",
    "from sentence_transformers import util, SentenceTransformer\n",
    "\n",
    "embedding_model = SentenceTransformer(model_name_or_path=\"all-mpnet-base-v2\", device=device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "# embedding model ready!\n",
    "# time to perform a semantic search.\n",
    "# lets say you were studying lstm\n",
    "# and wanted to search your textbook for \"lstm neural network\"\n",
    "# well, we can do so with the following steps.\n",
    "# 1. define a query string (eg: \"lstm neural network\") note: this could be anything, specific or not.\n",
    "# 2. turn the query string in an embedding with same model we used to embed our text chunks.\n",
    "# 3. perform a dot product or cosine similarity function between the text embeddings and the query embedding\n",
    "#    (we'll get to what these are shortly) to get similarity scores,\n",
    "# 4. sort the results from step 3 in descending order (a higher score means more similarity in the eyes of the model)\n",
    "#    and use these values to inspect the texts."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Query: lstm neural network\n",
      "Time taken to get scores on 21 embeddings: 0.00399 seconds.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "torch.return_types.topk(\n",
       "values=tensor([0.5522, 0.5183, 0.5035, 0.5023, 0.4935], device='cuda:0'),\n",
       "indices=tensor([15, 12,  4, 14, 11], device='cuda:0'))"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# 1. define the query\n",
    "query = \"lstm neural network\"\n",
    "print(f\"Query: {query}\")\n",
    "\n",
    "# 2. embed the query to the same numerical space as the text examples.\n",
    "# it is important to embed your query with the same model you embedded your examples with.\n",
    "query_embedding = embedding_model.encode(query, convert_to_tensor=True)\n",
    "\n",
    "# 3. get similarity score with the dot product (we'll time this for fun)\n",
    "from time import perf_counter as timer\n",
    "\n",
    "start_time = timer()\n",
    "dot_scores = util.dot_score(a=query_embedding, b=embeddings)[0]\n",
    "end_time = timer()\n",
    "\n",
    "print(f\"Time taken to get scores on {len(embeddings)} embeddings: {end_time-start_time:.5f} seconds.\")\n",
    "\n",
    "# 4. get the top-k results (we'll keep this to 5)\n",
    "top_results_dot_product = torch.topk(dot_scores, k=5)\n",
    "top_results_dot_product"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 0.00883 seconds to perform a dot product comparison across 21 embeddings on my machine.\n",
    "# gpus are optimized for these kind of operations.\n",
    "# so even if you were to increase our embeddings by 100x (21 -> 2100), an exhaustive dot product\n",
    "# operation would happen in 0.883 seconds (assuming linear scaling)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "metadata": {},
   "outputs": [],
   "source": [
    "# for larger datasets we would like to look at a vector database/indexing libraries,\n",
    "# such as Faiss.\n",
    "\n",
    "# lets check the results of our original similarity search\n",
    "# torch.topk returns a tuple of values (scores) and indices of those scores.\n",
    "# the indices relate to which indices in the embeddings tensor have what scores in relation to \n",
    "# the query embedding (higher is better).\n",
    "# we can use those indices to map back to our text chunks.\n",
    "# first, we will define a small helper function to print our wrapped text (so it doesn't print out a whole text chunk as a single line.)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define a helper function to print wrapped text.\n",
    "import textwrap\n",
    "\n",
    "def print_wrapped(text, wrap_length=80):\n",
    "    wrapped_text = textwrap.fill(text, wrap_length)\n",
    "    print(wrapped_text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "# now we can loop through the top_results-dot_product tuple and match up the\n",
    "# scores and indices and then use those indices to index on our pages_and_chunks variable to get the relevant text chunk."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Query: 'lstm neural network'\n",
      "\n",
      "Results: \n",
      "Score: 0.5522\n",
      "Text:\n",
      "C. 210 days Look-back We could test the model LSTM only on M ∈{400, 700} and N =\n",
      "300 as 210 days look-back demands more sophis- ticated computational resources.\n",
      "The result is presented in Tables IX and X. The best result is obtained with M =\n",
      "700 and N = 300 whereas there is no improvement for DPA.1218 Authorized licensed\n",
      "use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC\n",
      "from IEEE Xplore. Restrictions apply.\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.5183\n",
      "Text:\n",
      "TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL Internal Units Learning rate\n",
      "Dropout Rate Batch Size Epochs 495 0.00385 0.35302 60 600 Nevertheless,\n",
      "minimizing adversarial loss solely cannot assure satisfying predictions. For\n",
      "example, G might produce samples to “confuse” D, without being similar to ¯ YT\n",
      "+1, and then D will train to distinguish these samples, directing G to generate\n",
      "other “confusing” samples, and this futile process will continue. To ﬁx this\n",
      "issue, the generative model G must decrease the prediction error loss; that is,\n",
      "Lp loss LP ( ¯ Y , ˆ Y ) =| ˆ Y −¯ Y | (8) Besides, direction of stock price\n",
      "prediction is also critical for trading, so we deﬁne direction prediction loss\n",
      "function Ldpl as: Ldpl( ¯ Y , ˆ Y ) =| sgn( ˆ YT +1 −YT ) −sgn(YT +1 −YT ) | (9)\n",
      "where sgn represents sign function. We combined all these losses deﬁned, and\n",
      "achieve the ﬁnal loss on G: LG(X, Y ) = LG adv( ˆ Y ) + LP ( ¯ Y , ˆ Y ) + Ldpl(\n",
      "¯ Y , ˆ Y ) (10) We perform one SGD step on G to minimize LG(X, Y) while keeping\n",
      "the weights of D ﬁxed. The values of the hyper- parameters for G are mentioned\n",
      "in Table III.b) Discriminator: The Discriminator is a simple dense layer of\n",
      "leakyRelu activation functions followed by a sigmoid function layer. The only\n",
      "job of D is to determine if the input is ¯ Y (real) or ˆ Y (generated), our\n",
      "target loss here is equal to the adversarial loss for D. We keep the weights of\n",
      "G ﬁxed, and perform one SGD iteration on D to reduce the target loss: LD(X, Y )\n",
      "= LD adv( ¯ Y , ˆ Y ) = Lsce(D( ˆ Y ), 0) + Lsce(D( ¯ Y ), 1) (11) The values of\n",
      "the hyper-parameters for D are mentioned in Table IV. F. Evaluation Metrics\n",
      "Based on a speciﬁc model, for every stock at each time t, a prediction is\n",
      "obtained for the next time point t+1. Assuming the total number of time points\n",
      "to be tested is T0; we employed the following measure to gauge the performance\n",
      "of the different models.1) Root Mean Squared Relative Error (RMSRE): RMSRE = v u\n",
      "u t 1 T0 T0 X t=1 ˆ Yt+1 −Yt+1 Yt+1 !\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.5035\n",
      "Text:\n",
      "LSTM and GAN models to predict the stock prices on the Indian market and compare\n",
      "their performance. A. LSTM LSTM can be trained to learn dependencies ranging\n",
      "over very long time intervals of time. It beats the vanishing gradients problem\n",
      "faced by a general RNN by substituting the ordinary neuron by a complex\n",
      "architecture called the LSTM unit or block [3]. The working of LSTM can be\n",
      "summarised by the following set of equations.zt = tanh(W zxt + Rzht−1 + bz) it =\n",
      "σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo)\n",
      "st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate\n",
      "and ot denotes the output gate. The forget gate, memory cell, and hidden state\n",
      "are denoted by ft, st, and ht, respectively. The σ and tanh functions are deﬁned\n",
      "in 2 and 3 respectively.σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B.\n",
      "GAN Generative adversarial networks was introduced by Good- fellow et al. [4],\n",
      "where image patches are generated from random noise using two networks trained\n",
      "simultaneously. It has shown very good results when applied to various tasks\n",
      "particularly image synthesizing.\n",
      "Page number: 1\n",
      "\n",
      "\n",
      "Score: 0.5023\n",
      "Text:\n",
      "Tables V and VI show the average values of RMSRE and DPA with different (M, N).\n",
      "The numbers clearly indicate that LSTM has performed better than GAN model in\n",
      "terms of RMSRE metric. Although we have used speciﬁc loss function component for\n",
      "direction in GAN there has been no signiﬁcant improvement in DPA. There is\n",
      "strong possibility that the models have not achieved the optimum set of weights\n",
      "to solve the optimization prob- lem and produce good results. The best RMSRE has\n",
      "been observed for (M, N) = (700, 300), in LSTM model. B. 90 days Look-back We\n",
      "have also tested model LSTM on past 90 days data. For this purpose we have taken\n",
      "M ∈{400, 700, 1200} for N = 300 and M ∈{800, 1400, 2400} for N = 600. The result\n",
      "is shown in Tables VII and VIII. We can observe that the best RMSRE was again\n",
      "obtained for M = 1200 and N = 300. The model performance for DPA is very poor.\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.4935\n",
      "Text:\n",
      "The green bar shows the entire dataset, the blue bar shows the training set for\n",
      "one round experiment, and the yellow bar is the corresponding testing set. Image\n",
      "Source- [5] D. LSTM Model Architecture We have used a single layer of LSTM\n",
      "internal units followed by a Dropout layer. The activation function in LSTM is\n",
      "tanh and the loss function used is Mean Squared Error. The values for the hyper-\n",
      "parameters are mentioned in Table II. E. GAN Model Architecture As discussed\n",
      "earlier a GAN is a network of 2 models- generator G, and discriminator D. Here,\n",
      "we used the loss functions deﬁned and used by Zhou et al. [5].a) Generator: We\n",
      "have used a layer of LSTM followed by a dropout layer as our generator. To make\n",
      "the discrimina- tive model D as “confused” as possible, so that D does not\n",
      "distinguish between actual and closing price, the generative model G should have\n",
      "an adversarial loss which should be minimized such that D will not discriminate\n",
      "the prediction correctly. Consequently, we label actual closing price ¯ Y as\n",
      "class 1 and predicted closing price ˆ Y as class 0, and deﬁne the adversarial\n",
      "loss for G as 6 LG adv( ˆ Y ) = Lsce(D( ˆ Y ), 1) (6) where, Lsce is the sigmoid\n",
      "cross-entropy loss, deﬁned as 7 Lsce(A, B) = − X i Bi log(sigmoid(Ai)) + (1 −Bi)\n",
      "log(1 −sigmoid(Ai)) (7) 1217 Authorized licensed use limited to: Carleton\n",
      "University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.\n",
      "Page number: 2\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print(f\"Query: '{query}'\\n\")\n",
    "print(\"Results: \")\n",
    "# loop through the zipped together scores and indices from torch.topk\n",
    "for score, idx in zip(top_results_dot_product[0], top_results_dot_product[1]):\n",
    "    print(f\"Score: {score:.4f}\")\n",
    "    # print relevant sentence chunk (since the scores are in descending order, the most relevant chunk will be first)\n",
    "    print(\"Text:\")\n",
    "    print_wrapped(pages_and_chunks[idx][\"sentence_chunk\"])\n",
    "    # print the page number too so we can reference the textbook further and check the results\n",
    "    print(f\"Page number: {pages_and_chunks[idx]['page_number']}\")\n",
    "    print(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAjQAAAMsCAYAAABDeThBAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjkuMCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy80BEi2AAAACXBIWXMAAA9hAAAPYQGoP6dpAAEAAElEQVR4nOzdd3gU1f748feWZJNN7yG9UAIhlIAQkQ5SRVEQBBVR79Vrb9ivBRR7QUFFLIggCNhogvQiBAgdAqGk914228v8/uC382UJYEAUwz2v5/GRzM7OnCk785kzn3OOQpIkCUEQBEEQhBZMeaULIAiCIAiC8GeJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEb4Q6+++ioKheJKF+OqFxcXx+TJk690Mf4RNm/ejEKh4Icffrhsy1QoFHzzzTeXbXlXO/G7F1oaEdA0U2ZmJnfccQeRkZFoNBoiIiK44447OHr06JUu2mWVl5eHQqFg8+bNf2o5JSUlvPrqqxw4cOCylEv46xw9epRXX32VvLy8K12Ui/bNN99c9pvu5MmTUSgU+Pr6YjQam3x+8uRJFAoFCoWC995777Ku22nHjh28+uqr1NXVNWt+Z5md/2k0Gtq2bcvLL7+MyWT6S8r4T3Sx1x3n+eP8z8PDg7Zt2/Lwww9TXl7+1xZWuOxEQNMMP/30E6mpqWzYsIG7776bTz/9lHvvvZeNGzeSmprKsmXLrnQR/3FKSkqYOnWqCGhagKNHjzJ16tQWGdD8VdRqNQaDgRUrVjT57LvvvsPDw+MvXf+OHTuYOnVqswMaAI1Gw/z585k/fz4ffPABcXFxvPbaa9x7771/XUH/YS71ujNt2jTmz5/PrFmz6NWrF5999hnXXnstBoPhrymo8JdQX+kC/NNlZ2dz5513kpCQwNatWwkJCZE/e+yxx+jTpw933HEHhw4dIj4+/m8tm8FgQKvV/q3rFJrH4XBgsVj+8hvf1cZms+FwOK50MdBoNFx33XUsWrSIcePGuXy2cOFCRo4cyY8//niFSnduarWaO+64Q/77wQcfpFevXixatIgPPviAsLCwK1i6f7bhw4fTvXt3AP71r38RFBTEBx98wLJly5gwYcIVLp3QXKKG5g+8++67GAwG5syZ4xLMAAQHB/P555/T2NjIu+++K0+fPHkycXFxTZZ1vnfSCxYsoFu3bnh6ehIYGMhtt91GYWGhyzz9+/enY8eO7N27l759+6LVannhhRe46667CA4Oxmq1NlnukCFDaNeuHQAFBQVkZWVdyi44p3Xr1tG7d2/8/f3x9vamXbt2vPDCC8Dp/IdrrrkGgLvvvluuznXmLzi35dChQ/Tr1w+tVkvr1q3lfIktW7bQs2dPPD09adeuHevXr//D8jhzLpYsWcL06dOJiorCw8ODQYMGcerUqSbz79q1i2HDhuHn54dWq6Vfv35s377dZZ6LOY4KhYKHH36Y7777juTkZDQaDWvWrAHgvffeo1evXgQFBeHp6Um3bt0uOTfE+UrwvffeY86cOSQmJqLRaLjmmmvIyMhoMn9WVhZjx44lMDAQDw8PunfvzvLly+XPv/nmG2699VYABgwYIB+rzZs38+STTxIUFIQkSfL8jzzyCAqFgo8//lieVl5ejkKh4LPPPpOnVVRUcO+99xIWFoaHhwedO3dm3rx5592WGTNmyNtyvte4ZrOZG264AT8/P3bs2HFJ++9iTJw4kdWrV7vUkmRkZHDy5EkmTpx4zu/k5ORw6623EhgYiFarJS0tjVWrVjWZb+bMmSQnJ6PVagkICKB79+4sXLgQOH1+Pf300wDEx8fLx+Ria9AUCgW9e/dGkiRycnJcPlu9ejV9+vTBy8sLHx8fRo4cSWZmZrOW+0fXq4cffhhvb+9z1m5MmDCB8PBw7HY7AMuWLWPkyJFERESg0WhITEzktddekz93cl4zjh49yoABA9BqtURGRvLOO+/I8/zRdediDBw4EIDc3Fyg+b9ho9HIo48+SnBwMD4+Ptx4440UFxejUCh49dVXXeYtLi7mnnvuISwsDI1GQ3JyMl9//XWTZV7ua/fVTAQ0f2DFihXExcXRp0+fc37et29f4uLizlk13RzTp09n0qRJtGnThg8++IDHH3+cDRs20Ldv3ybVzdXV1QwfPpwuXbowY8YMBgwYwJ133kl1dTW//faby7xlZWVs3LhRfmKbNGkS7du3v6Qyni0zM5MbbrgBs9nMtGnTeP/997nxxhvlgKB9+/ZMmzYNgPvuu0+uBu/bt6+8jNraWm644QZ69uzJO++8g0aj4bbbbmPx4sXcdtttjBgxgrfeegu9Xs/YsWPR6XTNKttbb73Fzz//zJQpU3j++efZuXMnt99+u8s8GzdupG/fvjQ0NPDKK6/wxhtvUFdXx8CBA9m9e/cl75eNGzfyxBNPMH78eD766CM5GProo4/o2rUr06ZN44033kCtVnPrrbee80bXXAsXLuTdd9/l/vvv5/XXXycvL49bbrnFJbDNzMwkLS2NY8eO8dxzz/H+++/j5eXF6NGj+fnnn4HT5++jjz4KwAsvvCAfq/bt29OnTx9qampcbnTbtm1DqVSybds2l2nOZcHpi3r//v2ZP38+t99+O++++y5+fn5MnjyZjz76qMm2zJ07l5kzZ3Lffffx/vvvExgY2GQeo9HIqFGj2LFjB+vXr6dXr16XvO+a65ZbbkGhUPDTTz/J0xYuXEhSUhKpqalN5i8vL6dXr1789ttvPPjgg0yfPh2TycSNN94o72+AL774gkcffZQOHTowY8YMpk6dSpcuXdi1a5e8XmetwIcffigfk7MfqJrDGQQFBATI0+bPn8/IkSPx9vbm7bff5qWXXuLo0aP07t37D4Om5lyvxo8fj16vb3J+O1/hjR07FpVKBZwOqL29vXnyySf56KOP6NatGy+//DLPPfdck3XX1tYybNgwOnfuzPvvv09SUhLPPvssq1evBpp33Wmu7OxsAIKCgoDm/4YnT57MzJkzGTFiBG+//Taenp6MHDmyyfLLy8tJS0tj/fr1PPzww3z00Ue0bt2ae++9lxkzZrjMezmv3Vc9STivuro6CZBuuummC8534403SoDU0NAgSZIk3XXXXVJsbGyT+V555RXpzF2el5cnqVQqafr06S7zHT58WFKr1S7T+/XrJwHS7NmzXea12+1SVFSUNH78eJfpH3zwgaRQKKScnByX71+Ks8v94YcfSoBUWVl53u9kZGRIgDR37twmnznLsnDhQnlaVlaWBEhKpVLauXOnPP23334773LOtGnTJgmQ2rdvL5nNZnn6Rx99JAHS4cOHJUmSJIfDIbVp00YaOnSo5HA45PkMBoMUHx8vXX/99fK05h5HSZLksmdmZjaZ32AwuPxtsVikjh07SgMHDnSZHhsbK911110X3M7c3FwJkIKCgqSamhp5+rJlyyRAWrFihTxt0KBBUkpKimQymeRpDodD6tWrl9SmTRt52tKlSyVA2rRpk8u6KioqJED69NNPJUk6/XtQKpXSrbfeKoWFhcnzPfroo1JgYKC8P2fMmCEB0oIFC1y2+dprr5W8vb3l34lzW3x9faWKigqXdTuP59KlSyWdTif169dPCg4Olvbv33/B/XMhzTmPJOn0cffy8pIkSZLGjh0rDRo0SJKk07+18PBwaerUqXLZ3333Xfl7jz/+uARI27Ztk6fpdDopPj5eiouLk+x2uyRJknTTTTdJycnJFyzDu+++KwFSbm5us7bNWebKykqpsrJSOnXqlPTee+9JCoVC6tixo3xsdDqd5O/vL/373/92+X5ZWZnk5+fnMv1Sr1cOh0OKjIyUxowZ4zLfkiVLJEDaunWrPO3s34YkSdL9998vabVal/PWec349ttv5Wlms1kKDw93Wc+FrjvnMnfuXAmQ1q9fL1VWVkqFhYXS999/LwUFBUmenp5SUVHROct5rt/w3r17JUB6/PHHXeadPHmyBEivvPKKPO3ee++VWrVqJVVVVbnMe9ttt0l+fn4u6/sz1+7/NaKG5gKctQI+Pj4XnM/5eXNrEZx++uknHA4H48aNo6qqSv4vPDycNm3asGnTJpf5NRoNd999t8s0pVLJ7bffzvLly13W/91339GrVy85r2fz5s0urw7+DH9/f+B0dfGl5jt4e3tz2223yX+3a9cOf39/2rdvT8+ePeXpzn+fXWV+PnfffTfu7u7y386aNef3Dxw4IL8yqK6ulve5Xq9n0KBBbN269ZK3qV+/fnTo0KHJdE9PT/nftbW11NfX06dPH/bt23dJ64HTT8FnPnWfvZ01NTVs3LiRcePGodPp5O2srq5m6NChnDx5kuLi4guuIyQkhKSkJLZu3QrA9u3bUalUPP3005SXl3Py5EngdA1N79695ddwv/76K+Hh4S65B25ubjz66KM0NjayZcsWl/WMGTPmvLUP9fX1DBkyhKysLDZv3kyXLl0uYi/9eRMnTmTz5s1yjWdZWdl5Xzf9+uuv9OjRg969e8vTvL29ue+++8jLy5Nfpfn7+1NUVHTOV4R/hl6vJyQkhJCQEFq3bs2UKVO47rrrWLZsmXxs1q1bR11dHRMmTHC55qhUKnr27NnkmnOm5l6vFAoFt956K7/++iuNjY3y9xcvXkxkZKTL/jnzt+E8T/v06YPBYGjymsXb29slR8jd3Z0ePXo0+9pwIYMHDyYkJITo6Ghuu+02vL29+fnnn4mMjGxSzvP9hp2vmB988EGXZT/yyCMuf0uSxI8//sioUaOQJMllXw4dOpT6+nqX5V7Oa/fVTiQFX0BzAxWdTodCoSA4OPiiln/y5EkkSaJNmzbn/NzNzc3l78jISJebtdOkSZN4++23+fnnn5k0aRLHjx9n7969zJ49+6LK01zjx4/nyy+/5F//+hfPPfccgwYN4pZbbmHs2LEolc2LkaOioprkofj5+REdHd1kGpy+iDRHTEyMy9/Om77z+86b8F133XXeZdTX17sEC811vqTwlStX8vrrr3PgwAHMZrM8/c80N/6j7Tx16hSSJPHSSy/x0ksvnXMZFRUV8gX7fPr06cOvv/4KnA5cunfvTvfu3QkMDGTbtm2EhYVx8OBBl5t8fn4+bdq0aXIuOKvN8/PzXaZfKJn+8ccfx2QysX//fpKTky9Y1r/CiBEj8PHxYfHixRw4cIBrrrmG1q1bn/PVTH5+vksw7nTmdnfs2JFnn32W9evX06NHD1q3bs2QIUOYOHEi11133Z8qq4eHh/zqu6ioiHfeeYeKigqXm7Hz/HfmiJzN19f3vMu/mOvV+PHjmTFjBsuXL2fixIk0Njby66+/cv/997uc95mZmfz3v/9l48aNNDQ0uCyvvr7e5e9zXTMCAgI4dOjQecvcXJ988glt27ZFrVYTFhZGu3btXM7f5vyG8/PzUSqVTc7n1q1bu/xdWVlJXV0dc+bMYc6cOecsT0VFxZ/epv9FIqC5AD8/PyIiIv7wB3Po0CGioqLkYON8N6qzE90cDgcKhYLVq1fL75TP5O3t7fL3mRemM3Xo0IFu3bqxYMECJk2axIIFC3B3d2/SOuNy8fT0ZOvWrWzatIlVq1axZs0aFi9ezMCBA1m7du05t+Vs55vnfNOb+4TyR9931r68++67533ad+735h5Hp3Mdn23btnHjjTfSt29fPv30U1q1aoWbmxtz586Vk0AvRXO3c8qUKQwdOvSc8559oT2X3r1788UXX5CTk8O2bdvo06ePnGy6bds2IiIicDgc580xa47zndcAN910E99//z1vvfUW3377bbMD5stFo9Fwyy23MG/ePHJycpokdl6K9u3bc/z4cVauXMmaNWv48ccf+fTTT3n55ZeZOnXqJS9XpVIxePBg+e+hQ4eSlJTE/fffLyeCO8+L+fPnEx4e3mQZavX5bwkXc71KS0sjLi6OJUuWMHHiRFasWIHRaGT8+PHyPHV1dfTr1w9fX1+mTZtGYmIiHh4e7Nu3j2effbZJTemfvTZcSI8ePeRWTme73L9h53bdcccd532w6tSp00UvVxABzR8aNWoUn3/+Ob///rtLVanTtm3byMvL48knn5SnBQQEnLP/iLOfTBMTE5Ekifj4eNq2bfunyjlp0iSefPJJSktL5Wall1LL0FxKpZJBgwYxaNAgPvjgA9544w1efPFFNm3axODBg/+xPYwmJiYCp59Ez7z4n0tzj+OF/Pjjj3h4ePDbb7+h0Wjk6XPnzm32Mi5FQkICcPqp+Y+280LHyhmorFu3joyMDDlZs2/fvnz22WdERETg5eVFt27d5O/ExsZy6NAhHA6HSwDifIUQGxvb7O0YPXo0Q4YMYfLkyfj4+Li0pPq7TJw4ka+//hqlUunymvRssbGxHD9+vMn0c223l5cX48ePZ/z48VgsFm655RamT5/O888/j4eHx2X5/bRq1YonnniCqVOnsnPnTtLS0uTzPzQ09A/Pi7Nd7PVq3LhxfPTRRzQ0NLB48WLi4uJIS0uTP9+8eTPV1dX89NNPLom7zpZFl+KvuO409zccGxuLw+EgNzfXpRbr7FaWISEh+Pj4YLfbL/oYCBcmcmj+wJQpU9Bqtdx///1UV1e7fFZTU8N//vMffH19efjhh+XpiYmJ1NfXu9TslJaWurR0gNOtGVQqFVOnTm3ylCFJUpP1XciECRNQKBQ89thj5OTkuLxrhsvb9K+mpqbJNGdth7M61svLC+CiOgb7O3Tr1o3ExETee+89l/f7TpWVlfK/m3scL0SlUqFQKFxqdfLy8vjll18ubQOaKTQ0lP79+/P5559TWlra5PMzt/NCxyo+Pp7IyEg+/PBDrFar/FqkT58+ZGdn88MPP5CWlubyZD9ixAjKyspYvHixPM1mszFz5ky8vb3p16/fRW3LpEmT+Pjjj5k9ezbPPvvsRX33chgwYACvvfYas2bNOmethtOIESPYvXs36enp8jS9Xs+cOXOIi4uT86vO/l27u7vToUMHJEmSW6ldrt/PI488glar5a233gJO19r4+vryxhtvnLOrhzPPi7Nd7PVq/PjxmM1m5s2bx5o1a5rUGDtrXM5clsVi4dNPP724jTzDX3Hdae5v2FkTenb5Z86c2WR5Y8aM4ccff+TIkSNN1nf2MRDNtptP1ND8gdatW/Ptt98yYcIEUlJSuPfee4mPjycvL4+vvvqK2tpavv/+e5f3prfddhvPPvssN998M48++igGg4HPPvuMtm3buiR7JSYm8vrrr/P888+Tl5fH6NGj8fHxITc3l59//pn77ruPKVOmNKucISEhDBs2jKVLl+Lv79+kqeCkSZPYsmXLZamenTZtGlu3bmXkyJHExsZSUVHBp59+SlRUlFyLlZiYiL+/P7Nnz8bHxwcvLy969uz5t3c+eDalUsmXX37J8OHDSU5O5u677yYyMpLi4mI2bdqEr6+vnIfQ3ON4ISNHjuSDDz5g2LBhTJw4kYqKCj755BNat259Wd79X8gnn3xC7969SUlJ4d///jcJCQmUl5eTnp5OUVERBw8eBE4HoyqVirfffpv6+no0Gg0DBw4kNDQUOB28fP/996SkpMi1fqmpqXh5eXHixIkmSbL33Xcfn3/+OZMnT2bv3r3ExcXxww8/sH37dmbMmPGHSfbn8vDDD9PQ0MCLL76In5+f3OfR30GpVPLf//73D+d77rnnWLRoEcOHD+fRRx8lMDCQefPmkZuby48//ijXVg0ZMoTw8HCuu+46wsLCOHbsGLNmzWLkyJHyvnHWeL344ovcdtttuLm5MWrUKPmG3VxBQUFy7+bHjh2jffv2fPbZZ9x5552kpqZy2223ERISQkFBAatWreK6665j1qxZ51zWxV6vUlNTad26NS+++CJms9nldRNAr169CAgI4K677uLRRx9FoVAwf/78P3WN+iuuO839DXfr1o0xY8YwY8YMqqurSUtLY8uWLZw4cQJwrT1666232LRpEz179uTf//43HTp0oKamhn379rF+/XqXh8bLee2+6v2dTapassOHD0sTJ06UwsPDJaVSKQGSh4fHOZvpSpIkrV27VurYsaPk7u4utWvXTlqwYME5m/tKkiT9+OOPUu/evSUvLy/Jy8tLSkpKkh566CHp+PHj8jz9+vX7w6aezmaR9913X5PPLmez7Q0bNkg33XSTFBERIbm7u0sRERHShAkTpBMnTrh8b9myZVKHDh0ktVrt0pTyfNsSGxsrjRw5ssl0QHrooYcuWMYzm/meydm89uxmnPv375duueUWKSgoSNJoNFJsbKw0btw4acOGDS7zNfc4XqiMX331ldSmTRtJo9FISUlJ0ty5c8+5jItptn1mc+Ezy3Bm01BJkqTs7Gxp0qRJUnh4uOTm5iZFRkZKN9xwg/TDDz+4zPfFF19ICQkJkkqlatKE+5NPPpEA6YEHHnD5zuDBgyWgyT6TJEkqLy+X7r77bik4OFhyd3eXUlJSmhyDC23L+Y7nM888IwHSrFmzzrV7Luhc58G5nNls+3zOV/bs7Gxp7Nixkr+/v+Th4SH16NFDWrlypcs8n3/+udS3b1/53EtMTJSefvppqb6+3mW+1157TYqMjJSvNxdqwn2hMmdnZ0sqlcrl3Nq0aZM0dOhQyc/PT/Lw8JASExOlyZMnS3v27JHn+TPXK6cXX3xRAqTWrVufs2zbt2+X0tLSJE9PTykiIkJ65pln5K4azjwHz3fNOFfXCue77pyLs9l2RkbGeeeRpOb/hvV6vfTQQw9JgYGBkre3tzR69Gjp+PHjEiC99dZbLvOWl5dLDz30kBQdHS25ublJ4eHh0qBBg6Q5c+a4zCeabTefQpJE2Hcpvv32WyZPnswdd9zBt99+e6WLA5xuRj169Gi2bt36p5I0BeFqpFAomDt3rhjRXPhbHThwgK5du7JgwYImnXwKl5d45XSJJk2aRGlpKc899xxRUVG88cYbV7pIfPHFFyQkJJwzeVkQBEH4axmNxiat9mbMmIFSqbykHouFiyMCmj/h2WefvSJJimf7/vvvOXToEKtWreKjjz76x7YwEgRBuJq988477N27lwEDBqBWq1m9ejWrV6/mvvvua9LHlnD5iYDmKjBhwgS8vb259957m/RSKQiCIPw9evXqxbp163jttddobGwkJiaGV199lRdffPFKF+1/gsihEQRBEAShxRP90AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQhD9FkiQkSbrSxRAE4X+c+koXQBCEliE9PZ2KigrKysoIDw/H29ubgQMHsmfPHlJSUvDw8KCoqIhVq1YRFxcHQJ8+fTh+/DgHDhzglltuwc/Pj8bGRn766ScCAwPx8PBAqVQyYMAAFAoFJpOJlStX4uvrS0hICHv37mX06NEEBwdf2Y0XBOEfT9TQCILQLAqFgmuvvZYNGzaQmpqKp6cnOp2OmTNnsmvXLiRJIiAggPXr19OpUyeOHDnCr7/+ipubG+np6Xh7ewOg1WrZvXs3oaGhXHvttQQGBsrr0Gg05OTkoNfr8ff3Z+vWrfj7+1+hLRYEoSURAY0gCM3Ss2dP/Pz8UKlUeHt7c+2115KVlcXEiRP5+eefsdvtKJVKrFYr+/fvJy8vj4SEBNRqNWq1GoVCAZwOjJRKJVlZWaxevbrJepzzq1Qql+8JgiBciAhoBEFolrMDC5vNxq5du7BYLOTl5XHixAkA3NzcSExMpH379vz+++8u35EkCbvdjkKhIDY2li5duuDn53fedYrcHEEQmksENIIgNIskSVitVkwmExaLhcOHD9OpUyeGDx/OpEmTWLBgAWazGZvNhru7O1qtFp1Oh8lkwmg0Ultby7Fjx+QgSKFQ4O/vz+HDh9Hr9fI6TCYTJpMJs9mM2WzGbrdf4S0XBKElUEjiEUgQhGaQJInKykp27NhB165dqaqqIiAggPj4eHJzc8nMzCQ5OZmDBw8SEhICQOfOnSkuLiYrK4vQ0FB0Oh1t2rQhKysLT09P3NzcsNvt9OnTB6VSidlsZvv27Xh6euLv78/Jkyfp06cPAQEBV3jrBUH4pxMBjSAIgiAILZ545SQIgiAIQosn+qERhEtUXl7Ozp076dGjB5IkkZGRwbXXXktoaOiVLprw/1VVVbF9+3a6du1KdHQ0CoUCu93Onj17MBgM+Pv7065dOzl52dlSy/nqrEePHoSEhOBwONi6dSsBAQGkpKSgVIpnQUH4pxEBjSBcIk9PT+bOnUvv3r2RJIm5c+cyYMAA9u7di8ViITU1lYyMDNq0aUNGRgZBQUHExsZSW1tLaWkpvr6+cl5JeXk5fn5+dOzYEU9PT+B0zkpmZialpaX4+fnRqVMnVCoVubm5tGnTBrvdzpYtW/D09MTHx4fk5GSUSiUNDQ3U19cTHR2NwWBgy5Yt9O3bV+7/xdfXF7PZTE1NDV5eXrRq1YqioiIaGhpwd3cnKChITsx1d3eXc2SsVive3t6kpKTg5uaGw+Fg9+7d+Pj40LZtW/bs2UNqaiqVlZX4+Pjg5+dHcXExmZmZ9O/fHzc3N4qKisjKyqJv374cOXKEsLAwjh07hq+vLzabjS5dulBUVERlZSUhISG0a9cOSZLIycnh5MmTBAcH4+PjQ3BwMCdOnEChUJCcnIzBYCAjI4PQ0FDi4uIICQlBoVDg6enJvHnz6NGjh3zccnJyWLNmDffffz/r16+nVatWuLu7c+rUKWpra7nuuuuwWCz8+OOPHD16lClTplBWVsarr77KK6+8gkKhQJIkTpw4QX19Pddccw0Gg4HNmzfTv39/jEYjJ0+ebFK2sLAwQkJCiIqKYtu2bXh4eOBwOGjfvj379u2jTZs2xMTEiGbqgnCJxGOGIFwiNzc3FAoFBQUFFBQUoFQqcXNzIygoiCVLlpCXlweAl5cXGzduJCAggFmzZlFRUUFeXh7x8fEolUpKS0spKSnh8OHDTZo5V1RUUFRUxJEjR9i2bRsFBQXMnDkTvV6PSqUiPT0dHx8fvv32WyorK5EkiX379vH1119jt9txd3fnt99+Y/369ej1er788kvq6uooKSmhrKyM2NhYubVRTk4OKpUKDw8PTp48icFgIDo6GqVSydGjR7FaraxZs0Zunu288a5btw6r1Up9fT1qtZrly5fz22+/yU20t2/fjlp9+tnJbreTnp6Ou7s7JpMJT09Ptm/fTkBAAKGhodjtdhYtWkR4eDg6nU7eD0ajkUOHDhEfH49CoeCTTz4hNjYWPz8/5syZgyRJbNu2DTc3N7766iv5e2q1Gg8PD9zd3eXyOltW/fbbb4wcOZLQ0FD69u2Lr68v3t7epKWlERMTQ9++fdmzZw+VlZUcPnyYpKQkNBqNvByLxcKHH35IQ0MD+/fvZ+nSpXLZ4uLi8PX15YsvvkCSJLZv305CQgJLlizh6NGj7NmzB09PT1q1aiUfR5vN9tecqILwP0IENILwJ5zZzNiZXx8bG0v//v35/vvv6dGjB0qlEoPBQFZWFn369EGtVlNQUMCGDRtQq9UolUoOHjxIXl4enTt3Jjc3l6VLl5KdnY1SqaSwsBCApKQkTpw40aR/l+XLlxMXF0dAQABmsxmdTofNZiM3NxeFQkH//v3Ztm0bhw8fJjExEaVSKdf0bNy4EQ8PD9q3b49Go0Gr1dK2bVuUSiUnT55k8+bN8vAEW7ZskfuPgdMBTWpqKmVlZezatYu2bdtSXl5OXFwcO3fuxGg0olQqXTrHU6lUqFQq4HSvwGq1GkmSOHToEHv27EGr1TJgwAA+/PBDVCoVkiTJHfGVlpayYcMG9Ho9lZWVhIaGEhMTw8mTJ7HZbFRVVfHTTz8xZMiQCx4vf39/Zs6cSWZmJu+99x6SJLm8QlIqlSiVSgICAujTpw8LFixAkiSXHo2dtT/dunVjzZo1NDQ0EBwcTGNjIxUVFYSFhREbG8vx48ex2+2o1WoCAgJITU1l//79AGRmZpKeno6npycqlUoug2inIQiXRgQ0gnCJHA4HKpWK9u3b0759e1QqFQ6HA4D27dvj4eGBm5sbkiSh1WoZMmQIw4YNQ5Ik4uPjueGGG9DpdPJrB4Dq6mqCg4Pp1q0boaGhSJJETEwMd911l/yap1u3bqxfvx6r1YpCoaBHjx6UlJTIrzr8/Pzo0aMHy5cvx+FwyEHK4cOHCQwMRJIkHA4HrVu3ZtSoUdTV1QGna0+c5ZckiQ4dOjBy5EiqqqqQJInU1FTq6+tpaGiQb7pubm7079+ftWvXEhUVRWZmJhEREURFRZGeni7X0litVnJzc3F3d8dqtWK1WrFYLLi5uaFUKunWrRv9+vWjuLiYgIAAHnnkEZYtWyYPfClJElFRUdx88820bdsWX19f6uvrqaioICoqCrVaTXBwMPHx8RQUFMjlc/adYzabaWhoYNu2bezdu5fa2lqmTJlCcXExVqsVSZKw2WzYbDZ5/+j1ekaPHs2mTZvo0KEDNpsNu90ul8dutzNgwADWrl1LTEwMkiTh5eWFn58fdXV1lJWVER0djVqtxuFwYDKZOHXqFB07dkShUNClSxcGDx5MRUUFDocDu91OXV0d5eXlf/OZLAhXB5FDIwiXqK6ujuuvv56GhgYABg8eTH19PVqtFkmSSEpKkm/c11xzjfxKwdvbG39/f/bt24eXlxcBAQG4u7szfPhwDh8+TJs2bUhISECSJHx9feW8isrKSjp06EBERAR5eXlUVVWRnJxMQkICAQEBZGdnYzAY6NSpE5Ik0djYSF1dHQqFghtvvBG73U5mZiYeHh4EBARgsVjYu3cvAQEBOBwO4uLi0Gg0SJJEREQERqORffv2ERkZSWxsLIGBgbRv357s7GzCw8NRKBQoFAr69u1LXFwcDocDtVpNcnIyHh4e5ObmYrPZ5Bwii8VC3759GTRoEFu3biUmJga1Wk1KSgr5+fmYTCY6depEZWUl7u7uTJgwwaVmp127digUCrRaLQ8++CBHjhxBkiQeeOABzGYz3bp1Y+jQoWzevJnGxkY56Bk0aBCHDh3CarXi6+tLREQEBw4cIC8vjyeeeAKNRoPNZiMgIABPT08sFguNjY2o1Wq0Wi1vv/02gYGBcg6Ts9bIYrGg1Wp55plnCAgIoHv37kiSxIMPPsjhw4flf5tMJjp06MD+/fvp06cPbdu2lV/5FRUV0a5dO5KTkykoKODkyZN07dr1ip3TgtCSiX5oBEH4U5yXkOYms555yTn7O2dfjs63zAst42LWfynfv5jliwRfQfj7iIBGEARBEIQWT+TQCIIgCILQ4omARhAEQRCEFk8ENIIgCIIgtHgioBEEQRAEocUTAY0gCIIgCC2eCGgEQRAEQWjxRMd6whXh7GlVjF8jCC2fcxwz0e+OcCWJgEa4YpYtW4bJZDpn52rO3lj/qk7Pzlzumetz/n2udTscDpcxfy7EZDJRVlZGXFxck8+cwwtcju27mDJdLjqdjvr6eqKiolymO/cjXJ5ta47Lvc6TJ0+SkJAgjzd19noudh1/1OngmeeC08Usv6GhgcbGRiIiIs5Z3jOXeTnLfTa9Xs8dd9whjxQvCFeCCGiEK0KSJCwWC7fddpvLzcNut7Nq1SpWrVrFf/7zH7p06eJyUTWbzdTV1REWFnZJ69yzZw/Lli3jpZdekrv537FjB9988w2TJk2iQ4cOLF68mFtvvZWQkBD5uw0NDbz22ms8++yzwOkRtC908a6urmbnzp2MHDnSZbrJZGLOnDkUFRXx2GOPERkZedHb4WQ2m5k2bRr/+te/iI+Pv+TlONlsNioqKmjVqtUFb2SFhYVkZ2fTv39/l+mVlZW89957hIaGEhkZSXBwMP3795dH2v6zJEmivLycwMBA3N3dAaivr2fGjBnyoJnu7u6MHTsWNzc3l+9VVlbi6+uLh4fHBZe/dOlSbrrpJjQajTzd4XCwZs0ali9fzvXXX09paSmTJk3C19fX5bt1dXWo1Wp8fHzkaRkZGaxcuZL//ve/cpmdnx07dowTJ07IY0QFBASQl5fHv//972YHEnl5eRQUFNC3b1+X6WceC+egpg8++KDLdl3I3r17+eWXX3jxxRcvuM+cli9fjtlsFgGNcEWJHBrhH8U5yvG6desICwujoKCA0tJSjh8/jslkYvbs2XzzzTeYzWaKi4spLS3FbDaTm5tLaWkpubm58r9PnTqF3W6ntraWU6dO0dDQgFqtZu3atS6vuoKDg1m7di3h4eEEBgYSFhaGu7s7ubm5FBYWUlhYiN1up3379uh0Oh599FGOHj1KbW0tubm51NfXN3v7NBoNer2eo0ePykGZJEnU1NSQl5eH1Wqlvr6e7OxsamtrMZlM5ObmUlJSQn5+Pjk5OZSUlJCTkyOPk6TRaMjLy6O0tFS+QdbV1XHq1ClKSkqwWCzyeqqrqyksLCQnJ4fGxkaKi4vJycnBYrGwePFiPvroI0wmE6WlpRQVFck1CM0RGBjI4cOHsVqt9OvXj2effZa1a9dSWFhIWVmZvJ78/HzKysrk41ZYWEhubq48IGR2djYNDQ3U1tZSWFhIVVUVBQUFFBQU8Mgjj5Cfny/XIPj6+lJcXExZWRmdO3fm6aefJicnh4aGBrKzs6murqa8vJxHHnmEU6dOYTAY5G1vbifpCoUCf39/1q1bR3JyMp988gkZGRkYDAZyc3MpKyujsbGRKVOmkJGRgdVqlc9bhULBb7/9Rk5ODhUVFS7r/PLLL6mrq2PQoEH4+fnh4+NDfHw8tbW15Ofnk52djdFopKSkhOLi4os6FkFBQWRmZmI2m+nUqROdO3emsbGR3Nxc8vLyKCkpkYMw53ocDgfFxcUUFBSgVCr57bffyM7OpqKiApvNRn5+PkVFRdjt9maXQxD+TqKGRvhHcVaNO59QX3rpJVq1asXhw4e5//772bVrFzExMaSnp7Ny5Urc3d256aabeP7550lJSUGlUrF//36GDRvGypUree+995g3bx79+/dn+fLlPPLII+ddn3OdarWaiooKxo8fz913383SpUuZOXMmc+bMISEhgT179lBbW8vLL7/MsGHD8Pb2pl+/fpe8jc6Rn3v27IlSqeTQoUP069ePpUuX8v777zNx4kSuvfZa3N3d2bJlC+PGjeOHH35g1qxZzJ07l3bt2vHmm2/SvXt3Nm/ezLPPPsvcuXO56aab+PLLL1m4cKEcPGVkZPDKK6+QmppKhw4dqKiowGg00rlzZ/bt20dDQwOZmZl89dVXREZGMmDAAK677rqL3jZnLc2aNWv47bffqK6uxtPTk+TkZNzc3MjKymLcuHG8/PLLjB49mnXr1vHUU0/x008/cf3117NixQpuuukmPvroI55++mneeustXn/9dQ4dOoTBYHBZp1KppLKykpUrVzJo0CBCQkJ49dVX6d27N/Pnz+eFF15g//79NDQ08Prrr9OuXTsyMzOZPn26S03OH22TxWLh119/pW3btnTs2JGPPvqI1q1bM2/ePF577TV27drFjTfeyHfffUdjYyOHDh1izJgx6HQ6li1bxq5du/j+++/lmppu3brxzDPP8Ntvv/HKK6+wZs0acnJysFqtFBcX89VXXzFt2jRWrFhBUFAQI0eO5JprrrmoY3Hq1ClmzJjB1KlTqaqqYsyYMTz88MN8//33vPXWW3z66aeMGDGCzz//nP79+7Nv3z7Ky8sZMGAAjY2NrFq1iu3bt/Of//yHjIwM2rVrxw033ICXl1ezyyEIfxdRQyP8Y6lUKhQKBW3btqVVq1aUlZXh7+9PZGQku3btoqCggLZt2+Lm5obD4WD06NG89tprAKSmpuLp6UltbS1du3alqKiIY8eOYbVam6znzNwZOP2KwdPTE7vdTp8+fdDpdNjtdsxmM97e3mi1WuLi4vD09GT69OkoFIpmP+2fzWKxkJ6eTkZGBv7+/gQHB9O1a1eKi4vJysqSk6fvvPNOnnvuOex2O2lpaUiShF6vx2azyetPSUnB39+f0tJSLBYLkiQRHByMn5+fvJ1arRaNRsP777/PTTfdhJ+fH+Xl5Zw6dYqgoCDCwsLIzMzk1KlTxMXFNcklaS7nK8WwsDB5BO6PPvqI1atXExMTgyRJbNmyBYfDQdeuXQkJCeGXX35hy5YtdOjQgX379lFSUoLBYMDLywuDwYC/vz9arZaoqKhz5l0tX76cpKQkfH196datG8XFxZw8eRI3Nzc8PT0JCgrit99+w2w206ZNm4uuaXBzc+OGG26gvr6en376ia5du1JRUUFeXh4mkwkvLy+ioqJYvHgxMTExvPHGGwQEBODr60vfvn0pLi6Wzz9JkujTpw9LlixBp9Px+uuv4+bmhtFopHfv3qxYsYKJEydy6tQpcnJySExMvKRcqaioKHr16iWv126307dvX4xGI0uXLuXkyZMkJSWxevVqvvjiC/z8/Jg6dSoxMTH4+PjQv39/ioqK0Gq1rFq1ip07d17yOSEIfzUR0Aj/KJIkodPpsNlsNDY2YrVaMZlMWCwWLBYLHh4e5ObmEhYWRkVFBa1bt8bPzw+bzYbZbEahUGCz2TAajVitVkpKSvjkk09Qq9Xyqxir1eryyikoKIjIyEj2799Pfn4+arUah8OBzWaTgwaDwYDNZsPhcKBQKNi/fz89evTgzjvvZM+ePc3ePudrFbPZTGVlJfPnz6e+vh6NRoO7uzshISHMnDkTpVKJw+GgsbFR3jY4fUNylsVoNGKz2TCZTC7b7HA4CA0Npbq6mkcffVSuhZAkCbPZjNVqRZIktm7dyvr16/Hy8qKxsRF3d3eKi4vx9fWloaGByMhIWrVq1exts1gsmM1mdDode/bswdvbm9tvv12ertFo6NmzJ1VVVVgsFrp06YLD4SA3N5eqqir69etHQkIC5eXltGrVisTERKxWK9nZ2VitVux2O3a7nWPHjsmvX5zHyNvbm1deeYU5c+awdu1aPv74YxwOBw6HQ37llpOTQ0JCApWVlXTp0qXZN2bnOWm322lsbMTDwwNvb29mz55NbW0tSqUSo9Eo1zy1b9+eZcuWUVJSIp/LzmPmDKIcDgfffPMNMTExPPXUU0RGRmKxWLBaraxevZry8nISEhJQKpXU1tYSGxtLaGhoswNn5z632Wx069aNr7/+msrKSmw2G8ePHwdg4MCB+Pv7U15eTnJyMn369GHt2rXk5eVRV1cnl9tut9PQ0MD7779PcXExOp2u2eeEIPydxGjbwhXhcDhYvHgx48aNc7mxOBwOjhw5QklJCYmJiZSXl+Pt7Y1er8fX1xcvLy+ys7Pp1asX6enpuLm50aVLF/bs2UNkZCQxMTHs3LmT8PBwysvLiYiIIDc3l8DAQCorK4mPj6ekpISePXu6JG8WFBSwf/9+goOD6dKlCyaTif379xMbG0t+fj4xMTEUFBSQkpLC8ePH8ff3x2KxYLPZaNeuHQEBAS61BudLCjabzezZsweTyURQUBA1NTWkpqZSUFBAbm4uvXv3Zu/evfj5+VFVVUWnTp04fvw48fHxhIaGsmvXLqKjoykqKiIiIoKSkhJiY2MpKioiODiY6upqPDw8ePfdd/Hw8KCiooIPP/yQ5ORkJEkiJyeH3NxcrrnmGkwmEzt27CAyMhKr1UpcXByHDh2id+/eHDhwAKPRSN++fZskhZ4vKbi+vp49e/bg5uaGr68vcXFx+Pj4kJGRgUKhoFu3bhiNRg4dOoSHhwdJSUkMHz6cO++8k9atW9O7d29KS0vJz88nIiKC6OhoNm7ciI+PD5Ik0alTJ3bt2kViYiLx8fEoFAr0ej0ZGRnA6Vq5Q4cO4e3tLb/iqqur45prruHw4cNERkbi6+vL9u3b6datG3FxcU1au50vKfjo0aMUFRUREhKCt7c3iYmJ7Nu3j8bGRux2O+3ataO0tBS1Wk3r1q3ZvHkzsbGx+Pn5kZOTQ1RUFIWFhaSlpeHt7Y3D4WDPnj1YrVaUSiXt27cnPz8fvV6PVqulvr4eNzc3kpKSOHToEHa7nd69ezdJ6j1fUnBdXR179+7Fzc0NLy8v9Ho98fHxjBgxgqlTpxIdHU1qaionTpygqqqK1q1b4+/vz+bNm/Hz86NVq1ZkZ2fL5314eDgGgwEPDw86duzYpLZo+fLl9O3bF39//wv97AXhLyUCGuGKOF9Ac7U4X0Dzd6ivr2f+/PmkpqZSUVFB//79L+uN5nwBzcWqra3lscceY+zYsYwaNeqK92FyvoDmn+x8Ac255OTk8Mwzz/DCCy/QtWvXy7q/RUAj/BOIgEa4IiRJYt68ecTHx//t/aj8HXQ6HUePHqVnz55XZP0OhwO73Y5Kpbrs+7eyspKSkhI6d+78p5YjSZKcA3S5mnb/WVu2bKFXr17NShb+JygvL6eiooKUlJQ/nNd5TiiVysv+EHH06FEmTpwo13oKwpXwz7iKCP+TVCoVYWFhf1kNjcPhQKfT4efnh9FoRKfTERgYiFqtprGxUU44/StabHh4eODv7094ePglL8PZ5Fyr1aLVauXpjY2NmM1mAgICgNM1HRqNBm9vb4xGIzU1NfJ+bWhowG63y0/OtbW1OBwOAgMD/1TCr9Fo/FPbZjKZ0Ol0hISEyMGMw+GgoqICT09POS+qrq4OhUJBQEAA9fX1WK1WQkJCkCSJ2tpaPDw85Byg+vp6wsPD/9R2+fn5yc32Lwdn02hnYvK5juefYbfbsVqtF3UsnAnlVqsVPz8/6uvrUSqV+Pr6yvvc19f3omqp8vLyLjkxXhAuFxHQCFeEJEmo1WratGnzlwQ0kiRx4MAB1q9fzyOPPMKiRYvo1KkTGRkZjB8/njlz5lBXV8eECRNo3br1ZV9/dXU1FRUVtGnT5pK+L0kSP//8MxEREezcuVN++q2oqGDr1q3Ex8fLfamo1WoyMzMZOnQoeXl5HDhwgPj4eFJSUigqKgLA09MTLy8v1qxZQ1hYGPfcc88lv1bx8PDA4XBc8raZTCbmzZtHt27d2LdvHxMmTEChUHD8+HHKy8tJT0/ntdde4+DBg6xZs4b27dsTFxfH8ePH2bJlCw8++CBZWVn4+vpy6NAhhg0bxu7du9m3bx+pqalMnDjxkl6nOM+Z1q1bX5ZXTjabjV9++QVfX1969erFmjVriI6OZufOndx+++14e3v/6XW4ubmhVCqbfSycHUkWFBQwZMgQTp06RVVVFbW1taSlpbFr1y66dOlCeno6kydPbnZN1ZmJ2oJwpVx9df2CwOmkyMLCQmw2G3l5eRQXF9OlSxd27NhBZmYme/bsITo6mpiYmCtd1HMyGo1s2rSJLl26ALB//34ANm/eLDft/uWXX1i+fDldunQhKCiI33//nX79+skJzw0NDezduxeLxYJKpWLLli1UVlbSo0ePy1YDcSlOnDhBXV0dXbp0YdeuXXLHhPHx8Vx//fVER0cDsG7dOkwmEz179iQwMJChQ4fSvn17FAoFa9asoUuXLnh5ebFz505Gjx7No48+Sk5OzhXbrrPt2rWLrVu3EhAQgNlsZsuWLXTp0gW73c7BgwevSJnKysqYPXs2UVFRSJLEsmXL6NSpE/Hx8Xz77bccPXqUTp06yR1KCkJLIgIa4apjs9nYs2cPbdq0wWazERISQk1NDbt27UKn0xEfH89bb73FiRMnSE9Pv9LFPSeTyYTZbEalUqHVaqmsrASgoqICLy8v3N3dqauro6amBo1Gg5eXF9XV1ZjNZtLT0zl48CBJSUnY7XaWL19OcHAwEydO5F//+hefffYZer3+im2bsyWWSqXC4XBgMBhQKBS4u7uTmZlJRkYGtbW1PPnkkwwePJiZM2eiUqkoKSlh165dnDhxQm5m7u3tjcFgQKvVkp+fz6hRo67Ydp1tx44dXHvttZw4cYLPPvsMi8WCUqnE09OTqqqqK1KmY8eOERsbi0Kh4I033qC8vFyuvcvOzkapVKJUKlGr1dTV1V2RMgrCpRIBjXDV0ev1HDlyhJ9//pm9e/dSX1/Pq6++SmNjI126dMHHx4fw8HBGjRol91HyT6PVavHx8ZH7WYmKisJisRAXF4dOp8NoNNKqVSsiIyMxGo00NDQQGxuLl5cXjzzyCFarlfT0dAYMGMCECRP49ddf5SbunTp1uqLd10dERMh9pDhzf2w2G5Ikce2113LDDTdQVVWFVqulX79+tGrVCofDQVxcHA8//DAVFRUEBwdjNptpaGggJiaG/Px8fHx8aNu27T/m1Yefnx/h4eFcd911NDQ04OXlJfcj9GfG8PozvLy8CAkJoXv37pjNZlq1akVjYyMNDQ2kpqaiUCjk/nvOHMtMEFoCkUMjXHV8fX157LHHqK2tJTQ0lISEBDIzM6murua+++7j999/p7y8nKCgoCvWCumPaDQahg8fzu7du/Hx8SExMZF58+YxZswYVq1axf79+xk3bhwOh4O9e/diNpvp2bMnc+fOpUOHDowaNQovLy/S09MJCQmhY8eOfPHFFyQlJdG9e/cr2holISGByMhIdu/ezaBBgygqKpLHD1IoFISHhxMeHs6sWbPo0qULQ4cO5eTJkxw6dIiIiAiGDRtGfHw8GRkZcv7Ixx9/jLu7O61ateK+++77R3QFMGLECNasWYPFYmH8+PHU1NSwe/du/P396dix4xUpU0pKCvv372fPnj2MGDFC7nvIaDQyceJEduzYwe7du+nQoUOTEbwF4Z9ONNsWroi/ox8a57ABzp5cPT095Z6EDQYD3t7e8k30crsc/dBIkoTBYECj0aBSqbBYLLi7u8u95jo7vHP2UutsveXm5iYntZrNZiRJwsPDA6PRiCRJaLXaP7XNf7YfGudxcY7OLEmSXCvgHF5CoVDQ2NiIWq2Wk5AbGxvx8vKSzxfndjv3DZxuOadWqy85Kfhy9kPj7JnZOZSGs3WY83hejvPuYvqhcZbJ2fO08zwwmUwolUrc3d3lMjpfCTaX6IdG+CcQNTTCFfVXvx5wjnPk4eGBJElIkoRSqZRbmPxV63eu68++2nEGLQ6HQx6SQaVSyfkngHzzdTgccnNg52fOVioOh8Nlvj/DGXz8mW1TKBTyMQFccjec++7sbTn7mDm3R5Ikl9Y4f2b7Lse2nenM/Q+ux/NycDgcF32eOfN4nPvZmSB+dhkvZpniuVj4JxA1NMIVIUkSa9eupbS09Ir3EPtXsFqt1NfXExwcfKWLctmZTCaMRqPcD87VpLy8nNDQ0BZzTv5TjoVKpWLcuHFXtPWcIIiARrgixGknCFeflhIIClcn8cpJuCLEhU8QBEG4nESzbUEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCiycCGkEQBEEQWjwR0AiCIAiC0OKJgEYQBEEQhBZPBDSCIAiCILR4IqARBEEQBKHFEwGNIAiCIAgtnghoBEEQBEFo8URAIwiCIAhCi/c/FdBIkoRer6eqqgq73Q6A2WymsrKShoYGJEn60+twOBxUVVVhMBjOO4/NZqO0tBSr1fqn19cckiRRXFx8zvVJkkRZWRlGo/GCy7BarRQXFzfZR5IkUV1dTUNDw2Uts3PZOp2Ompqay3JszsXhcFBUVITD4Wiy7sbGRqqqqs65brvdTnl5OSaT6S8p19llaWhooLq6uslnNpuN8vJyzGbzX16OK0WSJEwmE/n5+ej1eiRJQpIkLBbLOY+NJEmYzWYaGxtpbGzEZDLJ35EkCavV2uR4Xwy73Y7NZrvoc9K5HY2Njej1eqxWa7OXIUkSNTU1lJeX43A4/vC7F9o/F1NW53Xyn+DMbTIYDOfcj3a7vclxd37XaDS6/P1n9o/wz/Q/FdAAVFVVceedd/Ldd9/hcDgwGo2888475OTk/KnlOm+AFouFl156iQMHDpx33oqKCh566CGqqqr+1Dqby+FwsGTJEurq6s75+apVqygqKrrgMurq6liyZMk5bwS///47R44cuRxFdSFJEkuWLOGrr7667Mt2MplMLFy4sElA4HA4WLFiBR9++OE5v9fQ0MBTTz3FqVOn/rKyOdntdhYtWnTO/VBXV8ejjz5KYWHhX16OC3Ge/3/FDUKv1zN37lzefPNNHnjgAQwGA5IksX379vMGJrm5uYwdO5a1a9eydOlSvvjiC/kYf/LJJxe9vyRJwm63I0kSGRkZ/PDDD5e0LQcOHGDcuHFs3LiRl156iePHjzd7n/36668sWbKE4uJiZs6cec7vORwOHA4HdXV1vPfee9hstksqp91u59133/3D68LfRZIkcnNzycrKwm63s2DBAh5++GG2b9/OzJkz2bFjhxzoPPLII3z55Zd8/fXXfPXVVzgcDkwmE1OnTkWn08nLTE9Px2KxXMGtEi63/6mARqFQEBkZSWpqKp999hnp6en4+fnRuXNnEhMTkSSJ0tJSuTajuLiYmpoarFYrpaWl2Gw2ampqyMvLw2q1YrVaqa6upqioiOPHj/P2229jMBjw9/dHr9eTnZ0tPxXodDpycnIwGo2EhITg4eGBJEk4HA55nc4nr7KyMqqrq5Ekibq6OkpLS/+wJsBut1NVVUVdXR25ubmYzWby8vKor69HoVAwadIk/Pz8MBgM1NTUUFBQQH19PQBjxowhMjISs9lMdXU15eXllJaWotfrycnJwWKxEBAQwB133IFCoaCqqoqSkhKKiorQ6XQMGjSIjh07YrVaqayspKKigtLSUvmJuKKigsLCQgwGwzlvQM4LUWFhIfX19fJTdGlpKV5eXlgsFmpraykuLsZms1FVVSU/geXm5lJXV4fD4aC2tpbq6moqKyvR6/WUlpai0+nkWqTCwkKsVis2m43q6mq5xunuu+/G3d0dq9VKSUkJ5eXlKJVK4uLiznvB8/Pzw8/PD4fDIS8/Pz9fPt6NjY2UlpbS2NiIw+GgoqKCioqKcz7xOmsUSktLqaurk2sS8/Pzqa2tRaVSERsbi9lslpedk5ODXq8nMDAQLy8v7HY7ZWVllJeXy7VH56u1sdlsFBcXU1FRIT+JFxUVyTVhdXV11NXVkZOTg9lsxmKxUFVVRXl5OeXl5fJ5W15eTkFBARaLhYqKCl577TUqKysve1Bjs9mYMGECM2bMwGg0UlFRgUKhoLS0lMrKyibzKxQK7HY7SqWSESNGMGHCBPbu3cvKlSux2+2MHz+eiIgIeduc56XNZkOn08nH1Gq1ysevqqqKb775BovFQlJSEsOHDwf+r0bAZrPJD0g2m03+3tkaGhoIDQ1lxIgRaDQa1q5di81mw2g0YjKZ5OU0NjbKNQ8OhwO9Xk9eXh5du3YlMDCQO++8E4VC4TKvyWRi0aJFlJaW4unpyT333INarZZ/X84aC4vFgtlsxmQyyefUmSRJ4ujRo6xbt478/Hx5mrMcZ5fLYrFgt9vlc99qtWK32+UyOWt6nOe1wWCQl+HcxzabDb1eL3/v7NrkmpoafvnlF9q1a4dKpaK0tJSePXty/fXXc/PNN/PKK69QXV2NUqmkpKSEkSNHMmjQIL777juMRiM7duxg06ZNLg+RDQ0N5Ofny/vnz9TaCf8M6itdgCshMTGRa6+9lhdffJF58+ahVJ6O69avX49SqWTlypWMGDGCzZs3065dO26++WaWLVtGz549yc/PJzc3F6VSSdu2bfn666/p3LkzkZGRHD58WL6BZmRkkJ6ejlarZcKECWzduhWHw8HBgweZOnWqXJZ169ahUqlYsWIFo0ePprKykpiYGPLy8hgwYABLly6lR48e5Ofnk5aWdt5tqqqq4j//+Q+33XYbW7duJSEhgeDgYPbt28cLL7zAE088wSuvvMKOHTvYvXs3Xbp04fDhw0yfPp3//ve/3HrrrSgUCmbNmsWECRP47rvvGDduHJmZmbRp04b27dvz/vvv8/XXX7Njxw7CwsJ48803mTJlCps3byY2NpZOnTrx3//+lzvuuIOff/6ZGTNmUFhYyJEjR+Qb5NSpUwkMDHQpu8Vi4e233+bGG2/k448/Ztq0aSxcuJCkpCROnTqFJEkcP36cjz76iLlz57Ju3To6d+7Mvn37CA0NZeHChbz44os899xzXHfddfj5+dHY2MigQYPYs2cPcXFxbNy4kZCQEJRKJa1atWLGjBlcc801dOrUiXnz5jF79mx++eUX4uLi2LBhA/fdd1+zz6f09HROnjxJXFwcH3/8MS+88ALz589n8ODBHDhwAF9fXzkw7tevH0FBQU2WMX/+fHr06EFWVhb9+/fn22+/ZfDgwcyaNYs777xTnq+6uprffvsNDw8PNm/ezLvvvit/tnLlSvLz83n55Zf55ZdfGD9+PBqNxmU9NpuNhQsX0r59exYtWsT999/PTz/9xODBg1m0aBHPP/88r732GjExMdTV1REdHU23bt147bXXmDBhAr/88guzZs3i999/R6FQAKdrHW688Ub2799PfX09ISEhzd53zeHv748kSZSUlJCUlESrVq1QKBSYzWaKiooIDw93mV+SJHbs2EH37t3l7Y+Ojmbv3r3ExsayePFiXn/9dWbPno23tzcnTpzg9ttv59ixY5SVlXH99dfj7u7O4cOHMZvNhIWFcfz4cQ4ePMiAAQN47733eOaZZzCZTKxdu5bWrVuTnp7OrbfeyrRp07jttttYtGgRU6ZMISkpyaVsmzdvJikpiYMHD5Kfn8+YMWPYunUrW7Zswd3dnT59+lBaWopKpcJut3PjjTeyYsUKWrVqxd69e7n33nt5++23GTJkCBqNhl9//ZWIiAgOHz5MWloa8+fPp2PHjvzwww9ER0czcuRIli1bRkxMDFu2bOHOO+9k0aJF8kOX1WrlySefdCmjyWTi+PHjDBs2jJMnT9KnTx8aGxvldR04cIB77rmHVatWERkZya5du7jxxhv55JNPeOutt/jwww8ZO3YsmZmZZGZmYjKZGDt2LLm5uYSEhPDzzz9z3333YbfbOXbsGHq9npCQEPbs2UPPnj3x8fGhtLSU0aNHy8fzl19+IT4+Hnd3dywWC3v27GH69OkoFApCQkLQ6XQUFxejVquxWCxYrVbWrFnDiBEjMBgM6HQ6unTpQkFBAQkJCSgUCrnWJz4+nldffZVHHnmE6Ojoy3ruCn+v/6kaGieFQsHQoUO55ZZbePbZZ9HpdFgsFpYsWYLRaKRTp05IksT48ePZvn07BQUF9OjRg19++YWGhgaio6Px8vIiKioKX19fnn/+eXr16kVwcDBRUVEoFAquu+46xo8fT1ZWFmvWrKG0tBQfHx8iIyPlpw+z2cySJUswmUx07twZi8VCY2Mj8+fPJyUlBbVazbZt28jOzqZdu3YXfPINDAyUL4h9+vRBpVJx8803U1JSglarRavVIkkS4eHhtGrVigkTJlBWVoZarSYoKAibzUZkZCSenp6MGjWKoKAg2rVrx9ChQzl8+DAxMTHYbDaUSiXDhg3j2LFjdOzYkbS0NCIiIrBYLISFheHp6cnNN99McHAwZWVl7N27F0mS6Nu3LyEhIQQEBDQpu1qtZuTIkTgcDk6cOEFubi67du2iZ8+epKamolQq6datG56enpw6dQoPDw/0ej179uzBbreTkpKCu7s7Wq2W0aNHc+edd3Ly5EnWrl1Lt27diIiIIC0tDZPJxNGjR+Xj9/TTTzN48GD54tarVy8CAwPlGqbmcDgcfPvtt3Tt2pU+ffrQ0NDAvn37OHr0KJs3b6Zr167Y7Xa+++47wsPD8fHxOedyysvLWbx4Md27d2fLli0oFAo6d+5Mz549WbhwoTzf1q1byc3NxcPDw6UGSaFQMGrUKI4dO0ZhYSFhYWH4+fk1WU9hYSFZWVl0796dZ555hqioKIYNG4ZarSY/Px+TyYS/vz9dunTh5ptv5ujRo4SFhaHVahkzZgw+Pj4UFRWxcOFC0tLSGDp0KPv27ZNri+Li4uRA53JxPkFv3LiRhx56CI1GgyRJVFZWnrPm0uFwsG3bNvr16yfXYpw4cYLExESqqqqIiIjA3d2doqIi/Pz8eOaZZ5g1axaBgYFYrVZ8fX2ZMWMGvXv3Zvjw4XTp0gWz2cxtt91GYGCgvK1vvfUWycnJJCcns2PHDrn2MyYmBg8PjybH2mQysX//ftq1a4fVauXNN98kJSWFuro6zGYzjzzyCDt27ECv11NZWUlSUhJffvklgYGBxMTEEBMTQ2BgIKWlpcTFxfH5558THh5OWloaN910E5Ik0adPH1JSUigsLKRNmzYsXbqUmpoaevToQWlpKUVFRZSUlODl5UVsbCze3t5N9vWmTZswGo04HA6ysrKQJIkvvviC4OBg0tLSuPnmm1mwYAHu7u707NmTW2+9FYPBgK+vL5WVlfIDT1VVFQ6Hg2eeeYbs7Gyqqqro2rUrer0eLy8vPv74Y/r27cvIkSPp2bMno0aNYv369URHR7sEgg6Hg127dtGlSxcUCgVFRUWYzWYSExMBqKysxGq1EhISwo4dO0hOTqa+vp5hw4bx6KOPsmrVKiRJQqFQcOLECbnmuLq6GoPBgFqt5vnnnyciIuKynrfC3+9/MqBxOBwolUruu+8+goKCWLp0qfzeuXv37kyePJlOnTrRvn17JEliw4YNtGvXjoaGBuLj4xkzZgyjRo3C4XCgUqlQKBTyTdFJqVSiVCrlVwQBAQGMGjXK5WlbkiTq6+vp1q2bvM4RI0bQt29fXnzxRfR6Pa+88grbtm1jwYIFzUoGVCgUKJVKlzI5f8xnzuP8++xlOac7l6FUKpu8h8/Pz2flypU8+uijTV5rOL/vvJEMGzaMgoICysrKuOOOO85ZZr1ezy+//CLfOI1GI5WVlS6vZ9zc3OSnwPj4eHk/DBkyhAceeACtVotCoUClUqFUKnnqqacwGAy8/fbbHDp0iAMHDhAbGysvzznfmfvh119/xWazERISclGvTWw2m/waJDAwEI1Gw7PPPktlZSXvv/8+7du357777uPtt9/m6NGj51zGXXfdRXR0NP/9738xGAxyQOXn5+dSy2IwGPD29mbEiBHce++9Lsc1NDSUHj16MGvWLDp27HjOwMJsNsuvJP39/TEYDPz888/4+vri6+srb/fZ5/SZx9WZYFlXV4darSYgIACVSiWv43K/cjKbzWzevJnrr78eLy8v4PRvOC8vj1atWjWZv7a2lqKiIlJSUpAkib1791JbW8uoUaPYuXMnPXv2RK/XU1hYyIABAygoKECn0xEbG8t//vMfGhoasFqtBAcH4+Pjg4+PDydPnqRdu3acOHGC6OhoamtrKSkpoWPHjhw7dkz+rHPnzvj7++NwOAgMDHTZF868nWHDhtGzZ0+57Onp6YwaNQqFQsHvv/9OSkoK48aNIzo6mm3btpGWlsa6devo2rWrnKTu4+PD7t276dixI2q1moiICDIyMkhNTaWxsZHKykoiIyNZs2YNgwcPpqGhgYaGBsLDwykqKmLQoEEcPHiQ7t27u+y7qqoqampquP766+nVq5cc5O7cuZOUlBRUKhWhoaFs2bKF1NRUVCoVkZGR7N27l5SUFDZt2kRcXBzu7u5kZGQwatQoPDw8WLFiBYMGDWL//v1oNBpKSkpQKBQEBATg6+uLn58fiYmJVFdXk5ubS0JCglwm5/U6KChIrv3u2LEjWq0Wq9XKggULuOWWWwgJCWHbtm3ccsstpKWl0b59e/Lz8/H39yctLY1evXpx4sQJebnOY2m32/H393c5h4WW6X8qoJEkifLycvbv309FRQUajYapU6fSsWNHPDw8uO6663jqqaf4+OOPqaiowM3NjZtuugkfHx+0Wi0jRozgrbfe4r333mP//v3U1NSg0+lobGzEx8eHuro60tPTaWhooLKyksrKSgwGA71792bx4sW88cYbrF+/HqPRiNFopLGxkV69evHUU08xc+ZMysvL5dcpvXv3prq6moMHD3LnnXfi6+vLtm3beP7558+Zh+F8l+3Mf6itraW2thaz2Ux5ebl8kautrUWv11NdXY3RaKSmpob6+nqqqqqorq7GbDZTX1+PwWCQL25Go5HS0lL5CXTmzJn07duXvLw8fvvtN6qrq6mpqZHXV1tbS2NjI/X19WRlZVFeXs6RI0fIzMw8Z9kNBgMHDhzg999/R6fTodfradWqFZ9++il79+6loqICs9nMgAEDcHNzo02bNnTo0IGSkhJefPFFvvvuOwwGAwaDgbKyMsxmM+vXr2fs2LHExMRQUFDA3r17OXbsGHl5eWRnZ6PX62loaKCurg6DwUB1dTWZmZns2LGDmpoaDh48KOfAnCsXxXn8ampqmDx5Mps3b+bw4cOEh4eTlJTExo0bGTduHJGRkWRkZKDVarnxxhtRq5u+5XU4HPKNp1OnTgwcOBCz2cyePXvIzs7mtttuo7KyEp1OR8+ePVm7di2vvfYaq1atwmQyodfrqampAWD8+PGYTCaX4O1M8fHx+Pn58fjjjzNv3jyMRiMHDx6U9/2BAwfQ6XTU1dXJT7CVlZXycdXr9RiNRiZNmsSqVavYv38/aWlptG3bFpvNxu+//35ZcxHsdjuzZ89m0aJFTJ8+XU7GraysxMvLi5iYmCb7cvv27QDk5eXx66+/snnzZj766COCgoIoKSnBarWSn59PUlISgYGB+Pn5oVKpKCsrIy8vj4CAAKxWK3v37pVfl9bW1qLT6cjOzpaDqsjISAoLC8nMzOTRRx8lMzOTUaNGUVxcjEKhkI+Jczs2bNhAeHi4/IobTp/7RqORLl264ObmRmhoKCUlJWRnZ2M2m3E4HPz+++/s378fpVIp187pdDr5dc/evXsxGAwcPXoUhUIhB9f19fW0bduWmpoatmzZIgdrHTt2xN/fn8LCQrnVGJxuyTh37lwSExMJDw+XayurqqqIjo6W12WxWEhISGDXrl3s2bNHzuly/k69vb0pKSlBrVaTlJQkLz89PZ1Dhw7h7u6Ou7s7RqOR/fv3k5mZiSRJeHp6ysfDzc1N3kfOoEmlUmE0Glm3bh1+fn5kZGQwf/58oqKieOqppygoKCAnJ4ewsDB53zpTAsLCwggKCuLkyZPodDp0Oh1Go5GkpCTeeecdl0BHaLkU0v9QuzVncm5dXR0BAQFydWtjYyNarRaHw0F+fj7e3t6EhYWhUCjkhDdPT08cDgfFxcXY7XaioqLkG39wcDAajYbCwkJ8fHwwGo24u7ujVqvR6/WEhoZSW1tLQ0MDsbGxclKqj48PXl5eLuusq6uTq2S9vb3latvQ0FAaGxtZvXo148aNa3JjNBqNVFVV4evri8lkwuFw4O3tTV1dHV5eXuj1enkbrVYr3t7e1NfX4+fnh06nw83NDTc3N/R6PQEBAdTW1uLp6Qmcrir38PDAYDAQEBAgN9F2XoScN3ytVktDQwMBAQHU1dXh6enJ6tWrad26NXD6iWjIkCHnzHkoKipCo9FgsVjw9fVFpVJRWFhIQEAAHh4e+Pr6AqDT6eSq/IaGBkpLS4mKikKtVss3OX9/f7kaOjAwEJVKRUFBAWFhYfJ+NxqN8hNfTU0Nfn5+2Gw2Ghoa5H3o5uaGxWIhNDTU5QIL/9fc39PTk4CAALmpfnh4OO7u7vL6g4KCsNvt1NfX4+bmRkhISJOaE2f1t8lkwtfXFx8fH/R6PRUVFQQGBspV+Xa7ndDQUOrq6uRXG3D65u7t7Y2Xlxd5eXkUFRUxYMCAc9bQOGtXCgsLiYiIwNvbm7KyMiRJQqlUolKpsFqtuLu7o1QqMRqNeHp6yudFXV0d3t7e+Pr6UlZWht1uJzw8HLVaTXFxMVqtloCAgMv22smZfOx8teZMgnYmiHbo0MFlXWe+inJzc5PPHWcQUVxcTFBQkPzbdp5LzgTxiIgIlEolFRUVGAwG+RVycXExUVFRcoAbEhJCTU0NJpOJoKAgNBoNtbW1+Pr6yr/viIgIuWx2u52Kigr5t+w8n+x2O3V1dXJeWW1tLTU1NURHR+Pu7k5xcbFcm6jRaNBoNFRXVxMZGUljYyNlZWW0atUKLy8vSkpKCAgIwM3NjdLSUpdEfz8/P3x8fOTEY29vb4qLiwkODsbDwwP4vy4AnLVSdXV16HQ6uWylpaXyuoxGI8XFxYSFheHj40NFRQUqlQqNRoPBYCAoKEi+FgDy78P5yi4sLEx++IiMjESpVKLX6zl48CC9evVqUuP3888/07dvX3ldcLrW1vmwCaevB/X19fj7++Pj44PFYqGyspLAwEA8PT2pqqrCZDIRHBzMtm3b8PX1pWfPnpSWlhIYGCjvB6Hl+p8KaFo6Z+sAX1/fy56n8Few2WxMmzYNtVpNTEwMcXFxXHfddU2CA+HPczgcvP7665hMJp5//vnz5upcDerq6qisrKR169Yt4ncgXJgkSfzwww/4+PjQv3//cwYWzmDw7IehS6HX6ykoKKBdu3YutWVCyycCGuEv42yWWVNTg0ajwc/PT1xA/iLOzvfUarWcT3S1OjuvR2jZnE3JVSrVOV/JOueBy3PMxflz9RIBjSAIgiAILZ54XP4LWa1WCgoKztnqw2azkZubS1FRUZPPdTqd3OnZ38HZKZvBYKC+vv4PO0czGAyUlJRctvJZrVa5M7+ysjLKysrkDub+CWw2GyUlJRc1vMCZHeAVFRVRV1d3Ub2Snu/ccXbCV1JS8pd1BObsAM2ZWyMIgtASiIDmMnP2vGqxWNDpdPz0009NWvZIksTy5cs5efIky5cvb/L5iRMn2LZt299WZmfexfHjx8nKymLHjh1N5nHe5Ox2O0VFRaxdu/ayrd9ms/Hxxx8zdepUjh07xq5du/j2228v2/L/rLq6Oh5//PFmdwMvSRLp6elMnTqV48ePc+TIER566CG519Xm0Ol0/Pjjj+dsFXbixAmmTJni0pzearVecPywi7Vlyxbeeuuty7Y8QRCEv9pV1VOwJElyNn2rVq1Qq9VUVVXJfYtYrVZqa2vlrHhni5rKykpatWoldwXe2NhIdHS03AwxOjraJeve2WLB2Uuls8WLs9XS+++/z8CBA0lNTWXChAkolUoMBgO1tbV4eXmhVCr55ZdfeOCBBxg7dqzc26nZbMZms9GhQwdiY2Ox2+1UV1ejUqnkfjIUCoXctX9wcDBarVYum7P7f2ermdjYWFQqFXV1dZhMJgIDA7FYLFRXVxMVFYVGo5GbHru7u+NwOEhJSZFbJRmNRrnVg1qtZvr06dx///3ExMTg7+8PnG7tU1ZWhpubG+Hh4RiNRgwGA3q9Hh8fH7lVQ319vdx64ux3156enoSHh6NSqRgwYABGo1FuQl5WVoZGoyEsLAyr1UpDQwN6vV7uHM153J01S1FRUXLeTkBAADqdjrCwMPmYeXh4EBISIg+xUF1dTUhICEajEYvFQlRUFHq9HrPZTE1NjdwZoI+PjzxOjnNYgbPPC6fq6mr++9//MnPmTDp06AAgt4Q581y1WCzU19fLLc8CAwMpLi4mJiYGX19fJk6ciFKpxG63y0NJBAUFERoaKg9wqlAoiIiI4NtvvyUgIIAbbrhB3i8Oh0Me0LK2tpa4uDgAuXVeWFgYRqOR2tpauc8V51AcHh4e8pATVqsVHx8fqqqqaNWqFSqVioqKCiwWi9wqqKysDIvFct59IgiC8Fe7qgKaw4cPs2nTJsLCwnA4HHTq1IlDhw4REBBAhw4dWLFiBQMHDmTnzp0kJCTw+eef8/bbb/Pf//6XF198kYULF7r0lKpSqdi0aRPXXHMNN998s3wj3rJlizwWSu/evfn0008ZNWoUM2bM4MUXX2TPnj2kpqZy5MgR3n77bb755hu+/PJLrr/+enbv3k2vXr2or6+nrq6O8vJyXnjhBebOnSv3I1FYWMixY8d44IEHuO+++7j99ttZvXo1jz/+OG5ubmzfvh2VSsXOnTt58803CQ0NBU43W3zggQfkDrH8/f0ZNGgQ77zzDr1796Zt27aYzWYMBgPZ2dk88sgjLFy4kMGDB8tjSS1dupT8/HyeeOIJFi5cyDXXXMM777zDI488wr59+6ivr+fYsWOsWbOG9957j08++YT+/fuzceNGEhISsNlsrFmzhn79+rFt2zZmzpzJl19+yZAhQ9i3bx833HDDeY9fXl4emzZtYteuXYwZM4aZM2cyePBgVq9eTadOnTAajfzyyy906NCByZMnExkZiSRJFBYWsn37dux2O1988QWPP/44zzzzDBMmTMBgMDB27Fg+/vhj+vfvz/fff88zzzzDp59+ilqtJjQ0lPT0dMaNG8fixYt5+eWXWbx4MZWVlURHR5ORkcHs2bPlMm7ZsgWr1XrO8+LM89DhcMiBoXM8LWe/Is7E6AMHDvDmm29yzz33MH/+fG655RZyc3PlHllfe+015s+fz5o1a1Aqlfz+++/4+flx5513UlNTw86dO/n555959dVX2b9/P506dcJms8kBTWVlJf/6178YOXIkBw4coH379rRu3ZpvvvmG1NRUOnbsSEFBAT179uSDDz7ghRdeYMuWLYSEhFBQUIDNZiM7O5u5c+fy1ltv8fzzzzN9+nSOHTuGm5ub3Ouyj48PkiSxbt06evfuLXcSJwiC8He6ql45RUdHk5aWhsFg4NixYyiVSn744QeUSiUhISGUl5ezaNEiunXrhq+vL2azWa7hcPaA2bFjR6ZMmcKyZcswGo2kpKScs5nxvHnz5O7lR40ahd1uJzc3V+4hNDExkfj4eDnPobS0lCVLlpCamkpwcDABAQG0a9eO5ORkUlJSSE9Pp7y8nNTUVOLj4+UaFU9PTwYPHkzXrl3JyckhMzMTg8FAnz59CA0NdRk3x8/PD19fX4YMGcLzzz/Prl275P52nnjiCQoKCqiuriYoKIioqChWr15NQEAAHTt2JD4+Xh6Q0Ww2s3v3bgC6du3KSy+9RGRkJP7+/iQmJpKQkIDFYiE7O5ujR4/StWtXRo0axXfffUdISAihoaHceuut1NfXywNILl++nNTU1Asev4CAABISEujQoQOnTp0iNzeXzp07M3LkSL799lsiIiIIDQ1t0k356tWrqaiowN/fn7i4OAIDA7n//vv56quvGDlyJO7u7gwfPhy1Wk1RURGNjY2EhISQmJjIzTffjNFo5PrrrycqKorS0lLCw8Pp2LEjjz32GBqNRu50y263s3jx4gueF3D69c+ZvRC7ubnx5ptvcuTIEZcedyMiItBqtQwfPpyIiAgSExMZNmwYmZmZREVFyfkrO3bswMfHh7S0NLlb/YCAAEaPHk3r1q0pLS0lJCSEuLg4ufYRTvfZ4uPjw6hRo3juuefkjt0CAwN55plnyMjIoHXr1nTr1k0eY2f16tVcd911dOvWDbVajY+Pj8vvxGg0snz5cnr16sU999xDr169XIYMEbUzgiBcKVdVQHPw4EEOHDhAfHw8ABERETz//PN8/fXXbN++nX//+9+0atWKV155RR5l1jlqrPPm4bwg19fXk5yczG233UZqaqpLcmSnTp24//77efvttzlw4ABLliwhLCwMLy+v8yZR/uc//yEgIIBp06a5JIcqFApuv/12vvnmG9zc3FxuSM7PnUMQSJJE//79aWhoIDc3l3vuueec61IoFGg0GkJCQnBzc5Nvrs7XVKNGjWLMmDFyB23O7zg584BOnTqFw+HAy8uryei3cHp4BGcHclqtVn6ddOZNG+CJJ55AqVTy5ptvYrPZzjt8g6+vLzExMYwcOZLw8HC54zYvLy952WcONeHkHOF85MiRjBs3Th6SIjAwkI0bN2K1Wlm6dCk+Pj4EBATIQ0Gc+Z+zvM6cFefQFWd2wOh8tXXmeXGu7UlOTsZoNJKbmwucfqXm4+ODv79/k5qLM4+v899n582MGzeOvXv3otFoGDFihFw+5/fOTA4+1351ng+hoaFy53nO/xcXFwOnB4BUq9VUVFS4jI905vAZzldupaWlVFVV4enpKXcamJKSwsSJE+nSpYtIJBYE4Yq4ql45lZaWkpGRgc1mIy8vjx07duDm5sbYsWPRaDSsXr2aQYMGyYPLGQwGFi1ahMlkoqioSM4XgNPjrTzzzDP06tWLfv36udQIpKenExYWxsiRI5EkiSNHjrB582a5+/Hw8HDWrVvHtddei8lkorq6mtWrVzNkyBA5WbixsZGSkhLi4uJISkoiPj6eDh06yMMzNDY2Ul1djcVioaamRu7x98SJExQWFqJSqaivrycmJsYlN8PZZXtWVhZjxoyhsbFRbr00fPhwnn/+eU6cOEHHjh0ZMWIEL774ImFhYZSXl5OXl4fdbken05GamsrcuXN5+umn6datGzfddBNeXl6sX79e7sk2PDycbt26sX79ekwmE5MnT5a7FHf2olpZWcmWLVu44YYb+O233ygsLGTu3Lm88sorcp8TJpOJkpISuZdkb29v2rdvT4cOHdi4cSO1tbXcc889VFdXy/kfzu7nAYYMGcKUKVPIz8+nQ4cOdOjQgfz8fF566SWeeuopkpKSOHbsGJs2baKxsZF9+/bJ4xA5e+itrq5Gr9fLrauOHj1Keno6rVu3JiwsTM6BGjp0qMt5sX//ftLT05k2bZocDEdERDBt2jRmz57N8OHD0Wq1qNVquUt2QO6h2JnX5RyQUKPRyPvDZDJRV1fHvn37KCwsxGKx4OHhQWRkpMuwFbW1tURFRbF161Y6derk0vmYcwgFvV7PxIkT5S78dTodEyZMYPbs2Rw8eBCAMWPGUFZWxgcffEBoaKjcu7FOp2PhwoXy0BrDhw/n8ccfp1evXtx0001cf/31PP3001x77bX0799fDPInCMIVcVX1Q2OxWCguLpa7JQ8KCkKn08kDmzlHtfXz80Or1VJZWYnD4UCj0eDp6Ul9fT0AwcHBAHJ3/M4nW6fGxkZ0Oh3u7u4EBARQUVEh16A4hwtwdtHf2NiIn5+fnHDsfEqvra2VR+Q1mUwsX75cHtKgpqZG7pa9oaFB7qpfoVCwYcMG4uLicHNz4+TJk6SlpcnJnpIk8e9//5u77rpLvhE7x0by9/fH09OTiooKjEYjUVFRqFQqqqur5bL6+Pig0+mw2WwEBwfL4/hERUXh7u5OVVUVcHp0bKPRKO+nsrIytFotgYGB1NfXy+O5NDQ04OfnJyfcBgQEyEGEc2gJOB2EOQfdcybPOjvlKysrw8vLS+5232w2ywNAOjmTwRsbG4mKipJHCvby8pKTtZ1dvjtfEzlzWTQaDTqdDl9fXxobG3F3d+fnn39Gr9czevRowsPD5eRhrVYrdxnvPC9qamrYtGkTY8aMcTlHnMNslJaWolKpCAsLw9vb22UwTGdg6+fnR0NDAxqNBoVCIQ814Rx2YOnSpfTs2ROLxUJmZiYjRozAZrPJw1a4u7vLx9Z5XJ37ddKkSTzzzDNERETIydBGo1Heh/X19dTX1xMaGoqHhwdWq1UewsPd3R1fX195+A2NRiNvQ1FREd7e3vLQDkVFRXh6ehIaGiryZwRBuCKuqoCmJdq6dStz5sxhypQpdO7c+YI3A4fDwXvvvUdDQwNxcXHEx8fTp08fOQm0sbGR++67j3vvvZcBAwaIXnkvgc1m48MPP8ThcPD444+7BE7n4hzcz8vL6y+5kev1el544QXCw8OJiIggOTmZ1NTUZh3bmpoa/vWvf/H888/TvXt3EWgIgnBVEwHNFWaxWOQalObccGw2G7W1tbi5ubkMuudcVm1tLe7u7s1enuDK4XBQU1ODJElyjdKVdGbzbq1W6xI4SZKEzWZzyf86MynXbDZTV1cnD9Aozgfhn0SSJI4dO0ZZWRkVFRWEh4djMBgYMGCAXNMtCBdDBDSC0EI5HA727t0rj36emJgov34UhH86h8NBeno6nTt35rXXXmPKlCnk5ubSsWPHJo0jBKE5xDuJ89DpdGzatEm+WVxOztYyjY2NciLqpS6nuLiY1atXc+rUqYv67okTJzh8+DCFhYXs3r3bpWWKM5F3x44d7Nu3j+LiYux2+yW1XsnNzeXQoUPydx0OB4cOHeLkyZMXvayzObc/PT39gmU7ceIER48e/dOtb2pqati2bdsfLufgwYPk5uZSU1PDhg0bLmrIhOaQJIns7Gzmz59PY2Mj/fv3p3///mRlZbFmzZrLvr7msFqt7Nq1i5KSkr993ULLpFAo6NGjB2azGTc3NwIDA0lJSaG8vPyyDq0i/O8QAc15SJLErFmz5ETY5n6npKQEo9F4wfkcDgcrV67EYDCwevVqudfai2U0Gvn888+JjY296KAoPz+fxYsXU1ZWJrdgcfLw8GD69Onk5+ejUCj45ptvePvtty/pRmm1WuVmwM5WVOvWrWPr1q0XtRxn8HL2vi0qKuKLL7644LhGFovlosZROh+73f6HxxZOv+qxWq2YzWY+/PDDZn3nzO82Z4iF6OhoDhw4QGFhIUqlEpPJRENDA507d5Zzqv5uixYt4ujRo1dk3cI/h7OZ/x9RKBRy44aEhASUSiWHDx+mrq6O7du3o9Pp/obSCleTq6rZtslk4sSJE1gsFpKTk6mtraWoqAi1Wk2nTp3IzMxErVaTnZ1NUlISNTU16PV6+vbty6lTp+S+Q5KSkkhKSpL7HzGZTPIP7brrrsNqtZKbm4skSS6didXW1vLkk09y//3307t3b3JycqioqJB7Kj4z3+XkyZMEBgZSUlLCli1bGDduHMeOHaO+vp7U1FQ0Gg1HjhxBp9PRrl07fHx8OHjwIJ6enpSVlTFw4ED27t3L3r17ufnmm3F3dyczM5OgoCCOHTvGddddR3FxMXV1dQQEBBAVFUVWVhaFhYVcc801dO7cmaioKCIiIpg8ebLLTVCr1aLVaomOjqZr164kJSVx++238+uvv3LLLbe47HNJkigqKiIrK4vOnTuzb98+UlJSUCgULnkfeXl5PPXUU7z66qv4+/tTV1fH8uXLadu2Le3ataOkpIS8vDxUKhVdu3YlKysLjUaDj48PJ0+epH379jz55JM8+OCD9O7dW+6Pxjmcw7p163Bzc6Nv377y8bJarVx77bVotVq5ufzhw4fRaDRUV1czcOBAFAoFmZmZ8hAICQkJcpNwSZLIycmhvr6ekJAQPDw85PUdPHgQHx8fsrKySE1NJScnB09PT6655hp8fX1xc3PDz88PDw8PAKqqquSOF1NTUzl16hQmk4mKigoGDRqEWq3G4XAwb948CgoKeOKJJ+SeeiVJolu3bvKynDeCW2+9ldmzZzN27Fj27NnDkCFD8PPzu2y5Ms6m5bm5udjtdlJTUzl58iRWq5X8/HxSUlKIi4ujsLCQ4uLiK1IzJPyz1NTUsGvXLpRKJUOHDr3gvM7WgxkZGYSHh2Oz2aipqSEuLk5+EPL19f2bSi5cDa6qGprt27eTmZmJ3W6nrq6OOXPmEBwczOzZs6moqOD7779nzZo1qNVqnnvuOTw9PZk/fz6nTp1i0aJF/PTTT3h4eDBlyhR5DBxJkli4cCEGg4ENGzawcOFCFi9ejMFgoKamxmWAQI1Gg9lsJiIigr1797JmzRqSkpL45JNP2Lt3rzyfUqnkpZdeYujQoTz00ENMnDiRb7/9FrVaTVlZGevXr2fu3LnU19cTFhbGM888Q3FxMa+88go6nY5ly5axb98+goOD8fDwICoqipKSEr7//ns8PDyYOXMm1dXVzJkzh5CQEAoLC9m0aRMnTpyguLiY119/naCgIJKSkvD19f3D1lUeHh6kpaWRnp5+zs8VCoU8lMDixYvlKuOGhga2bdvGokWL0Gg02Gw2uY+U8vJytFot06dPp6SkhFmzZtG2bVt27NjBd999x6lTp/jll1/QaDR89NFH8tABERERTcpaVVWFu7s7s2fPZu3atXz99deo1WrmzZvHhg0bWL9+Pb/88gsNDQ28/PLLmM1mFi1aRGZmJunp6Wzfvh2z2cwbb7zhUpPT2NjIV199Je/DU6dO8eGHH2Kz2Zg5cyb79++nrq6O119/ncDAQGbMmEFVVRU//PADmzdvlpcjSRJz585Fq9WyaNEisrOz+fbbb1m2bBkGg0F+mnV2lOfr64vdbueDDz4gPj6erKwsPvnkE5enXoVCQZcuXaivr2fZsmXysaypqaGqqoqKigp5XLMzy+FchvPfZyYVn8u8efPQaDQsXryYkydPsmLFCpYuXQrAO++8Q0lJCfPmzaNNmzYuvwXhf5NKpUKv1zf7FXhAQAD33XcfN910E2q1ms6dO1NVVUVYWBgBAQF/cWmFq81VFdDEx8ezZMkS9uzZg5+fH+PGjaOwsFDu1yUwMJCEhAQ6deqEp6cnHTp0kPvmCAsLo0OHDgwfPpzw8HC52t9isbB161ZUKhWjRo2iV69eJCUl8dZbb1FXV+fSqsTd3R2NRoOfn588plRwcDDdunVj8+bNLjeuM3trNZvN7Ny5k8TERG699VaGDRvGunXrSEhIoF27dkiSJHft37lzZxISEqiqqsLHx0fuBt/DwwNJknB3d0epVMp9k7z88ssEBwezYcMGVCoVKSkpTJgw4Zw95V6Is5fhszm78G/Tpg0HDhwgMDCQ9evXU1hYSKdOnYiKigJOB0UajUZubdOuXTu6d++OyWTiwIEDmEwmgoODGThwoFzbcub2uLm5yfv27CbLkZGR9OvXj1tvvZU1a9bIwePkyZNJSkqSx3zy9fUlICCAzp07Ex0dLdfQFRcXExsbS3JysjzoprPMANOnTycsLIyIiAhUKhUajQZ/f3/atWtHp06d0Gq1JCcn4+bmhtlsplWrVk2Cj1tuuYXKykpqa2sxGo2EhYWRnJzMzTffLPeNo1Ao8PT0xNvbm6KiIioqKggLC2PAgAFs27atyWszDw8PYmJiiIyMJDQ0lPLycrZs2cL27duZO3cux48fJyMjQ54/MzNTLpfdbmfbtm18//33FzzuN998M9XV1XLnf0FBQSQkJNCtWzfq6+vZsWOH3B9NZGTkBZclXP38/PyaHYg4r4MajQZ3d3cUCgXh4eGkpaUxYMCAK97CUGh5rqqARq1WM2PGDI4dO8batWt57733aNu2bZMhCZzduTud+dTqcDjw9vYmNDQUSZLk/AQ3Nze6dOmCw+EgPDyc9957j6VLl5KXl9ekHHa7nYiICJd8ggu1PlGpVNTW1pKZmYler6ekpAQ/Pz/5tZaPjw/BwcHnLTOcrvVx5opYrVYsFgu33347o0eP5oMPPkChUMg5Fl5eXn/4NO18gne+djl48CA333wzNTU18ojeTkqlkptuuolZs2YxevRodu/ejVqtloOsM5fpTC4+s/zBwcFyT7h2u52EhATUajUWi0XORzlz314oXyYlJYXa2loCAwPp3LmzPMTFubYPTo9VFRYWRlVVFQ888IDLPFarlXvvvZeBAwfy8ccfuwyRca5lnf3vM5fz3nvvER0djb+/v8tyzp7fObq2n58fFRUVGAwGbDYbsbGxLhd45/AUtbW1pKSkAKfHjBo+fDh6vZ6ePXvSvXt3kpOT5e84X3nB6XMuMTFRzr2y2+0UFha6DLtgt9t59913iYiIICAgoMn2S5KEh4cHJ06cwGazyf+JZE7hQurq6lixYgUrV67kt99+c/lv7dq1rF+/nlWrVrFs2TLWrl17WfLfhP8NV1VAU1BQwJEjR+jbty8dOnTA29ubJUuWEBgYyO7du7Hb7RgMBsrLy/H29qakpAS1Wi13d79nzx5WrVrFsGHDcHd3R6vVUlNTw7///W8++OADPvzwQ7y8vDh06BDl5eWMGDFCHukaTt8kOnXqxMqVKxk2bBg+Pj6sXbsWT09PRowYcd6aEI1Gw4MPPsg777zDZ599hq+vL8888wx79uxhw4YNjBo1Cl9fX/z9/SkrK8Nms6HX6yksLMTT05PS0lLatm1LbW0tP/74I7GxsVRUVLBhwwaCg4O54YYbmDx5MmvWrOHVV1/FaDSed2BFOJ1o6+Pjw+7du1m4cCGrV6/m9ddfp02bNqxatYr333+/yU0rNTWVfv360aNHD0aPHk1qaioOh4PKykrg9I06JiaGVatWodfrMRgMlJSU4OPjQ6tWrRg+fDgrV67kxIkTPPDAA6SkpFBYWMjy5cuJj4+XxwtauXKlS65GYGAgYWFhrF27FoCJEycyceJEXn75Zb788kv8/Pyorq7GZrNRWloq70M4PTp5dnY227dvZ8GCBcyaNUvuLRpOByIbNmygVatWDBkyhJKSErRardxTcG1tLdXV1ajVakpLS/H29qa0tFTunbm4uBhfX18qKioICgpiyZIl8n51Dl1w9hhZ7du3JysrC6vVyh133MHy5cs5cOAAjzzySJM+hzZs2IC3tzfl5eUAci/Lx48fJykpCY1GQ1BQEAcOHGD9+vUcPHiQjRs3smvXLpdRv5374umnn5YH4oTTgX9ISAhLly7F29ubffv2YTKZsFgslJWV4ePjQ5cuXfDx8eGDDz7AbDbT0NAgAhrhgux2O2+//TbTp08nKCiI6Oho+b+oqChCQ0NRq9UcPHiQ559/vkkrTEE4n6uqHxqHw4HVapWTJp1Pi2cPlqhQKOQLusPhQKFQ8Pnnn+Pr68utt94qd71vt9vlQQrNZjMqlUpO3rRarajVajk5Ff6vBsLhcODm5obD4cBisaDRaP6wZ1dJkjCbzfIy4XQnena7vUl5zn5KdpbROcoznA6unN93Vuc6B1F0/n0+Z9aCOKuFnf8uLy8nJyeHtLS0JgNanrlPneVw1sio1Wr52Jx9HJz70FkT5lyfxWKR/61SqVz27dn73DnOkXP/nLksZxnOLJtz+z755BOGDRtGYGAgR48eJT4+ntjYWHnZZ+7Ds7fxzDKcOe3MwS+d8zvLqVKp5PnOHGzz7PPA2UPxmefd2fOduV3O/V1RUcGbb77JG2+8gVarRZIkjEYjNpuNNWvWcP3118vBek5ODj/99BOPP/44KpWK33//nZSUFPmVgXMdZ5bbuZ4zj53z9+Dc36IDv7/HmecANP2tnv258zw5+/jY7XZ5RPU/y2QysWLFCk6ePMljjz3mMubameXevXs3t9xyC0899RSPP/74Oa+PzgYHv//+O+PGjZO3zbldzu081zb9rzrzXDjXv892te23qyqguVQWi4VZs2ah0Wi45557RC+VF+AMMs7MHWqpJEli+fLl7Nixg+joaNLS0ujatWuL3Tabzcbx48eprKwkOTmZkJAQl8/Xrl3LwIEDUavV2O12srKyqK6upmPHjgQEBMiB1NV2kbtaORwONmzYwPLly7nhhhvkQWjHjh2Lt7c3ZrOZ77//npMnT9KvXz8qKirw9fVl5MiRcgAhSRJbt24lMzOTBx544E8f+6qqKo4dO4bVaqV169bExMSct+yff/4506ZN4/vvv6dv377nXLckSTQ0NODt7S3/LnU6HQsWLKCyspLbb7+dhIQEcc5yep8eO3aM2NhYlEolK1asoEuXLrRr1w6dTsdPP/1EVlYWt99+O4mJiZw8eZLk5OQWe707F5F1xencg8ceewxJkq6qg/tXuFJ9nPwVFAoFN954IzfccMNFJUj/U6nVapecmbP16dPHpcbr7HmdSdBCy6BUKvHy8sLb25shQ4YgSRJLly5lyZIl3H333bi7u2O1WmnXrh2DBw+mpqaGSZMm0bdvX/z8/IDTyf47d+6UX1v+WcHBwfTp06dZZZ88eTLp6ek8+eSTrFixglatWjX5/SkUCrmsTs6Bb50NJFryb/ZykSSJHTt2oFar8fLyorGxkfXr1xMeHk7btm0pLCxk8ODBhIaGsmDBAt58803g9FiC/fv3v2r24VWVQ3OpnFW1Z1frC1e/M1ucXe3H3tPT86rfxv8lkiSRkZHBNddcI796iYqK4uDBg/Ln+/fvl/PZDh06RHJysvxqSZIkDh06RO/evZsk+kuShMFgoLGx0eU/vV5/waT8i+HswNNms8ndKTSHzWbj6NGjf9jdxP8So9HIzz//TKdOnVAoFHh7e7u0NktKSiI0NJT6+noGDx4MQLt27VixYsVf0hv+lXJV1tCYzWbS09Pp2LHjOZsa/xGLxcKBAwfkqk4fHx/i4uLQarXN/gGdOnWKxsZGEhMT2blzJ9dccw3l5eXU1NQ0yT9pjuLiYgoKCi7puxdSVVXFgQMH6Nu3L1u2bCEtLQ0fH59zztvQ0MDu3bsZOHDgJY3kbTab2bhxI4MHD5aTkp3DQGRkZNC/f/8LJivX1tayd+9e+vbt+4+qKbLb7Rw8eBB3d3eSk5Mv+fjU1dWRkZFB37592bdvn9wk+2JJksTRo0dRKpW0b9/+or9vt9s5efIkeXl5eHh4oNVq5X5uhH8Oh8PBkSNHuOGGG4DTxz0rK0s+Z3Q6HQUFBZSUlFBRUYHdbufll1+WW8uVlJRgsVgIDw+nsbERm80m/67sdjs7duxocrNzc3Ojf//+8jVCkiTKy8vR6/UXLKtWqyU8PNzlt6FQKIiKiuL999/njjvu4LrrrmPy5Ml/+Puprq7GarXK3SPU1NTg5+f3P93Mu6ysjIaGhnPWsjr3Z15eHhs2bECr1dK/f385LzA3N5cuXbr8zSX+a1xVNTR2u53s7GwUCgVff/01xcXFTZ46ztW52NlpRCqVisOHD/PFF18QFBTEiRMnePTRRzl27Fizs+2dCaTOhOPKykpyc3Ob9PvhHJfH2ST2XOU7c3kXKjecjtQLCwubbPP5ymg0Gpk5c6bcZPpc+6a4uBi9Xu+SjHeuMlxoXU5nNgt2zmsymfj444+xWCxNvi9JErW1tVRWVmI2m/noo4+aNMX+o3Kcr1zN/d75jseZf+/cuZMNGzb84ffOR5JOj6r98ccfYzQaz3ssLlRGh8Mh9yx8dtPw5u4TZ2/FCxYsoE2bNkRFRfHVV19RXFzsMl99fb3La4rmbKNwedXW1mIwGIiIiABO//bT09O5+eabUSgU5ObmEhMTw6BBg+jfvz+DBg3C29sbhUKBzWbjwIEDwOnWoXq93qV5tFKplPtZOvO/jh07ygnrTlVVVRQWFl7wP2drx7MpFAoGDBjA+PHjm11Dc+zYMRISEnB3d0ev1zNnzpz/+dZ1f9S0XaFQ0Lp1az788ENqamrkDjed3ZJcLa6akFaSJNavX8/333/PtGnT8Pb2Jjc3l6ysLPr3709ISAjZ2dmcOHGC9u3b4+vrS2ZmpnxABw4c6JJfEBAQgI+PD61bt6Z169ZYLBZefvllFi5cKD/F5Ofnc/ToUVJTU9m9ezddu3YFTl/sPT09cXNzQ6vVyheR4ODgJi1VDh48yPTp05k6darc2qC8vJy0tDSys7MxmUz07NkTOF1FazQaycjIwMfHh7y8PIYOHSq3JHA4HMyZM4eamhoee+wxeSgHh8NBjx49XKJ3i8XCoUOH5HGkrFYrvr6+OBwO6urqyMnJweFwkJCQwJQpU5g0aRKpqan4+PhgsVjYv38/Wq2WEydOMGDAAIKCgsjKyqKkpARvb29iYmI4duwYycnJlJeXY7FY+H/snXd4G1X6tm/Jki0Xucm9t7gksdOcSjppJITQl7ZLW2BDX9ilswRY2tJC6Ak9JEAgC4H0hBScBmnuvcRN7kWyJVl1vj/yzfwk22mUBYKe6/JlWzpz6syc97znPc+TlpZGQEAADoeDpqYmmpqaUCqVJCQkoFQqOXToEJ2dncybN08KzLZarTz55JNkZGRIbKKHDh2iu7ubefPmIZPJyM/Pp729nalTp0pGrcPhYOTIkbS2tlJYWEhISAgjR46UPEsOh4PS0lIMBgNxcXGo1WqKi4vp7e1lzJgx9PT0SH1nMplISkri6NGjTJs2TZK+0Ol0eHp6MmPGDEJCQmhqakIQBCorKyW5BrEfTCYTycnJaDSaAfeuzWYjPz+fnp4eyWj08/NDEASam5uprq6ms7OT7Oxs2traqK2tZezYsYSGhlJRUUFTUxNRUVF0dXXx8ssv88QTT0irr66uLvLz8/H09KS7u1tyN+fn59PR0UF4eDhDhgyRtiHq6+tZsWIFa9euJSoqCkEQ+Ne//uXiEbNarTz33HNERkZy3XXX0d7eTkNDA0qlkjFjxpzUy+bGzwO73U5hYSGRkZHAcUHZb775hssuu4yUlBRsNhv79u1j+PDhA2LDxO8SEhIYOnQoPT09KJVK9Ho9vr6+0qmYjo4OF6ZpOP5uFNm+4fhEebK4LTHNiSBueyUlJXH99defMq3NZmPv3r2SR6GhoQGdTueyUPojIiwsTDJUxWffYrFIJ33r6+tRqVTIZDLGjh0rnYC02+1nFSHmWeWhUalUeHt7o9FosNvt9Pb20tPTw4oVKyguLmbnzp3I5XLuu+8+6urq+Pe//42Pjw/vvvsudXV1J8xXJpMxfvx4ampqXATTPDw8WL58OUqlkv/+97+0tbXR1NSEyWRiz549fPHFF6ess/gCCQsLY+XKlWzZsgVBEFi7di02m41du3Zx6NAhCgsLeffdd7FYLDz33HO0t7dz4MABdu7c6VJPpVKJRqPBYrHwyiuvkJaWRmlpKW+99ZbLin3t2rW0t7cTExODzWbDYrHw8ssv09zczOeffy7tn4vHnuPi4mhtbeWFF17AarWyfPly8vLyaG1t5aOPPqKuro6VK1cSFRXFY489htlsZuPGjZSXl9Pb28vKlSvR6XQ888wzGI1GVqxYgVqtprW1VdqvVygU7Nq1iz179rj0MUBUVBReXl4YjUaUSiU7d+5k7969fPLJJ1Jg4wcffMAXX3wh7fe3tLSwatUqAgICePTRR11EH1taWvjss88ICQmhubmZnJwcamtrqa2t5ZtvvqGhoYHnnnuOwMBA3n77bcrKyujt7eWjjz6isLCQF198keDgYN58800XSYjCwkJ2794NwP33309JSQkbNmwgMDBw0FWqIAhs2rSJmpoaScPGbrezcuVKjhw5QlFREc8++ywOh4Pt27dTWlpKW1sbTz75JD/88AOHDx8mNjaWjz76CJVKhYeHB6GhoWzevJktW7ag0+l44okn8PDwYPXq1VRUVLBjxw7y8vIwmUwsXbrUhWSxrKwMb29vicixtbWVjRs3UlBQIE0aYsxRZGQkra2tvPfeewwbNozt27ezdu3aP/RK+X+Frq4uWlpaGD16NAcOHKCwsJBZs2YxZ84cZDIZ9fX1qNVqYmJiBqzAm5qaqKmpkZ6turo65s+fT11dnTR2Hh4ejBw5kkmTJrn8jB8/fsApULPZzKpVq9i8eTNvvvkmq1evZv369axZs+ak94IgCNTW1rJ582auv/76AZ4f53RiPhUVFURHR0unuPr6+qQJ+o+MwMBAJk2aJO1KdHd3M3LkSImuo6amhs8//5yqqiquvfZaPDw8aG1tJSUlRfLwnQ04awwamUyGSqWSjBqR5n/kyJG0tbWxd+9eBEEgLCyMe+65B41Gg0ajISsrC41Gc9LAKHHC9fX1dXnoRMr/o0ePEhQUxLZt26ivr2f48OGndZOIdRap9MPCwhg6dCgLFixgwYIFEhOsTqeTNIx8fHwIDAxk2LBhpKamukySYn6iGGR3dzchISFMmzaNnJwcaeKy2+1s3ryZjIwMIiIiJAFJkfZ/+PDhPPfcczQ1NUm05P7+/kRERKBUKvH09CQoKIj09HQyMzNpb2/HZrPR0NCAn58fmZmZREZGSvIFIo9OcHCw9OJJT0/nvvvuk1Zkfn5+jBw5ckCb5HI5KpUKtVqNQqHAz8+PESNGMGTIELRaLTk5OXh6erJgwQLOPfdchg8fzvPPP49Wq6Wqqor29naUSiUPPvigy2mJgIAAOjs7Wbp0KSEhIYwbN04yOjo6OtBoNAQHB5ORkUFYWBhxcXGkp6fT3NxMeHg40dHRTJgwgfnz50tBmHBcTwwgIiKCe+65h8jISMrLy3nvvfcG9c4IgsCGDRukctRqNZ6enhJhY0REBJGRkZx//vkUFRXh6elJZmYm1113nSSPkZSUxL/+9S/p/vT395dW7qLcQ1ZWFhEREXR3d2M2m6mvrycmJobMzEyXmCkfHx/MZrN0r2g0Gnbv3o3FYnHhoFGpVPj5+ZGfn49cLicoKIjp06ezffv2U973bvx0hISE8Kc//YmrrrqKGTNmMGPGDOLi4qTnKTExkT//+c8sXLhwgAESGxvLtddeS3p6OjKZjOHDh3PLLbcwYcKEHxUb19DQQEZGBjNnziQvL49hw4Yxe/ZsySMwGMTYl48//pjrr7/+hHF7giBQV1eHzWZDJpMxdOhQrr/+eubNmyflr1AozqrA1h8DuVzOZZddJnn1g4ODue6665g5cyaenp5MnTqVW2+9lfHjxxMQEIDVaqW4uJirrrrqrDrZe9YYNPB/tPHOcR7ib3GLJj09naCgIPr6+k4Y49A//sFoNLJq1SquuOIKF6IomUzGhRdeyKuvvsqll17K999/j5eXlzSB98dgn8lkMux2u8sqWRAEli5dKkkeOMfXDJaP8/9yuRybzUZQUBDNzc0YjUasVitJSUkupFsymYyqqiqJrt65DI1GwyuvvML69euprq4GGJCmf39FRkYyYcIEKioquPPOO/H09JR0qoxGo4vkgSAcVyl/7LHHeOutt6RtrxONh+hKdY4rET8XY3+GDRuG1WolMDCQZcuWsXHjRpqbmyXPR1xcHL29vdK1ZrOZe++9l8zMTN555x1WrlyJXq8nOjp6UImD/n0vpjGbzaSlpUnp/Pz8qK2tJT09ncDAQHQ6HUuWLCE8PJzVq1fT1tZGZ2enlI/oVRPlA0QPzWBjq1KpaG1tZejQoRI53vbt2zEYDFRVVUn3fn95if7tGDNmDBqNhr6+Pm644QaX70VivZycHAThOGmfQqFwITN0HpOwsDDpPrJarQwZMmRAmW6c3VCr1WRlZUkLsOTkZDw9PaXTV4PBZDLx7rvvctlllxEeHj5gK17cXvr+++/ZuHHjCSfdrKws5s+fPyDg+I8IHx8fpk6dOuiWb39aCg8PDyZPnnzWCYCeNTE0cHzlYTabycnJwcfHh7a2Nmw2G97e3kybNo29e/fy0EMPsWDBAiIjI9FoNGi1Wry9vaVJVSY7rurc1NSETCZjzZo16HQ6Jk2axOzZs13Kk8lkjBo1ijlz5pCdnc2ll17KmDFjcDgcdHR0IJPJaGxsxM/Pj+bmZrq7u1GpVPT29korksDAQDQaDdu2bcNqtUoR/HFxcaxbt46goCCKioqA48duGxsbCQ4OprW1FbPZLDESiw98VlYWH374ITNnzuSqq67im2++QS6Xc/vtt0s3s1wu54477uDdd9+lvLyc6Ohoqqur8fX1paWlRdJCmj9/PtHR0YwbN47169czZswYAgMDaWxsxNvbWxJa9PLyoqmpiZycHI4ePYqfnx+33347s2bNYu3atYwaNYrQ0FBqa2slQ+uHH34gLS2Niy66CLPZTHBwMI2NjVgsFvr6+lzaNG7cOHbt2oW/vz/BwcE0NTVJmlW33HILb775JhkZGVxzzTXk5+cTEBDAeeedx8yZM6moqOD++++XxCtFmEwmdu3aRWJiIhkZGbS0tLBz504yMzNpaGjg2LFjBAYGSqrgra2tWCwWvL29MZlM1NXVsWnTJgICApg4cSKffvopNpuNWbNmsX//fh5++GEWLFhAQkICOTk5kgfm66+/RqvV8sgjj0h1Wbx4Ma+++irNzc1ER0fT2NiIzWZDp9Oh1WpRqVQSh8iSJUuoqqrikksu4c9//jOPPPIIjz76KFdffTVDhgzB19eX3bt3o9frsVqtNDc3o9FopHilnp4eysvL2bNnD3l5ecTExPD3v/9dOsHk7+/P66+/zqpVq2hsbEStVjNkyJABp6Wys7PZtWsXixcvZtq0aWzYsAGTyTTAQHLj7IfoTayuriY8PFzaRvf09KSiooLw8HDUarX0/hG1zTo6Ojh48KCLgKq4SGhpaeHQoUN8//33fPbZZyf0HIkLAjeQ+vx0cLaQo/bHWcUULAa3enh4SMaJ+LkYC3IyGnuRi0QMlhIhfn4iJsvBKP/FFfxgNPn95RJsNptEjy/Ww+FwDKCcd87PWQLBmUNFDAYTb2xnZt/+qyBRikC83pmSX6yTWL4zzX3/NgGsWbOGiIgIySBwOByMGjUKi8Xi8sJxrrvNZpO+O1mbHA4HNptNkp3on06UfBDH3bnuzv3h/FIUx1iUUhD7Q5TMEK8dbPwKCgp45513ePHFFyUZCfEEmrOXUBwDMTBPqVTS1NSEVqtlzJgxLnk6j4WYR39pBXFcxLwAlzYA0v0vwll+QbwnX3rpJS655BICAgLIz89n6NCh0haV8zj19fUhk8kk9mDn8RbHRCxXNK7/CHw+bgyEIAisWLECT09Prr32WkwmEx9//DEXXHAB33zzDTfccIN0X+p0Oj799FNMJtNJ84PjBvZVV13lZm9347RwVnloTmWhKhQKiatAnDyAAda/uC97umU6n44S4fz3yfalT7TCcLag+1vSp8rPOc7nRIF2g/WVc77O3znX5UR9lpqayueff86RI0cYM2YMkyZNGlCX/tedbh/J5XKpPoOlc65r/3aJ8R790X+Mnet6sn4RBIG2tjZ8fHzo6emR5AWc2+J8nwEu4xsSEjIoH0f/Mk90//W/V/qnG+z+dx4zh8NBamoq7777LrGxsZxzzjmEh4cPes3JAi2dxwTcLMN/ZIiLiEOHDnHbbbcBx7li9Ho9oaGh1NfXYzKZ8PPzA44bKTfffPOvWWU3zlKcVQaNG78eRo8eLR1b/71LCJwKs2bN4txzz/1RAZS/NiGgXC7nwgsvZNGiRcDZP1Zu/PLQ6/Xs37+fiIgIamtrSUhIGLBYcX5WRC+46Dk9kee7p6cHs9k8QJPMDTdOhLNqy+n3AueA0N9CPeDUXBGnStM/vfjS+qXb+L/uy1OV1//70+3j082/f5rfyr3kxh8Xg00hVquVNWvWMGrUKEpKSrj44otdBDEtFgvvv/8+N9xww6AeapvNxs6dO2loaHDHZblx2jirTjn9HiAIAhUVFRw4cOBH8XUIwnGq8b6+Pjo6Ok5JOX6yfEpLS/niiy9oamo6adq6ujr2799/WvW1WCxs2LCB7u5utmzZQnt7+4+q3+lAlBw4fPjwL1aGCEE4zpi8ffv2E2rZ2Gw29u/fT01NjUsd169ff1osqEajkX379vH111+7nHrrX48dO3bQ1tZGXV0d27Zt+9m0ddxw48fA+QSN+KNUKrnsssuIiIhg0aJFA7yZhw8fpqamhu7ubg4dOsT27duln4MHD+Lh4UFycrKb08iNM8JZZ9AMdlx1sCPBJ/pssDTOP1VVVVit1hOWdbJyq6qqsFgsVFVVsXbtWunzU9XTGSaTiW+++Qaj0chXX30lBdYNdpR6sP/Fz7RaLV9++SWpqamD0mY7p/Xw8HDRXjpZWSILaXl5Obm5uRK9+onyHizPE/V9/+9kMhklJSVs3LjxpPmcCqeqjwi9Xs+KFSsGGBCCIFBeXg7Apk2byM/Pd7neeZvpRHUShONkh3a7HbVaPaAMs9ksHaEXg29F2vf+khgna+Pp9vPJ+sX5/1O1y40/JsSYNI1GM8ADIwaqz5w5k6CgICIiIoiPj5d+BlPddsON08FZFUPjcDgoLi6msbERX19f0tLSUCqVEivqlClTyMvLw8vLi7KyMqZNm0ZkZCSVlZWUlZVJrIlHjhzBZrMRHh6OIBxnXYyOjkahUPD3v/+dRx99lGHDhpGfn09LSwvTp08fIHFvNpvJy8tDp9ORmJiISqXi7rvv5sEHH8Tf318yTPz9/Zk6dSptbW0SRf348eM5cuQIAL29vcyePVt6wGUyGUVFRYwbN47a2lpKS0sZNmwYJSUlGI1GzjnnHMxmM4WFhRiNRkaPHo2HhweHDh2SjrLPmTOHnTt3UlRUxAUXXEBtba10Mqe+vp5p06ZJ+cXFxUneAlH0U61WU15eznnnnYe/vz8FBQU0NzcTEBBAamoql112GWlpaXh7exMXF+fSL6KHSqfTERQURHx8PHl5eej1etLT05HL5RQVFeHl5UVHRwfDhg3j4MGDTJkyBZVKRW5uLoJwnNL//PPPJzg4WNp6qauro6ioiPDwcEaMGEFxcTFGo5GEhAQXunZniBIQer1eEpYsKCjAw8MDo9HInDlzsFqtHD16lI6OjkHzyM/P57HHHuPpp5/G39+f5uZmPvnkE8aMGUNMTAyenp44HA66urokvpgxY8a4BPS2t7ezZcsWZsyYwcSJE9mxYwfnnHMOe/fuJSsri82bN1NQUMC9996LQqGQuIYUCoXktcvPz8fLy4tzzjnHJYbBbDbz/fffo1AoKC8v59xzz0Umk1FdXU17eztTpkyhubmZtrY2/P39GTVqFL29vZIUxKRJkyTtH5PJxIwZM+js7KSxsRG5XE5WVhZFRUWYTCZSUlJ+lCCsG38sdHd34+npSXt7u8ToLkLk5WpoaJCCi93CqG6cDs4qD01tbS2rV68mMTGRp556CoPBwLvvvotKpeKLL75g27ZtfPrppxw6dEjSPSorK2Pr1q14e3tz//33o9Vqefzxx7FYLHR0dPD1118THh7O0qVLJbXtuLg4PvvsMwwGA/n5+axYsWLAanXVqlV0dHSQkZHBo48+il6vRyaTkZiYiEwmo6uri/j4eJYtW0ZDQwPvvfceQUFBLF26lNzcXF5//XVKSkoGrG5UKhUvvPACmZmZLFmyhOzsbJYvX05iYiIlJSUUFBTwwgsvEBYWhp+fHw8++CBdXV08++yzeHl58fXXX1NaWkpUVBTBwcEkJCSQm5vL/v37USgUvPPOO2i1Wj7//HPCw8Npb2+nsLCQjz/+GIfDweuvv45Wq6W+vp6vvvqK4uJi1q9fT3h4OM888wx2u52RI0fi7e3N8OHDB7yI+vr6WLFiBVFRUbS2tlJRUcH+/ftRKpWsWLGC9vZ2XnjhBYKDg/n8888lFesVK1bQ1tYmGQ3fffedi9CnKNkQGBjIo48+Sm5uLmvXriU8PPyEhghAcXExhw8fxm6388EHH9DZ2SlJHoiSDu+99x4eHh5ER0cPqhmjVqtRKpXExMQgCMfVw0NCQnjhhRcwGAy8+OKL6HQ6vvjiC4noUPTyiQgICCAwMJDExERCQkJYvnw5ALt27aK4uBhfX1+Cg4Px9/fnvffeo7KyUrrWYDBI989LL71EcXGxS942m42lS5dKoqWPPPIIubm5vPLKK3h7e/P999/z7bffMnr0aD744AN2797N8uXLiY2NpaamhiNHjvDOO+9Iz9HWrVt599138fX1RafTUV1dzYYNG9BoNHR2dp6wr91wQ8Ts2bMZN24cERERZGRkMG7cOOknIyMDDw8PsrKyuOGGG054UtMNN/rjrDJoHA4HjY2NqFQqMjMz8fb2libEW265RSKGS0lJYejQoXR2dvLDDz8gCAKhoaE88sgjhIWFERISwtSpU5k8eTJTp06lqqqKzs5OlEolSqUSLy8v9uzZg0qlYv78+Zx//vkD6rF9+3bi4uKIjo5Go9FQWVmJp6enRDoVHR3NsGHD8PHxobKykrq6OpRKJQ888ADJycmEhIQwatQoZsyYMeCIr8gpI5fLaWlpob6+nvDwcG699VYSExM5evQoUVFRjBw5Eq1Wi8ViITAwkIyMDKKiotDpdPj4+ODl5YWvr6+0JSKeOAgKCkKn0/H8888TFBQkuYBFWv20tDTS0tJob2/HbrfT0NCAv7+/JCPRfz/dGSKt/3333YdarZYE8hoaGujs7CQwMJCgoCBSU1OJiIggOjqatLQ02traCAoKIiwsjFGjRnH55Ze7bGcVFhbS09ODr68vjz32GHFxcXR1dfHSSy8RGBh4wu2QIUOGkJycTGNjI52dnQQEBBAcHMywYcMIDg6mvb2d3bt3k5KSQkRExKAvV09PT0mIVCaTkZaWxvDhw+nt7cXPz086rjpy5Eief/55qqurBxy3Fu8rPz8/VCqVxOcijo2Xl5ckAdFfQqGhoYH6+npJ4kEUlRTh4+MjyXxcdNFFdHV1oVariYqKYt68eRQXF6PRaAgICGDs2LF89dVXVFZWEhERwY033khGRobLczRy5EhGjBjBAw88gMlkIjw8nNraWt566y0CAgLcW09unBKi3MuJtpZEOY2QkBC3QePGaeOsMmhE+v3KykruuusufHx86OrqwtPTk6SkpEEl5gMCAqioqCAxMRG1Wo3JZJLSFBYWsm7dOknzRIwTEInmDAYDqampLjIKgGSwFBcX43A48PLyklbvzhICcNybo1KpqK+vJyQkhNjYWKmepzMx+Pj4UFVVRWVlJS0tLXR3d6NQKNBqtdhsNiIiIlx0UgaLL1EoFJhMJsl7YDQaufvuu8nOzua99947oRSAIAgkJiYyfPhwGhoauPPOO095lNlisbBo0SKuu+46li1bxtatWzl69ChJSUlS354slkj8vre3l8zMTCmNv78/lZWVxMbGEhISQnt7O/fccw8jR47kgw8+QKfTScJtzti0aRMlJSUkJiZK4+rcRpHk7tixYxI7cf/4FjEmQCS9c66782+1Ws2rr77Ktm3bqKqqGtA3zrFBDofDRTZClLQYTNbAx8eH+vp6QkNDiY2NRafTDZq3SJYXHx/vQlQWFxdHUVGRdH9mZWVRX19PWVkZ7e3ttLa2ujxHOp2O5ORknnnmGT744AO0Wi0PPfQQCQkJfPLJJycdfzfccMONXwpnlUHT2dlJTk4Oq1ev5vnnn6e5uZkbb7yRZ599lo8++oigoCBJeqCjowO1Ws3EiRPx9PTk4Ycfpra2FpPJREhICA0NDQQFBdHd3c2BAwck+YKMjAy2b9/OLbfcwsqVK3n55Zfx9/cf4EW56667qK+vZ/v27cyaNYvhw4czYsQINm7cSHt7OyqVCq1WK9Vp4cKFPPLII2zcuBFfX198fHzQarWDbnE4Q6PRcO211/LEE0+wYcMGoqOjeeihh/j222/ZvXs3ixcvxmKxEB4eTnNzM56enuj1eunvrq4uJk+ezNGjR9m/f78kDvndd98RGxvL7NmzaW1txcfHh5aWFoKCgqTTVQqFgvr6evbs2cPKlSv5z3/+Q2Nj40nr63A4+O677/D29ubiiy8mJiaGsrIyKisrsVqtlJSUEBYWRnNzMz4+PrS3t9PR0UFgYCAGg4HOzk42b95MW1sbl1xyCZ2dnSgUCslr9NBDD3H06FFJAiAuLo7Zs2dz+PBhnnzyyQGnh6KjoykoKJDIvyorKwkNDUWr1aJWqzEajdx11118+OGHbN68mbi4uAEGg+jF2rZtG3A87qmpqQmNRkN1dbUkPVBeXs6xY8dYsGDBAPFSnU6Hp6cnWq0WT09PJkyYwPLlywkNDcVisZCWloZWq+Xo0aN4enrS0dFBQ0MDGo0GlUrFggULeOSRR9i0adOgujaicvu3337L3/72N7q6uiRJhYULFxIdHc2WLVtQq9VceeWV3HzzzTz99NP897//JS4ujr/+9a/ScxQREcGRI0fo7u7mkksuQS6Xs2fPHlJSUpg+ffpJx98NN9xw45fCWcNDIwgCq1evJjIykhEjRkjxAiNGjJDo4AejsXeWGRAp8MVVskg1L8omKBQKiZJeXDE7bwH1r49zvoC0snfmaHEuy2q1olAoBsgQnCriXxAEl2sBaeIWtzbE/MTynNsvpncWrxS9FWLQaf/6itevWLGCUaNGkZycTFVVFf7+/gN0f07VL2LdRamHE5XV3t7OHXfcwfvvv4+3t7eL12wwOQXRmyGq8RYUFDBp0qQBzL+nKl8cG5EgrP+YCILgIlXgXOf+7RC9Lf0lAsR+EfMW83S+t5wlGQYbQ7Ed/b1kgiBw0003cffdd5Oeni7lIV4rjrco/SDm7Zyf2E/icyT2tXjPOfe1+4SKG2648WvgrDrlNGzYMNasWUNubi7Z2dlMnDhxAK38YIJcztT+/V/G4uQoThInorTvD9HQOZEcwmA4VT1PVlZ/Btr+dTtVfoOlP1GfiBCE46rZ69atIzw8nPHjx59SbXmwfjmZrIFzWc3NzYSFhdHS0iIFV59MQsB5rHx9fZk4ceKgMhenU/7JGH7PRCrjZCJ7/TWY+qc9lQjfierY3d1NYGAgTU1NZGRkDGoky+XyAbEKp5KUOFFfu+GGG278GjhrPDTgGhfipnT/3+B/1efO8Sj9PSFunBzu58INN9z4I+CsMmhOBucX+mBN/rle8s7l/NroH6h8ptf1p+//OeURfix+S/3rjP91vU5W3ul8J+K31o9uuOGGGz8Wfwg/sSAIHDp0CD8/P0JDQ9mzZw/nnnsupaWlAGRnZ/9sZWm1WkpLS5k5c+bPMllYrVZ27doFHKfG9/b2xmq1kp2dPahKMiARz4lBqPPmzSMwMPCUZYnbOkePHmXu3LnSFkhRUREWi4XRo0ef8NqioiLMZjNjxow54zaeLkRpgfDwcFJTU3+xcn4MDAYD27dvZ+HChWe0Xfhj4HA4OHr0KIBLfwvCcSXwffv2MW/ePBcFbEEQqKmpobCwEKVSia+vL3FxcSQkJPyidXXDDTfc+F/hrDrldDKIvCtyuZz33nsPk8lEaWkp33777YC0/Y8Kn+x/8aeiogKLxYJSqZR4R/pfdyKcLI3BYKCwsJCYmBheeuklKXizvLz8hNc5HA7efvtt4uPjSUhIOKWOkBhw29LSgslk4r333pOCPOE4mZ94zPdEZZ4szan64HS+LykpQRAEdu7cOUBa4GT9PNh3JxvLU6Xp/39jYyNdXV14eHiclM30dOp3qno4o6CggJ07dw6a5t1336Wvr88lj507d/Lss8+SmprKmDFjOHToED/88MMp++t0x04co1OdynPDDTfc+KVw1nhoBEGgsLCQuro6RowYwd69e5kxYwYdHR3I5XIsFgsymQw/Pz/JsAkKChpwzNjhcFBRUUFDQwMBAQGMHj2a8vJytFotcrmcCRMmUFBQgNVqpaamhlGjRqFWq7n77rt56KGHiI6OlrhccnJyCAoKory8nIULF1JSUoKPjw+BgYEUFBQwd+5cOjs7yc3NxcPDg+nTpw8IrvT19eWvf/0rnp6eeHl5ERUVRUZGBo2NjXz33XcYjUbGjh0rEdoBlJWVsXPnTkaPHk1XVxeBgYF4eHhIbLA1NTWMHz9ekmuwWq08/vjjZGRkcN555+FwONixYwc9PT0sXLgQi8WC2WzGZrORn5+PyWQiNTWV0NBQqc/ENA6HQ5JdiIqKwtfXV6L7T0pK4sCBA8ycOZN9+/YxZMgQYmNjyc/Pp6amhkmTJg2q41JYWMhDDz3E008/jVqtpqmpiY8//phRo0YxbNgwamtrKSwsJCQkhHHjxrmo+h47dozKykpsNhtpaWlERUWRm5tLY2Mj06ZNo7m5mc7OTinQeMyYMWi1WvLz8/H29mb8+PHSxN/X10diYiKtra0oFAqGDRvGfffdx6JFi5g8eTJ2u52uri6+/fZbJk2ahJeXF+Xl5YwePZr8/HwaGxuZPn26JNfgcDgoKSmhs7NTOpru5+dHWVkZixYtwsvLi6NHj9Ld3U1ERATDhw+npaWFsrIympub8fDwwGazUVhYSFVVFdnZ2YSGhg4I7jWZTDzzzDM89thjpKWlIZPJ+Nvf/kZra6vL89Pc3ExBQQFeXl60tbWRlZXF999/z9SpUxEEgbKyMsaPH8/u3buZPHkyJpOJpqYm5HI5SqWShx9+mKeeeorMzEz3VpYbbrjxP8dZ5aFRKBSsXr0atVot0cw3NDSgUCjYvn07e/bsOWUexcXF5OTkkJmZKVH7r169mrFjx7Jv3z4+//xzdu3axdatWyVJBF9fXxQKBcnJyRw7dowPPvgAh8PBG2+8gV6vp7q6mq1bt9LY2MjWrVvx8/Pjvffeo7u7m3feeYfg4GDeeuutQYUclUqlCzEeHJ983nzzTeLi4vD39+fhhx92odIPDQ3F29ubrKwsqqurycvLIzAwkNLSUj755BMcDofkRRL7zcvLi8TERLy9vTEYDERFRUn6Qfv372fTpk0cO3aMjRs3Eh4eTnd3t0t9Dhw4wKZNm+jo6GD16tXExMTQ2dnJ2rVrMRqNCIKAh4cHH374ITabTVLJ3rx5MzU1Nej1ep566qlBlaP9/f3x8vIiPj4eOM7zEhsby4svvohWq+Wzzz5Do9HwxBNPoNVqpetMJhMvvvgiw4cPZ8WKFTQ3N/PFF1/Q1dXFsWPHWLZsGUePHuWzzz6T8uvo6OCDDz4gJCREkhF4/fXXqa+vR6VSsWrVKhITE3nrrbfo6+tDqVSSkpKCyWRi2bJleHl5kZOTQ3l5OV1dXdjtdtauXetSpgiZTEZeXh6fffYZKSkpPPvss3h4eFBZWcmuXbtYt24dNTU1jBkzhueff57c3FxeffVVhg4dKnmDcnJyKCwsxGq18thjjw2q0t3Z2Ulzc7N0MsxgMPDf//6X3bt3u3hydDodL730EuHh4Xz99dcUFBQQEBDAihUrUCgUvP/++6hUKjZt2iRJQvj5+WEwGFCr1Xh5ef1sW1g6nY5du3YNKpzqhhtuuDEYzhqDRiaTMWTIEPz8/CgsLCQgIIBt27bR09NDQkICYWFhp5XPnj17CA8PJzQ0lCVLllBSUoKHhwd+fn5MnjyZvXv3EhgYSFxcHOnp6RgMBkmN2tfXl/DwcORyOSqVShJrTE5OlqQTAInbo7m5mbq6OlQqFQ8++CAJCQmn3J6C48dwnUUY6+vrXQwMLy8vlEol/v7+kuGiUCi48cYb2b59OykpKS5HgsUjuD4+PsjlcgIDA0lPTyc6Ohq9Xk94eDgymYzw8HC0Wi2vvvoqarVaqquHh4cUz+Pv749cLuexxx5DrVYzZswYXnrpJUpLS/H09JTKFTl/cnJy8PX1JTs7m8WLFw+6shelBcQtrSFDhpCRkYHJZKKoqIje3l7UajVPPPHEgFghnU5Hd3c3w4YNIyEhgT179uDn58esWbO46qqrCAoKIjo6mvT0dGw2G9XV1bS0tODt7c2//vUvkpKSCA4OZuTIkUydOpVFixZRXl6OXq+XuHREaQEvLy8UCgVXX30133zzDZWVlWRlZbF3716XMp37PTg4mPDwcJKTk/H39yc+Pp7Y2FhaWlrYuXMnkZGRBAcHk5KSwrp162hrayM0NFTyZH333Xf4+PgwdOhQ7rnnnkH7z8vLCw8PDwwGA3B8+7Wzs5Pa2lrJmyOTySQ9qeTkZCIiIoiIiCApKYn29nYXuQ2Ra2b06NHcf//96PV6PD09USgUEj/QT0VfXx8vvviii8HlhhtuuHEynDUGDRyfWBcsWMCyZcv429/+xrp164iOjpaIwAaLT+iPkJAQNm/eTFdXF2VlZYSGhlJRUYHNZqOvr8+FNK5/rIHFYhkga+CcTqFQYDQaMRgMWCwWvLy8qK+vR61WExERQVdXF83Nzej1+kHbJ5Ld+fr6IpfLJZ2m6OjoQb04zr8dDgcFBQVcc801vPrqqwPEEcVtOZG+37ne4o/RaOT+++8nPT19AMW9mMZkMnHjjTdy3nnnsWLFCry9vXn99dfZvXs31dXVwHHPSW9vL3a7HR8fH2pra0lISMBisdDe3s6xY8cGnNCy2+1YrdYBdQoICKC8vJzw8HACAwNd+s7Ly4v58+dTVVXFn//8Z8LDw/Hw8KClpYWEhAQMBsMAqQNfX19qamoICgoiNDSU7u5uqazu7m5efPFFhg4dilKplMj0LBaLS8zRqFGjaG1tRafTERAQ4FJmb2/vgDIH+xsgISGBwsJC6fNhw4bR2NhId3c3ZrMZi8WCWq2msrKShIQEqY/6Izg4mHnz5vHFF19IHg9xm+hkHEPOf8vlcqxWq4tERkJCAs8//zwff/wxOp1Oiu86HaMcjpMM5uTkUFxcTFFREbt27cJkMnHo0CFkMhlKpZLc3Fy++eYbTCYTVquVvLw86fmsq6vj8OHDbNiwgZ6eHgoKCti0aROdnZ2nXQc33HDj7MFZZdDIZDKmTp3K+eefz8iRI7nyyisZNmwYVqtVigFpaGiQPA16vR4PDw+XVeC8efNQq9U89thjdHd3M2nSJObMmcP69evR6XRcffXV0gTX2tpKeHi4dAJo8+bNkmSCWE5bW5sUv5OZmUl7ezvbt2+XxDP/9Kc/8cQTT7Bp0yaioqJYv349RUVFLu0SBIHq6mpiYmKoqqqSpBp27NjBnj17uOOOO1xOtLS3txMZGUlTUxMymYy+vj6qq6vp7OzkiiuuQC6Xc+jQIZeXvuh9Eo0DrVaLl5cXer0enU6Hh4cHXV1d7Nmzh8TERGbNmiVda7PZ0Ov1KBQKenp6yMnJITg4mAsuuEDSmVqwYAFJSUnMnj2bFStWSHE91157LUeOHJG2W8rLy/n3v//tstUQGBhIQkIC27ZtQyaTYTQaJZK92NhYRo8ezaOPPsrRo0eluB44Hhu0d+9evvrqK1566SV27tzJrbfeyoYNG3j55Zfx8/PDbDbj5eVFU1MTkZGR+Pv7M3/+fB577DF2796Nt7c3AQEBaLValEolERERfPnllyQkJFBUVMTkyZMlbaaIiAhaWlrw8vLi2muvZfz48chkMpcyRQ8WHJ/QDQYD3t7eNDc3ExERQWNjo2QcXXvttVitVrZt28bo0aNZtGgRF198MS+++CKNjY34+/tz6aWXcuzYMZ566imsVis6nY6IiAiam5ul8ZXL5ZLw6ttvvy0ZNgsWLHC5x9ra2oiKiqK5uRk/Pz8ptkej0aBUKsnMzOT999+Xttjy8vJob2/n0ksvJTY2lsTERHbt2nXaxoRcLicvL48dO3ZgtVp59913kcvlFBcXS8a/KPSak5PDpk2bqK+vp7y8nOeff55t27bxxhtvYDKZ2LRpE1VVVRw7doxnn33WbdC44cYfEGcdD41obAxGnQ8DZQfEz5zTCcJxGnrRzS4IgkQ73/86MS9nWYOTlSNS3zunFSnk4f8kCvrT8/dvA+Aiw9CfRn+wsgfrF+d2D0b97wyxnWI8TP8+cy5H7D+xnmIdxe+c6+ycZ29vL/n5+S6svuI1/VmBnespSlM4pzl06BAHDx7k8ssvl2KZZs6c6dJvg/VP/zFxTuPsXXGuv3O/ien6yxj0H6vB+m2wMXNuW/97sX//ifkORj4oXutwOCTvTP/n42Rj73xfOddNrMuZyHWIqK2t5dFHH+W2225jyZIlUjzY+PHj+fOf/8z777/PqlWr8PPz4+DBgyxYsICAgACUSqVEMfDwww9z7733cu655xISEoJcLmfMmDGnFEp1ww03zi6cdU+886ThrLvjrJ/j/Lu/MdA/TkD8TKlUDnqd+Fuk8z9VOWL8jHPdnCcXMY/+beqfj5jXYDpSJyp7sH5xvkZsc/9rnfPw8PAYYDgMVo6YxrnNzt8518E5Tx8fHyZMmDAgxse5bwdrz2DbJ3FxcXR1dbFixQqKi4sZP348gEu/DZZf/zHp3/7+Y92/3wCXv080VqdzX/ZvW/97sX//nWh8na8VT/n1r8upxl4ul0vtdq6bc9/0z/dUiImJITg4mK1bt3Ldddfx9ttvM2zYMOl7Z8NKLpdjMBgYMWIEnp6eLsal6B3MyspCpVK5j4+74cYfEGfNsW03zg78nKR0oaGhPPjgg9L/ZzLRuvG/gVwu5/LLL8doNDJ69Gh6e3sltXt/f3+ampowmUx4eHhw/fXX8+yzz3LkyBEuvvhidDodVqsVg8HA9ddfz5NPPkl+fj6XXnqpW1vKDTf+gDjrtpxE9N+e+bXrAb9sXQYbxhOV97+q05nglxqvH5vvmV53Jv3/U8o5G+G8XStuR4rbWc7bgh4eHlI8mhiULQbbA1JAsqen5x+6P91w44+Ks27LCaCnp4cvv/zyhKeF/lcQBIH6+no2bNjwiwYpiqSCu3fvxmw288UXXwx62kWEw+Fg7dq10jHeXxMi+d26det+9m0Cg8HAV1995XKk/XRRXl7O3r17z+ia3NxcXn75ZbZs2XLKvhUEgcrKStavX/+HD2Dtv6XnvJ0lbqeJW2peXl6SwaJQKFy2BkXySbcx44Ybf0ycdQaNGBy5evVqdDqd9Fn/Y6gnOpp8srQnu7Z/HZqbm2ltbcVgMLBq1apT5tv/+pOVM9i1jY2NrF+/HrlcTlhY2IDAU0EQpKOuAGFhYS5BpIPV7cfU72T9eaI+ttvtrFy5Ulqp/5gxcIbZbKa0tBS5XM6nn34qHb0e7HqbzUZRUdGAPHx9fSVOm5P1hTNCQ0M5cOAAU6ZMwdfX95TpzWYzq1evPqN2u+GGG264MTjOqo1mu93OkSNH6O3tlfSLOjo6OHr0KL29vcybNw+j0UhRURHd3d1MnjyZ4uJirFYrra2tzJ49WzqKOm/ePPr6+igpKaG3t5dp06ZhNBqpqqrCYrEwceJE6uvrKSgoIDAwkMmTJ0uBrGazmUcffZTx48czZcoUbDYbmzdvpre3l4suuoje3l6OHj2KXq9n3rx5LhpIJSUlNDQ00NzcTGhoKNOmTWP//v3IZDJMJhMjR44kLy8PhULBzJkz6erqorCwkJqaGgCJ58Zms2EymcjPz6erq4tx48axbNky/Pz8+Mtf/iJx4QiCwOHDh+np6ZEI1Xbv3k1QUBDV1dVceumlLpOz3W6noKCA8vJyRo4cSWBgIIcOHcJkMhETE0NHRwc2m42EhAQ8PDyko+OTJ0+mrKyM9vZ2GhsbufTSSyUSNn9/fxQKBYIg0NTURGlpKQaDgUmTJuHj40NeXh5ms5msrCw6Oztpa2tDqVSSlZXFhg0bmD59uiTjIAgCa9euZefOnTzyyCOoVCry8/Ml4Uh/f3+OHj1KT08Po0eP5siRI3zwwQc8/fTTEpOuIAj09vbS09OD3W4nLy8Pk8lEQkIC0dHRg957MpmMkpIS0tPT8fHxOaWXQGy3eM+0trZSVFSEwWBg/Pjx+Pr6UlBQgNlsZtSoUZJEg0qlIisry+2FcMMNN9zoh7PGQyMIAnv37iU/P59Ro0YBx1ffy5cvx8/Pjw0bNrBhwwbefvttUlJS6O7uJj8/ny+++ILc3FzCw8N577338PDw4MCBA6xZs4YtW7agUCjIy8tj//790jaNUqlEq9XyySefEBYWxnPPPSd5PuA4s623tzdDhgxBpVJhNBpJTk5mw4YNlJSUsGLFCnx8fNi8eTObN292aUNubi5r1qxhxIgRvP766xQUFLB8+XKam5tRqVS8++67hIaG8uabb3LkyBGWLVvGkCFDiIyMRBAEzGYzb775Jr29vbzzzjtER0cjCMfVxtVqNQkJCfj4+LB8+XI6OztZsWIFgiAwYsQIHn30UZqamnj99dfx9PSkoKCAffv2ufTz/v37OXz4MJ6envzrX/9Cq9Xy0ksvodFosFqtvPDCC/j5+VFfX8+aNWuYOHGi1J85OTmsX7+e2NjYQSdkk8nEf/7zHzIyMlCpVDz22GMcPnyY77//nrCwMHp6elixYgUhISGYTCbgeNxEf7mE4OBgQkJCCA8Px2634+XlhSAIrF69mtzcXOrr6+nr6+Ozzz4jKCiIwMBAF0NFEASKiopYt26dJJcQGxt70i1MQRD4/vvvGTdunMsR59OB2WzmP//5DykpKQQFBfHwww+Tk5NDfn4+wcHB6PV6li9fTkRExG9im9ANN9xw47eIs8agAfj222+Jj4/Hz89PYo0tKSnBz8+PO+64g+TkZKqqqggLC+Pqq69mypQpaDQaUlNTGTt2LLm5uajVaq699lpmz57NtGnT0Ov1dHd309PTw4QJE3jllVfIzc2lqqqK3t5eAgICePrpp9FoNFI9xKBFlUqFTCYjKCiIIUOGSIR1hYWFqNVqbrvtNsaPHy9NfHK5nJCQEGJjY8nKymLChAk0NjYSHBzM8OHDiY6OlkjPlixZgq+vLxUVFYSHhxMWFibJFqhUKgwGA3l5eURGRnL++eczb948qU7+/v6o1WqJqTUyMpLIyEiJQC4gIIDk5GRiYmKkbTsR+/btw9fXl9TUVB555BFCQ0MJCQnhnHPOYfjw4Wg0GiZMmIBer5dkAc455xz279+PRqMhPj6e6dOnuxABimhra6Ompobg4GDGjBlDeXk5UVFRHDlyhPfffx9fX1/GjBnD3//+d3p7e5HL5VxyySUEBQUN6HtRLkGhUJCWlsaQIUPo6upi6NChBAYG0tLSQnd3N0qlUkrvfOQ6NDQUmUyGRqPBZrPx+OOP4+3tfUIjxWq1UlpaSlZWFnBccqG5uRm73Y7D4XBhYO6Prq4uSktLCQkJISsri4aGBqKjo9m1axdr1qzB19eXESNGcNddd51SOd0NN9xw44+Ks8qgCQoK4ujRo5LyMxyfLCwWC2FhYRgMBrRaLUePHkWr1VJbWyutpBUKBWazWWJabWtr4+WXX0atVhMYGIjNZkMul/PGG2/www8/0N7eTmlpKYGBgXh7ew+Y+OVyucROLBKfCYKAUqlEp9PR19dHeHg47e3tA9ohXmO1WklNTZUmQj8/P44dOybpRBmNRrq7u2lqaqKvr8+Fgl+pVNLT08P+/ftpaWmhqqpKqpN4GkQulxMbG0tRUREOhwNvb29iYmJOGiMUHBxMYWEhkZGRWCwW+vr6BsS0wHHa/rKyMmw2GwaDgeHDh5/QcyF+FhAQgNVqpa2tDZPJRHJyMjabjaeeeorAwEA2bNhAWloazz77LB988AEGg4HS0lJ6e3td8hNPw4hbas4/a9asobm5mcjISInszmq1DqDsF9MbDAb+9re/MWvWLD766CM6OzupqqoakLatrQ273U5YWBgWi4UdO3Ygl8vZunUrn3zyCdu3b6epqWnQ+9bPz0+SsjCbzcTGxiKXy3n++ecB2L59OyNHjmTJkiW8++67bn0jN9xww41BcNYYNDKZjKuuuor29nbeffddUlJSEASB2267jddee43PP/+coUOHctddd/H666+zceNGScVZp9Mhk8m4++67+fDDD/nwww+Jjo5m5MiRbNq0ieDgYI4dO0ZVVRXl5eXMnz+f2bNnM336dJ544gny8/MlcUYR06ZNY8+ePRw7doyoqCgaGxslr8jtt9/OG2+8wWeffUZMTMyA7ZfKyko2bNjAuHHjiIqKIjg4mKamJsLDw7niiit4+umn2b17N8OGDeNvf/sbb7zxBiUlJcTGxlJbWyt5Vv7xj3/w0UcfsWbNGsLDw5k4cSL5+fmUlpYSERFBa2sr9957Lw0NDWzbto1LLrkEtVpNXFwcra2tKBQKF40igIsvvhiTycTTTz+NxWKhp6eHqKgompqaaGlpITo6moaGBrKzs5k3bx4bNmygr6+PK664QjJ+xO0iOG4MaLVaYmJisNlsPPTQQ2zdupWDBw/y97//nb6+Pr7//nuGDx/OpEmTyM3NpauriyuvvBKlUsm7777L7t27XfovJSWFvr4+Dh06JMkA6PV6/Pz8SEpK4ujRoxgMBoxGI4GBgfj4+HD06FHpervdTldXF35+fnR3d7N3715CQ0M5//zzKS4u5vnnn3c5RdbT08O3335LUlISu3bt4sMPP6Szs5Pg4GAEQSA8PByTyeSyLSnGC0VFRWGxWHj00UfZuXMn+/bt4+6775Y0jcaMGUN2djaHDx/GYDBw5ZVXSoKSbrjhhhtu/B/OKh4a55W4M1OqM/W++P+JCNz6pxV5MZx/9897MLZeQfg/iv9TleN8Imnr1q3s27ePRx999KR1dJZh6F9nZ/Sv72B1Eut6MlK7wU5NnQ61/In62jm/weoMSPn3/985z/z8fEJDQ4mKihrQnpP1vYeHx4CxPVGdnNvb09NDUVER48ePd5FmOBFee+01hg0bRkNDA5dccokUYH2m7XYeI3dAsBtuuOHGQJxVp5wGo3sHBrCGnoxFtP934sTZ//ep8hI5NU63HDg+gfX29uLj44PBYJBO7pzsWpGP40ToX4cTGRenq3tzoj4+EU7F2DpYXqeqs7PG0tChQweUcbp9f6IxHUwyQISvry/jxo0bIM0wGEwmEx0dHaSnp3POOefg6el50mtO1u4zGSM33HDDjT8izioPze8d/YfCvRL/faO3t5eKigqGDh3q3iZyw43/D0EQ6OzsJCgo6DdnpJvNZvr6+vD393e/f3+H+G3dTX9wiN6PM/WCuPHbhJ+fH6NGjXIbM2chxEnZLYJ5ZhCE46zm/ekgfisQBIH169fT0dHxa1fFjR8Bt0HjhhtuuHEGsNvtfP7557z77rvYbLZfuzq/K3R1dbFt2zZmzZrlEoNWWVlJW1vbr86G7eXlxfTp0/n888/dY/s7hNugccMNN9w4TQiCwI4dO9i1axe33XabS1zU/7oev/bkPxhOJZny9ddfM2XKFBceKpF/Kjc391fnWZLJZERFReHr60tubu5vso/dODHcBo0bbrjhxmnCbrfzwQcfcNVVV0nSHfB/J+t+zAQoCIILn5Pdbqe3t/eE21k2m02SYNHpdOj1eqlsg8GAXq//xbwLfX196HQ6enp6pDJ7e3vR6/USxUNVVdWgdTcYDBQVFZGZmemSl6enJzNnzmTGjBlYrVYpr58bImWEXq+nt7dX6u+enh6pPXDcqJkyZQqbN2/+SQaNIAgnJdR04+eH26Bxww033DhN2Gw2WltbiYqKkowZm81GYWEhy5Yt+1Gkh3a7nRdeeIGOjg7sdjtffvklq1atGsCvBMcnSVHbTafT8cQTT/D6669LvEjffPMNn332GUaj8ac1dBCIBJL33nsvn376qWTQfPzxx2zcuBGLxYJMJsNsNnPw4MEBE7nIai7GlIl5rVmzRqJFWLVqlZTXL4H6+npuuukmtm7dKhkcb731Frt373YxoqKiomhpaRkgNTIY2ehgsNlsbN++nR9++IFvv/3WvX31P4LboHHDDTfcOEP0P50TEBBAfn7+j/IseHh4cPvttxMUFERDQwM7d+7k6quvZtKkSQPS9vb2Ul5eTmJiIqGhoVxwwQWUl5fj4eFBc3MzcXFxXH/99fj7+//otp0IMpmMmJgYZs+eTWVlJXK5nNraWrKzs7n88sslj1VaWhq5ubkuxp0ovJucnOyS15w5c06Y1y9R/yFDhjBp0iRqa2uRyWRUVlZy3nnnsWDBApRKpZTW09MTtVo9KLv3gQMHKCgoGMAu7oy8vDzq6uqYOHEiJSUllJWV/eztcWMg3AaNG2644cZPgEKhICws7KTcRydDe3s7jz/+OI2NjRw+fJjq6mqKi4sHjc+pqqoiODhYOgk5ZswY2tvbKSgooKKigrFjx56S+wl+fAyOTCZjxowZFBQUUFJSQlNTE6NGjXIx8Dw8PCSBWmc0NDQQEREhebZkMhnTp0+nsLDwhHn9EvWfP38+u3btorS0FKPRyNChQwc9VRoaGjqoQTNy5Eh6e3v57LPPOHTokLRd6Iy8vDzp6Lefnx9FRUVnXFc3zhxug8aNXwyCIGC1WjGbzb/oPrIgCBiNxgGq22648XtAUFAQHR0dWK1WMjMziYyMJDs7e9CJXavVotFopAnY19eX8ePHs27dOsaOHYtSqcRgMFBfXy89c319fXR3d0t5dHd3s379erq6ulzy7uvro7i4mKKiIpef/jp1ISEhJCYmsmvXLrKzs/Hw8KCrq4uWlhYASZC3vzHQ29s7wPMSEhIiSYYMlpd4najXJm57rVu3zkVCRUzXv+7FxcUDto0SEhLw9vamoKCAkSNHIpfLaWlpcekPmUyGr68ver3e5VqZTIa3tzcTJ07ksssuQy6X89lnn5GTk+Py/jEajZKBK5fLf5EtQDcGwm3QuHHacDgc1NXVUVlZSWVl5YAXXX+I+8jffPPNgJfPzwVBEGhpaeGLL77gyJEjP8lw6ujokNrW2NjoDuZz40fjTDwIZ8ICLcq6OMNqtTJz5kxUKhUWi4Vt27a5TLAdHR1UVFRI6b28vKipqaGnp2dA3iKxnPhjNpsHXSiInhqlUonJZOLLL78kLy9ParMo19H/mhNhxowZeHp6uuQlltvQ0ODi7fHx8aGkpGTAiSiHwzFo/QeTMvHy8mLKlCl4eHjQ29vLqlWrBojOnooLTC6Xo1QqEQRhQMxPcHCwVLbZbCYwMPCkebnx8+Cskj5w4+QY7AV7qofW4XBgNBrx9fWVTnKILxrnl7ZzPuJndrudcePGoVQqXU6AGAwGfH19z4g80Gaz0dfXh5+f34ByFAoFf/rTn+ju7j6jF1L/fMT29W+bmJcgCOj1ejeL6P/HYGP/R4TD4aCjo4Oenh7a29vx9vbm0KFDNDU1sWjRolP2j9VqxWg0YjQasVqt9Pb2YrFYXI42i4iMjKS2tlbq+5aWFoqKirjmmmukNOXl5cyePRuz2UxTUxPNzc00NzcTFBREaGioJMrbH97e3owaNeqUba2traWxsZGgoCDp89raWqZOnQr8H+lgdna2y7U+Pj4uC5v+eYltEvMyGAy0tLRQV1eHIAgolUoiIyPx8fFxiXcR4e/vz+jRo09af7vdTllZGUajER8fH6ke7e3tA/TgjEbjgFgk0XjJz8+nrKyMpKQkl/ghEePHj2fjxo2YzWa6u7s577zzTlovN34euD00fyDY7Xa2b9/Oiy++SE5Ozikj7wVB4NChQ7S1tUkq0+Xl5SQnJ5OUlMQ333zDN99846I8DcfVp1euXMmKFSsAUKvVVFdX89Zbb1FfX09ZWRkVFRVn5AHp7OzkwIEDLnUrLi7mueee4/DhwygUCjQaDZ988gnr1q0bUKeTwWAwsHr1aj7++GMCAwNJTU2lp6eHN998k2PHjrmk3bZt2x/2xIIgCNTX12MymSgpKSEnJ4eDBw9is9moqan5xU6m/FgIgkBZWRnr16/nhx9++MXq53A46Orq4uabb0an0+FwOIiPjz9tQ6+vr4+bb74ZlUqFj48PN998MwaDYdDnIyUlhY6ODsng7ujoYPHixdL3oqcnIiICu90ueSksFgt9fX0uBrvz7zNpq06n4/bbb5f6UxAE1Go1arVaWhQYDAbi4uJcro2KiqKpqUkq81R5Oddf9Lz0X3Ccaf1tNhtms5nFixdL20AWi4WoqKgBsUcdHR2Eh4cPyOPQoUPI5XIuu+wyJk6ciI+Pz4CxTkpK4txzzyU3N5dFixYRExNzRvX8MRCPoIueoZaWlgFe9O7ubrRaLRaLZcAC8GyA26D5A8HDwwOlUklzczOTJk06ZfBgZ2cnNTU1xMfHk56eTktLC93d3cjlcnQ6HRqNhnPPPXfAakmtVpORkcHu3bvx8/PD4XDQ2dnJ7NmziYmJITMzkz179pzRBCPyczgjPT2d9vZ2urq6kMvl6PV6AgMDmTVr1qAruBPB19eXzMxMdu7cKb1I29vbOffcc4mPj3dJeyaG0tkEQRA4cuQI9fX1dHZ2snr1asaMGcO2bds4duwYSqWS7du3/+bimNRqNcuXLyc6Ovq0gmVPBZvNhsPhcPFwKBQKhg8fzsyZMxkxYgRKpRJvb2/S09NPK0/xnh0yZAjp6enMmjXLJU6mf3uSk5OlUzrDhg3j3HPPZciQIVL69PR0urq68PX1ZdiwYaSlpZGcnMzw4cMJCgqivb0dpVLJsWPHzvhUlkKhYOTIkdKzIW6XpaSkoNfrpZNDw4cPd/EwyWQyhg4dSnV19WnnFRAQwPDhwxkyZAhDhgxh+PDh+Pn5odVqUavVA7aITgdeXl5kZ2czc+ZMIiMjAVAqlcTExLh4j2w2GzqdTkrjjEmTJjF69Gi8vLxOaLSK7Z0wYQJpaWm/uBdTNGD27duHTCajvb2df/7zn+Tl5UlprFYrS5cuZevWrcjlco4cOXLWba27DZo/GA4dOsTEiRNRKBSnfMgOHz5MQkICcrkcuVzOJZdcwsaNG9Hr9RQUFDBz5sxBt45kMhmZmZk4HA4qKirIy8sjISGB5ORkad/Zz8+Pmpoa6RpxtSVuaTn/nGirrH+d8vLymDlzJn5+fmf0ApHJZKSnp+Pt7U1xcTGFhYVERkaSlpaGXC7/w2+pwHHl8LVr1zJq1Cg8PDzIz8+npKQEmUyGv78/0dHRVFRU/KaOp8pkMlpaWoiPjycyMnLQOBXnlf5gP/3TVlVVER4eTnBw8EnL9vX1JSUl5We/d2QyGZMnTz6hYa1QKFi4cCEZGRlSe4ODgxkyZIiUJiQkhBtuuIGpU6f+6JNZzvDy8mLRokWkpKRIfTZu3LgBbU9NTaW5ufmkiwLnvMTro6OjJQ+HyOR7yy23kJ2d/bP0r7+/P4sWLSIuLk7Kr6mpieDgYNRqtUva36rWnsPhYOXKlYwcORJPT09CQkKIiIhw8cQdOXIELy8vBEFAoVAwevRoPv7447PK4+yOofkDwW63U1hYyEUXXQT8n7vZZrOxZ88elEolXl5eREREEBMTQ1VVFQsWLACOP8ijRo2ivb2djRs3smDBgpOKLnp5eTFnzhzefvttFi9e7OKCl8lkREREUF1d7bKKzc3NpbS0dEBe2dnZg/JqyGQyRowYIZ3aWLhw4YC4A5E8q7/nwMPDAw8PD6lOSqWS+fPn8/bbb3PHHXeQnJz8m3tp/ZqoqanBaDSiUqlQqVRcfvnl3Hjjjdx6661oNBoAYmJi2LVrF+np6b+JvhMEgYMHD0oTn2ikOE9IOp2OPXv2DPBUBAcHM2nSJGnCF+OnVq1axd13331KyYOfwxt0IiiVShcDpT/69723t7fL6aJfwshyRmpq6qBl+Pn5kZGRQUlJCVlZWYOmOZFX6lRpfgr65ycIArt372bOnDm/ifv4dNDW1kZFRYX0LDpDEAQaGxvp6+sjKSlJ2mrz9/ensbGRuro6iR/o9w63QfMHQmdnJyaTiaioKOkm9/f3x8/Pj/3793PppZfi5+fH0qVLeeqpp+jr63PZulGpVISHh5OVlYVarT7pwy6TyUhMTKSpqWnQCU6lUtHa2uryWXp6OgkJCQPy8vHxcTl22j+fk9XJ4XCQk5NDW1uby+cxMTEuxGUymYyEhAQKCwsZNmzY7+ZF9r9CT0+PtOJvb2+npqaGjz76iEceeYSxY8cyevRofH19aW9v/5Vr+n8QBIGjR49y2223AcdZYo8dO8aUKVOkNGq1msmTJw/wxigUChePjiAIrFixghkzZjB27Nhf/f74tcs/EU71TrjgggtYvXo16enpv0kVenHrRgxq/q32c390dnYiCMKgHjebzcbnn3/OsGHDKC0tpa+vj7a2NkJCQvDy8qK1tdVt0Ljx+4LI1BkfH4+Xlxc9PT3k5ORw6aWXYrFYaGlpwWq1UlxczPTp01EoFPj7+0v7yoIgoNPpMJlMJCYmDjjVVFhYSFJSknQaSnRxTpo0adCXgsFgGHCUsaqqiqqqqgFps7KypBMJ/duk1+sxGAwkJiYO2m65XM6MGTNO2TcAR44cYeLEib+bl9j/EmFhYRIzaltbG3q9nszMTC688ELq6+sZNWoUer1+QCDorwm9Xk9nZyfx8fGYTCa++OILwsLCsNlskqGu0+nYvn37AA9NSEgIM2fOlCYImUzGZZddxksvvcSQIUPcHrwfiZCQECZPnszOnTuZM2fOaR9X/1/BbDazdetW/vSnP/2iXrafG84B2c5GjfhuS01Npauri+7ubinAGo577X8JVulfC7+fEXPjJ6GpqYkjR44QFRXF5s2bKSsrIzMzE4VCQWNjIxqNhoiICOLj46W4mPT0dOrq6khMTMRkMrFr1y6Sk5NpampysegdDoekVyIe+ywoKMBkMmGxWLDb7S4vB0EQ0Gq1TJ48WfpMJpMxfPhwhg8fPmj9nYm2RIh1SkpKoqmpiZSUlAFpTnfSKS4ulk6oOE94bhxHbGws0dHR6PV6UlJSyMzMZPPmzfj6+jJ9+nQEQaCpqYmLLrroNzHR9/b2smXLFslrFB8fj1wuZ+7cuS73YlBQEJdeeumgnCnO7ZDJZMTFxXHxxRfz/PPP8/rrr/+uJrzfCmQyGaNHjx70ef6tYMGCBRIb8+8FERERREZGotfrCQoKQqfTYTQaJS+4GDoQGBiI0WgkNjYWk8mEj4/PCReDv0fIhLMpxNmNE+JEw2yz2fjiiy/o6+vj2muvdVkxGQwG1q1bx+WXXz7AldnfQ1NbW4uXlxeRkZGn5Lvp7e1l/fr1XHbZZacdlNjS0kJubi5z5849aZt+zEvodPMRBIFPP/2USy+99A9n8IhGaHl5uURIZrPZpEm9tLSUrq6u34yHy+FwSKtQkfxs2bJlzJ07l/T09B89fnq9nosuuohVq1YRERHxc1bZDTd+NARBoLKykoaGBqZOnSrx5chkMlQqlYuQKhz3XO/fv1/iHvotPLM/B9xLjD8ITnTDenh4cN555w16msfHx4dx48ZRXV3tcix0MHh7exMaGnrSsuD4g1daWsrMmTPPyN2sUCjw9fV1+eznegjPJJ+AgICz5uE/E4inS3x9fXE4HCgUCskoEASB8PDw/8nx1NOFXC532aYUBIFLL70Uf3//n+RZ8fT0RC6XD1DV/jGEjqeDwQKZfyzEwPjf2jaPGz8dMpmMlJQUVCoVfX19+Pr6Dnqfi5/19fURFxdHTEzMb+aZ/Tng9tC4cVKI/C/OJ4J+jfzEI9y/pptfPDF1Okfe3Tg70dfXx8KFC1mxYoUUwC4IAnl5eRiNRmQyGePGjfvJx6HF13JnZyevv/46t956KyEhIT8pT5F/ZNasWT8pHzd+uzhd9u6zleXbbaq7cVLIZLKfdQL/sfnJ5fJfPWZBJpOhVCrPupeAGz8NjY2N7Nixg3HjxlFWVuZCZna6EAM6RY9MW1sbu3fvRq1WU1xcLLHkOnPkDPa/M3dT/+/HjRtHdnb2oNefjPPJjd8PTteT91vk0vk54N5ycsMNN9z4CSgsLMTT0xMPDw80Gg0//PDDGcUl2O12vv32W6xWK7W1tVx88cWsXr2a4uJiEhMTEQRBCuT/61//SmhoKPv376e0tJTp06djsVjYt28fPT09DB06lMrKSq6++mrefPNN/va3v9He3k5fXx+tra14eXkxdepUdu/eLTENOxwOCgoKKCkp4S9/+QuRkZFn5WTnxtkPt4fGDTfccOMnoKenR/IeKhQK9Hr9GV1fVlbGtm3bmDdvHqGhobzyyiskJCSQlpZGdHQ0giAwceJERo4cyaZNm/j444/p6enBz8+P1atX43A4yM3N5eabb+acc84hLy8PtVqNVquVpEFSUlIwmUyUlZVRX1/P1q1bycjIQKFQ8NZbb+Ht7Y1Op2Pr1q2/RBe54cb/BG4PjRtuuOHGT4Cfnx/Nzc3AcaFDURFeDOY9FRobGzEYDMjlcjIyMli3bp2kVC1u0QYEBODv7y+pU0+bNo34+HgEQaC6uprAwEBJdygjI4MNGzaQmJjIl19+yZgxYwgLC5PYgpOSkkhOTuaee+7hvvvuQ6vVkpWVNahcgRtu/J7g9tC44YYbbpwhnGNNhg0bhtFoxG6309LSwtixYykuLmbPnj2nFZOSnp5Oa2sr7e3t1NfXM3PmTLy9vens7KSjowOr1YrNZpN+x8fHs3r1aqqrqzl06BAWiwWLxSKdYrrwwgtZs2YNN9xwA6WlpSQkJCCTyTCbzVitVurr65kzZw433HADhw8fRi6X880331BeXk5JSckv1mduuPFLw23QuOGGG26cJjw8PPD396erq0syVmJiYjj33HM5cOAAGRkZjBo1Ci8vL2pra08rz5iYGB577DF++OEH/P39ueqqqxg/fjwJCQk0NzezYMECjEYjQUFBZGVlccMNN5CUlERubq4kFjp58mRJ7Xro0KH861//Ii4ujocffpihQ4dis9nQaDSkpaXh5eVFfX09QUFBXHnllTz55JP09PTQ0NDAiBEj3F4aN363cB/bdsMNN9w4TQiCwCeffEJxcTFLliyRYmf689D09PRQW1t72rpg/a8/Ga/NqThvnI/kDnY892SvfLcx48bvGW6Dxg033HDjDGA2m1m2bBmBgYFcd911g7IO9/X1IZfLT6nK7YYbbvx8cBs0brjhhhtnAJEcsqqqiqSkpD+cDIYbbvxWcdbE0IjaFSaTSRJFFEmjTgWz2YzRaJTSCoJAX18fZrP5J9XHbDZjNpux2Wz09PT8aNIqh8OBXq//WUivRLZbkZvip+YpCAIGg2GAWvHpXqvX66Vgxl8SgiBgMpmwWq2YTKYB1PX9YbFYMBqNZ1zOmY6VODkaDIafjdRssH4VBIGenp5TjtPJ6uNwOOjq6vpJ9/LZAPHkUVpamtuYccON3xDOGoMG4NixY1xzzTV8+OGHvPPOO7z44ouYTKZTXnfw4EG+/vpr6X+LxcJTTz3F7t27f3RdrFYrzz33HNu2baO2tpb33nvvjCcBm82G3W6nq6uLV155RRIW+ykQBIG1a9fy7bff8tVXX50wjWgQngqtra3ceuutaLXaM66L2Wxm6dKlGAyGM772TNHb28u9995LaWkp27Zt47vvvhuQxrndhw4dOmH/nAydnZ28+uqrZ2Tgfffddzz44IM/m2FnNpt5+eWXJaNEbNPbb799WuO0b98+7rvvPpc2CILA7t272b17N2+//faPMmDdcMMNN35JnDUGjUwmIyEhgZ6eHiZMmMCNN97I7t27OXLkiOQt6erqwm63IwgCRqOR3t5eHA4H48aN4/zzzweOGxEmkwlvb29MJhOCINDb2yutuh0OBz09PS4eHWeIK1yj0YiPjw8mk4mEhASuv/566eikyWSSjnl2dXVJnhJxdS96PFavXk1paSmBgYHceeedKBQKaaUtrsAdDofklRps5exwONDpdPT29iIIAh0dHXzzzTece+65XHzxxS511uv19PX10d3dzWuvvSZNiAaDAZ1OJ1Gj2+12uru76evrIyQkRPL6GAwGl5W92O99fX0ufd/X10dvby9yuZy77roLHx8fqRyxXWI7+7dpsDxFZWXRA9F/rK1WK3a7HblcjtVqZc6cOUyePFmqd3d3NxaLhd7eXl5//XX0ej3Z2dlccMEFLu0Vx9xut2MymaT+cq5fcHAwt99+u6RGrdfrB3j6xHuqp6cHgMTERLq7u4HjxrRer8dqtUp91dXVJf3vnIdz3cW0BoMBmUzG3XffjY+PD0eOHOGrr77Cbrdzyy23EBERIaXt7u52eR50Oh2CILjUx7nPP/vsM1JTU7ntttsQBMHl3rXZbPT19f1snkQ33HDDjTPFWUesJwgCx44do62tDW9vb+Lj42ltbWXPnj00Nzcjl8tZsGABe/fuxc/Pj4yMDH744Qeampq44447WLVqFcnJyRQVFTFs2DCOHj2KVqtlz549LFy4UJpYtFotF198sURWJcJms7Fq1SqSkpIoLCwkLi6OnJwc1q9fzz//+U8eeughMjIyUKlUxMbGArBz506WLFlCTk4Ofn5+bNq0ib/85S+sWbMGLy8v+vr6ePPNN3n11VfZsWMHDoeD5uZm/Pz8GDNmDEuWLGHu3Lns2LGDZ555hqioKOA4pfqaNWsICQkhPz+frKws/Pz80Gq1lJaWMn78eOkkxJdffklcXBw1NTXEx8ezceNG5syZQ1dXF01NTdhsNnQ6Hddddx3r1q0jIiKCb775hoceegiZTEZfXx+rV69mwoQJ0skOvV7P/fffz5AhQygvLyczM5Nx48bx2muvkZiYyIgRI9iwYQNPP/009fX1aLVaqqqqSE9PJyAggO7ubrZt28YNN9zA8OHDkclk6HQ6/vnPfzJ06FCKi4sZPXo08fHxfPXVV2g0Gi644AK0Wi0tLS04HA7+8pe/8MknnzB06FCqqqqw2Wy88847+Pv7c/HFF/PVV18RHh7Ozp07ueSSS1i3bh1TpkyhoqKChoYG7rzzTj766CNSU1PZtWsXF154IVVVVezcuZMhQ4ZQX1/P008/jVKpRBAECgsLee2111i2bBmff/45GRkZ1NXVcdFFF0l9vXv3blpaWqivr2f+/PmSKrTFYuGTTz4hIyMDo9FIYmIiO3bsIDY2lgMHDnDvvfdKpG0Gg0Gq+44dO7jyyiv5z3/+w9ChQwkLC2Pv3r08++yzbNmyhY6ODkaNGsWyZcu45ZZb8PHx4eDBg5jNZiwWC+PHj+fAgQMApKamkpycPOC5amlp4dixYxQXFxMQEMDXX3/N0KFDpfFZt24dHR0d2Gw2Hn300QHK6G644YYbvzTOGg+NCHFl397eDkBHRwdffvklTU1NhISE0NTURFtbG+vXrycyMpKIiAiio6Pp6OggPz+furo6pkyZQmZmJna7nffffx+LxUJ0dDRVVVVUV1ezbds2xo0bN+j+eVFREVVVVUyZMoWsrCwA4uPjaWtrIygoCLPZzMKFC5k+fTrbtm1DoVDg5eVFbm4u+/fvZ9q0afz1r38lKiqKiIgIRo4cSVJSkhS7sHLlSmbMmMFFF13ERx99hFKpxGQycckllxAREUF1dbVUF61Wy+bNm5kxYwYLFy7kjTfeID4+npCQELKzs6X6C4JAbm4uP/zwAxMmTCAqKorw8HCSk5NZvnw5U6ZM4cILL2TTpk1s3LiRlpYWpk2bxnXXXYdKpcJqtbJy5UrmzZvH8OHDkcuP31b+/v4IgsCMGTN47LHH2LBhA2q1GqPRyP3338/8+fMxGAyYzWY+/vhjpk2bxjXXXEN8fDwffvghdrudqKgoqqqqpDap1WocDgezZ8/mscceY926dfj7++NwOFiyZAn5+flotVpCQkJoaWlh8+bNeHh4cM4555CcnIxCoSA0NJTu7m6+/fZbVCoVM2fO5LLLLiM6Oprw8HAyMzOJioqis7OT3NxcKioqmDp1Kueccw5vvvkmGo0GT09PbrzxRmpqaly2zKKjo+ns7MRut/P9999TXFzM+PHjXe6R0NBQ0tLS6O3tpbCwUPrcarWyY8cOOjo6yMzM5MMPP2TIkCHMnj2bjo4OcnJypLQ7duzA09OTmTNn8qc//Yno6Gh6enpYvHgxV111leRNio2NZciQIaSkpEiG50cffcTYsWO5+OKLmThxImq1mhEjRgDwww8/DLinZTIZMTExhIWFkZ2dzffff4/ZbGbq1KkkJyezevVq/P39iYqK4oknnpAMNDfccMON/yXOOoNGLpeTlZXFFVdcQWpqKp9++iktLS2kpKRw6aWX8o9//IPk5GSuuuoqnnjiCY4ePSpxL7S0tEiTk1wux263097ezoQJE7j11ls5//zzWbhwIVFRUTzwwAN0dnYOUKhtbW2lt7dXygNcuR08PDzw9PSkr68PhULB3LlzpUmgsbERu91OUlISXl5ewP/Rp8tkMmw2Gx0dHZjNZnx9ffH19cVut+Ph4YFcLsfDw8MltsFsNtPe3o7NZiMoKEjasuoPmUzGzTffTEtLC0899ZS0ReJwOOjo6MBoNOLl5UVgYCAtLS00NzcjCAJpaWnI5XLkcjkdHR1s27ZtQP5yuRylUklAQABBQUHSUValUomHhwcymUxiLxW3sIKCguju7mby5Mncc889TJ061aWug+Xp5eWFp6cnbW1tJCUlcckll/DPf/4TrVaL2WyWrnMej5aWFjo6OpDL5aSnp0v97ZzGaDTS0dGBIAiEh4dL/atQKPDw8Bhwjfgjl8u5++67OXr0KC+//LJL/FNNTQ21tbVoNBqXuBkvLy/++c9/snbtWlauXElnZ6dElhYZGekSyNza2irVPS0tTVIC9/T0PKGaueghqq+vp7e3F19fX+Lj42lqaqKkpETaPjwVent76erqAiAqKkrqX5VKhaenp5vLxA033PhVcNZsOQmCQHd3N729vZSVldHd3U1ZWRl/+9vfkMvlPP/88zQ1NREQEEBUVBRBQUHceOONdHZ2AscF5oYNG8bbb7/Npk2bqKmpwdvbm2HDhvHYY48xZcoUkpOTaW5u5vzzz6e3t5fOzk5efPFFbr75ZslNn5mZyeuvv86GDRuoqqrC4XCQnJwsxYeIk0FSUhJVVVUsXbqU0NBQpkyZgsVi4YknnmDIkCHMnTsXtVrNwYMHyc7Opre3F4VCwbRp09i+fTtDhgxh6tSpUqyPwWCgp6dHipWRyWTS6nzfvn0AXHDBBZjNZik2xcvLC5lMhsPh4LvvvmPx4sW8/vrrKBQKDAYDpaWlLFy4kC1btjBz5kyGDh3KBRdcwE033cQLL7xATEwMs2bNwmw2c+utt7Js2TKioqKYNWuWRDhms9morKyks7OTGTNm4HA46O3txWg0IpPJ6O3txWq1MmbMGO6//36mTZvGqFGjiIuLY8mSJYwZM4bhw4cTFBQkjbXNZqOiooLGxkZmzZolxZ2YzWamTJnCs88+S2trK/7+/kyePJmnnnqK9PR0tFotx44do6urC71ez7nnnstdd90leeCmTp2KxWLh6NGjdHR00NPTw/Dhw4HjisrFxcVcfvnlGAwGjEajFEtlMpkIDAyU7iMx5mrfvn3cc889vP322y6Gi7iN1Nrail6vJzk5GaPRSHt7O2VlZTzwwAOsXbuWSy+9lG+++YYRI0bgcDg455xzpDymTJnCbbfdhs1mIyoqijFjxkhxOd7e3lIdgoKCOHz4MNXV1dK4z5o1i3/9618sXLiQ1NRUDh48SHt7O2FhYdTW1tLU1ITRaKSvrw8/Pz/phJher6erq4tp06axa9cuyWN50UUX8d1332G327HZbNLYu+GGG278L3HW8NAIgkBLSwt1dXUolUpkMhlhYWFSEGRRURE9PT2MGjUKq9VKY2MjCoWCuLg4GhoaMBgMpKWl0dzcTF1dHaGhoWg0GtRqNYcPH8bf35+hQ4fS0tKCXq/Hx8eHqKgovvrqK9LT08nMzJTqUVdXx7FjxwgPDycoKAiLxUJbWxtRUVE0NjYSGhpKbGwsTU1NlJaWkp6eTmRkJB0dHRQUFJCamkpUVJQUCxIZGUljYyNJSUn4+vpSWlqKp6cnSUlJGI1GqquriY+PR6vV4u/vT3x8vLQaNxgMVFRU4OvrS2JiIu3t7TQ2NkpbT2K6mpoaLBYLAQEBhIeHk5eXh0ajITIykvLycgRBICkpCW9vbxobG6msrCQzMxNPT08qKiqIiIiQto9SU1Px9PREEAQWL17MvHnzSElJITU1lY6ODhobG0lOTkYmk1FVVUV0dDQBAQEcOXKEoKAg0tPTMRqNHD58mLCwMMkTBMe9RjfffDMXXXQRCQkJDBkyBK1WS2dnJ6mpqXh7e1NcXIxer2fkyJF4e3tTXl5OZ2cnQUFBhIaG0t7ejt1uZ8iQIdTV1aHVahk1ahR+fn4UFBTg6+sr9V16ejoGg4Fjx44RHBxMbGwsDQ0N6HQ6oqKiqK2tJSkpSTJoWltbqa+vJykpiY6ODux2OxqNxqWvW1paqKqqkuK7QkJCaG9vJzExkZaWFgAiIyNRq9XU1NTQ09NDXFwcQUFBkvdDEASqqqqkupvNZo4dO0Z8fDxKpVLqV39/f3Jzc0lJSaGxsZGgoCCio6PJy8tDLpeTmZmJwWCgoKCA5ORk6uvrCQ8Pp6Ojg9TUVNRqtXQMvLKykpCQEMkY02q1hIeHEx4eLt0jQ4YMcZPJueGGG78KzhqD5teAw+Ggvr6e6Oho96p0EJjNZm688Ub++te/MnXqVMko+SkwmUzccMMN3H777UyaNMm9veGGG2644QbgNmjc+AVhMpmoqKjA39+fuLi4n8WgMRqNVFZWEhAQQGxs7M+SpxtuuOGGG79/uA0aN9xw41eByOkjQgyyFuO6xLgjMcha9MaJn5+pMSuWJ5PJpLJOllasw8/FBiwyl9tsNmlb/NeC3W7Hbrf/6vX4sRC5kcQ4QHEas1qtv7s2ifeFOB4/Z74WiwWlUvmHWfj9MVo5CMQXS09Pj0Tp/r+g4P854Fz3E5H79fb20tzcjNVq/UnlGAwGLBYLZrNZOr31YyDy2Pyv7GfxYTYYDNL4nqxskSjuTO8BZ1LCn3r/iOPW2tr6k/ISA+R/LJuvSMDY0NCAVqv9SffQyWC32/nyyy+56aab+OKLL3j77bf58ssvMZlMtLe3c8899/DSSy/xzjvv8P7770tEl3v37iUnJ+eM7yW9Xs/DDz9Mbm7uKdNaLBbeeOMN/vvf/w74ThzzMylfPA25bds2XnrppTOp9i+C4uJiPvjgA4AzbstvASaTiWeffVaiJ3A4HJSWlvKPf/zjZ2FUP1P070PxHX2qfhWJQT/77DNpPH4uOBwOXnzxRdra2n7WfH/LOKsMGtFqP118//33PPjgg5hMJl599dUfpd3za+GHH37ggQceGDDxCYJAQUEBu3fvZsuWLezatetHlyHyxeTn51NQUDDoy/1kEFl6HQ4HDQ0NP0r+4cdCEAT++9//8uqrr2IwGHjttdcGvTccDgcWiwWr1cprr72GXq8/47IqKipYvHjxT5ZwaG5u5ssvv+T999//UVISYlvsdjtvvfWWxMX0Y9DQ0MBtt91GY2PjL2boKxQKKcD44osv5pprriEvL49///vfBAcH09vbS0pKCn/5y19Yv34927dvx263s3HjRt55551BJy6RBVkkwBSZtG02G/7+/gADGJ8dDgdWqxWbzYbNZsNiseDp6UlQUJB0ZF9kEHc4HBiNRj788EOJJVnUbBNX2s7lw/Fx+fLLL6mvr0ej0aDVaqV7zvkaZ7kRZ29OX1+fNDk6t0dsg9VqddGuE+8Bq9UqlSH+LdYzPT2dK664ApvNxurVq6V2imWLmmdinmK54sTt/L/YZw6HA7PZLHm2bDabVF7/PnLuT3HB5Fx3sR5ms1lqn1iGxWJBpVJx11134e3tzb59+zh48CAajYampiaX/Pr3pVg/57EXyxX7UKy7c3v7p3ceB4vFwkcffSSxZIsLgk8//VQaJ+f2ixBJWLu7uwkICKC5udllrAe7l052j4t1tdls0nv3tttuQ6PRSG0U71nnvv29GbMnw1kTySoIArW1tXzzzTfcdNNNyGQy6WHz9fVFr9ejVColBtOenh78/f3p6elxeTjEG95oNKJWq/H09MRut6PT6fDy8kKlUrm4v53LF8UxfXx8UCqVEhW9v7+/ixtUJP8T3d5iHUXJBYDu7m78/Pwkl6pzOWLdxclXPLYcEBCAIAisXLmSK664ggkTJiCXyxEEQTombTabpRM5RqNR+t/ZJSm+0MxmM56enlgsFsaOHUtaWpp0hNfhcEicMjqdDh8fH1QqlSS1oFKp8PDw4M033+Tiiy8mJiaGG264QdpO0Ol0yOVyiRSvr69Pmjj9/Pxc+sput0sPo0qlchknu92Oj48P3d3deHt74+3tLZ3KCQwMRK/X4+vry5133imdvBK9cv7+/mzZsgWZTMa5557LnXfeKR1TFk9sBQYGIpPJMJlMLv3nPCaxsbEuLwrxKHpAQAByuVwSxfT19UUmk9HT04Onpyfe3t5SPna7nZ07d6LT6Vi8eLE0yfb19eHh4YFKpXJ5oarVauA4Y7BYVk5ODh0dHSxcuJBbb71V6gvxCHZAQAAeHh5SW4xGo8Tj4wyRGyg+Pp4xY8b8ou5qZ+4ePz8/brrpJhYtWsStt94qbSWIL2eVSkVNTQ2TJk3inXfeoaioiBEjRrj04ZYtW1AoFHz33XfcddddfP311yQkJJCbm8udd94ppRVZqGUyGa2trSQmJlJVVcVFF13Ec889x7///W8pbUdHB59//jkhISF4enoSGRnJypUrGTVqFHK5nNraWvbt28d1110nnVJrbGzk8ssvR6FQoNPpeOutt7j55puJjY1Fq9WyfPlyqqureeqppygsLKS5uZk9e/Zwxx13EBMTgyAIfP755xw5cgQfHx86Ojp48sknOXDgAHq9noKCAm6++WZeeuklYmJiUKlULF68GLvdzrJlyxg7diwKhYKtW7dy7bXX8p///IeJEyeyd+9elixZQmFhIYcPH+aaa65h+fLlREREMHPmTDw8PGhtbeXLL78kMDAQf39/pk+fzoYNG/D29qawsJDbb7+dLVu2SIzTV1xxBW+88Qb/+te/WLZsGXPnzqWwsFA6EXr++efT2NhIW1sbc+bMITAwkJycHOk+9PT0pLS0lAcffJDPP/+cqVOnEhsbi8Vi4eGHH2bRokXU1tbS3NzMrbfeyscff8z06dN57bXXeOSRR3j//fdJTU0lLCyM3t5ePvnkE77//nsef/xxiTX9u+++46uvvsLb25sLL7yQpqYmGhsbJWLNo0eP4nA4yM7O5tNPP8XhcNDW1kZ4eDiLFy9m165dtLW1YTAYuPnmm9m0aRPe3t4cOHCASy65hHfffZfk5GQmT56MTCYjJyeHL774gvHjx0unIpubm0lISGDu3LnIZDKam5tZsWIFkZGR0qnFV199lZ6eHh555BEOHDhAR0cHBw4c4L777pMME9E7qdVqueCCC9i1axfd3d3U1tZy9dVX8/zzz5OQkCA9L7fffjsOh4PCwkKOHj3KnDlz8PX1pbm5mc7OThYuXCgxkP/ecdYYNACHDx9mz549zJw5kzfeeEMKGg0JCSE+Pp5du3Zx9913U1JSQnNzs6S6XF9fz2OPPcZzzz3HypUrMRgMyOVy/Pz8uOuuu/j444+JiIhg1apVXHTRRSxatGiAQWO1Wvnoo48YMWIE3d3dpKSksH37dpKSkti1axf33XeftEIUBIG33nqLkJAQkpOT+fTTT3nppZdYs2YN5513Hvv378fDw4Ndu3bx1FNPSUaYIAjs27dPIosTtXO2b9+O2WymvLycxYsXU1paSnl5Ofv376ejo4Obb76Ze+65h3nz5rFjxw7uu+8+lEolJSUlFBcXk5SUxOWXX+5CJrdy5UqysrKoqKhAEATWrVtHSUkJf/nLX3jggQcYOXIkcXFxqFQq6cX52GOPsW3bNqKjo6WthHXr1pGZmUl1dTX//e9/eeGFF1izZg0REREUFBSQlpZGSEgIy5Yt49xzz+W7775j6dKlBAQEAMdXtx999BGHDx8mOjqa0tJSHn/8cZ566ilSU1NRKpUkJibi6enJ+vXrefzxx9m7dy/e3t7SS6q6upp///vfvPLKK+Tm5mIymTh8+DATJ05k48aNxMXFkZSUxFNPPcUzzzxDXV0d1dXVeHt7U1lZySWXXMJDDz3E/Pnz2b59O48++ihDhgwBGHAf7N69m7a2NhwOBy0tLVx22WVs2bKF6OhoIiIiqKysJDIykurqamnCA6TxMxqNtLS08Oabb/LnP/+ZhoYG8vLyOP/883nuueekMXzppZeorKyktbWViooKUlNT2bdvHwqFgqysLJ577jnuu+8+TCYTR44cITQ0lCNHjnDLLbfw0EMPMW7cOI4ePcpll13GjBkzBhjNBw8eZMyYMYMa7n19fSiVSpeTfeIq0dvb+ycZQGq1Gg8PD8mLIkqW/PnPf2batGmsXLmS0aNHM378eD766CNeeOEFqY7V1dXk5ubywAMPoFarkclk0rP/xhtvcP3110vl/Pe//yU1NZXs7GwOHjyIXC6noaEBjUZDc3Ozi/fHbrdLNAPr1q3jwQcfJDIykrS0NO6//37mzJmDn58fhw8fRqvV4uvry7x586R6iSzK2dnZdHZ2otFouPnmm7nxxhupra3l/fffZ+HChSiVSvLz84mJiZEMPIVCwQMPPMBtt93Gp59+SnFxMfPmzaOlpYXe3l5aWlq49tprpftRXCR1dXUxfPhw6urqCA4OlvizysvLKSsrIzQ0VKKniIiIYMyYMdK1drudlJQULBYL27dvx+FwIJPJmDt3Ln5+fhw4cICenh4uvPBCvL29CQ4OpqGhQVowGAwGlEolPj4+3H333ZSVlZGWlkZ9fT0HDx6kubmZ+fPnExMTQ1FREYmJiXz++ed0d3cjk8kkI8TT05ORI0eSm5vLuHHj2LBhA3q9nri4OMkw9PDwID4+nhEjRuDv749KpeKKK66grq6O4uJiKa/g4GA6Ojp4/fXXWbduHa2trcTGxrJr1y5iY2PZvn07d955J8HBwdJ4L168mKuvvprAwEDa2toYPnw433//vWTc3Hjjjfj4+BAeHi71oXjvJyQkEBcXR3h4OI8//jhvvvkmPT093HTTTUyaNAl/f380Gg0RERFkZ2dz4MABoqOjufXWW7n22ms5duwYn3zyCYsWLZLeDRMnTgSO7yyI91hBQQE5OTlMmTKFnTt3IggCjY2N3HvvvYSGhnLXXXdhMplYtWoVI0eORKPRsH//fomO45prrjmrFOPPmi0nkZ49OjqaoUOHIpfLmTp1Kn//+98ZOnQokZGR1NbWUltby4cffsj06dM555xz8PLyIiIigr6+PmQyGQEBAcTExHDttddy9OhRTCYTGzZsIDs7m6ioKIlCvj/sdjs5OTnU19czYsQIPvroI5KTk5k5cyYmk8ll60cmkzF79mwOHz5MXFwcVVVVtLW1ERISQmNjozQxqVQqiZEVjhtNznVXqVTs2rWL0tJS/Pz8sFqtqFQqiaJ+1KhRGAwGgoODsdlszJ07l1GjRlFUVMSqVavQ6/VER0ej1Wpd3I4iSdrEiRMlFtqYmBi6u7sJDw+nr6+P6667jmHDhrFjxw7kcjlqtZqcnBwaGhqYMmUKN998M1FRUYSFhZGVlUVSUhKdnZ00NTWxZcsWpk2bxqJFi3jrrbcIDAzE4XBw5ZVXSitmEXK5nLCwMMLDw7n//vvx9fWloqICh8PB3LlzWbhwIZs3bwaOv7TKyspYv349M2bMkDxU0dHR9Pb2YjAYWLt2LdOnT+fGG28kLS2N6OhoMjIySEpKktzHy5cvZ8KECSxYsIAffviBzs5OrFYrCxYsICMjg7KyskHvQYvFwjvvvMPkyZO54IIL2L59O2VlZWzcuBGVSkVcXBwVFRXs2LGDiRMnukz8Pj4+JCUlkZycTEpKipSfRqOhq6uL8PBwbDYbV1xxBT4+PtILb9q0afzlL38hMzOTmJgYSYtJJpNJ9Rk1ahTz5s2jsrKSyspKZDIZEydOZN68eRw6dGhAOxwOB0ePHpUMGnHbA44bNB9//LGLxIZY11deecWFzfhMIQjHddiCgoKIiIhAJpMxZcoUbr31Vi6++GJ0Op1E7jd06FAOHDhAQ0ODdH1ra6tElDlu3Di8vb1pamqit7dX0toSUVVVJT0vojTFiVzvNpuNqqoqacJ2bnN7ezuTJk3ivvvu4/zzz+eqq66irq6Of//73yfcwhYXAQqFQhKanTJlikTgCUiByyL7cnJyMk1NTZJH4ZVXXiE6OhpPT098fX3x8fE5YSCsTCZDoVBIDN0iu/iJ0lutViorK/Hy8sLhcHDs2DFMJhMKhYJJkyZJvF1yuZxzzjln0ABrDw8Pic28r6+Pqqoq/Pz8JKJNcQtw9OjRhIeHM3r0aN59910SExNdGL1nz57NgQMHKCsrQ61W89VXX0m6boMZziJbtpeXlxT7Jfalt7c3vr6+NDU1kZKSwqJFi3juueekd+U//vEP6urqpLRqtZqIiAhqa2uJiYlh/vz5vPzyyxLpplwuZ+LEiSc1CMxms6QrJ7K1WyyWQdN6e3ujUCiQy+XodDqsVivTpk3jueeeY/jw4dK9d/XVV0v3WHV1NRqNhtmzZ7Ns2TKCg4Nd7gm5XC5t+WdnZ3PLLbfw17/+lblz5+Lt7c2DDz4oPTNnA84qD40zxO0QuVzOkSNHGDt2LAEBAZjNZpqamqRVh/MP/N/DL27VeHl5MXv2bNavX8/IkSMlRmDn6+H4g3T//fezbNkyysvL0el06HQ6ZDIZ4eHhLoGVMpmMoUOHYjQa2bJlC5MnT+add97hiiuuoLm5GW9vb8477zxmzZrlcp3VakWr1UonNWQyGZ2dnYSFhXHBBRcwa9Yslz5wfmGJEgWiPEJbWxsXXXQRWVlZA4J1m5qaMJlM0nX9oVAoJA0pUb5h1qxZbN26VTKOkpKSXPaMxbpYrVZJjiEgIEB6GYj1Ex9A5zY4j0lUVJQkm+Dp6SnFxcyePZtZs2ZRWlpKW1sbgiC41F2c4EWJhbCwsAExL+IWh8g4rVAo0Gg02Gw2qW4KheKEwbZiYLHBYCAsLIygoCA0Gg233347r7zyCtdccw3XXHMNH3/8MY888ghvvfWWtHV0OhAlLkTV8Lq6OoxGI+Hh4QO2Jp3botfrkcvlhIaGYrVapRe8h4fHoPExotEQHx+PIAiUlJSg0WgkYdP29naCg4MpKiqSPDIajUbakj1dOPf1sWPHaG1t5bPPPuPhhx/G09OTnp4el0DtdevWMWnSJMaNG4fdbufrr79m1apV/POf/0ShUJCamkphYSErVqxAo9EwdOhQvvrqKzIyMtDpdNTX10ssz5MnT+bll1+mq6sLPz8/IiIiqKur4+DBg3R1ddHV1SUxWv/www/k5+dLpIziVrLo0Vm6dCnZ2dmEh4ej1Wq55557WLp0KT09PZKnyNPTk+LiYpRKJQaDgb6+PoxGoyRV8frrr5OWliaJtopobW2lpaUFk8nE5Zdfzr333sunn36Kj48P2dnZLgr34vhrNBqOHDmCzWaTWLH7+vqkehuNRsnAF99jJSUljB07Fk9PT77//nuKiookz8SFF17IU089JXldRo0axb333ktAQAC+vr5MmzYNQRDYv38/DQ0NkvyLmP9XX31FUlIS7e3tyOVyxo4dyzPPPMPll19OcHAwU6dO5corr+See+5h8eLFLvdxaGgocXFxKJVKLrvsMjZv3kxkZKTUbpPJhK+vL2VlZURHR0vM3c79Ase9zuL27JgxY3j55Zel7ezg4GAmTZqESqWirq5O8nIcO3aM2NhY5syZw+OPPy5tjw8bNoz33nuPsLAwfH19mTp1Kg6Hg5KSEkaMGIFSqUSlUtHW1oZOpyMrK4t9+/YRFxfHxIkTpS1/0agsLi6mu7tbui/EMIX29nbee+89IiIiyMrKkhY633//vXSPxcbG8t5770m7ERMmTJDaGhAQIOWZkZHBCy+8wLx58/Dx8cFqtXLVVVchl8tpa2sjMjLytJ/b3zLOKoPG398frVZLQUEBPT09dHR0YLFY+PbbbwkMDKSzs5P6+npGjhzJ8uXLSU5Olphre3t70el0UnyDTqeTHpja2lpCQkIIDw+nq6uLnTt3UlVVxe233y49fL29veTl5XHvvffyxRdfcNlll/HFF18watQo7Ha7tPIS4enpyZw5c9Bqtfz5z3/mmWeeITk5GY1Gw/PPP8/SpUsJCQlh4cKF0jXe3t5kZ2ezfPlyUlNTaW9vJz09nSVLlkh6TfPmzcNgMKDT6eju7pao+E0mk9RGlUrF5MmTefLJJ5k/fz7R0dHMmTNHKmfSpEk88MADDB8+nPr6emprawkICECv10sTkE6nY8iQIVRWVvLiiy8SHh7O+PHjWb58OU899RSJiYksWLAAb29vycUpeouGDx9OTk4OHh4eXHTRRVKsSG9vr/Tj/IIWhOPsyxUVFchkMoYNGyZJT2RmZtLW1sazzz5LTEwMU6dORaPR8OGHH2IymWhpaaGtrY3e3l5kMhkZGRncd999jBs3jqlTpxIcHMyRI0fIyMigp6cHg8HA5ZdfzubNm1GpVMTHxxMXF+fSf+LpMtFoECUQrFYrF198MZs3b2bmzJkMGzZM2qK57bbbqK+vl5SxxfQi7HY7nZ2dUlBgTEwMW7ZsITg4mK6uLjo7OzGZTJJ8gdVqZdy4cdx3331MnTqVCRMmSG0RZQ70ej1/+tOf2L59O9HR0QQFBTFs2DBJtkEcS4fDIa2y+/r62LNnDyqViubmZurr69m/fz933303H374IXPnzpVimkRP2759+xg7dqy0VXK6sNvthIaGct9999Ha2oqXlxePP/44gYGBdHd3c9NNN+Hr6ysZYampqYSGhkrX3njjjZJiuEKhICQkhNdee40jR46QnZ1NbGws999/PyqVigcffBAPDw8uueQSQkJCSE1Nxc/PD71eL01mN910E97e3jzzzDMEBwczatQoHA4HQ4YMke4dHx8fNBoNd999NwqFgieeeIJdu3YRGhrKmDFjCAwMpLGxkWuvvZaIiAjguHF5zz330NraSmhoKFdffbUUixEUFMTzzz/P7t27iY2NJTMzc8D2X2VlJbfccgtRUVEsW7aMvLw8Ro0aRXBwMDfffDMqlUq6H2UyGZdccgm7du0iJSWFRx99FA8PD+666y6sViuLFi0iMDAQDw8Prr32WhwOB/fffz92u13aQpw7dy6enp5kZGRIXpSnn36ahoYGJk+eTGBgIC+88AJVVVWcc845BAUFsWTJEpqamrjrrrtQq9W0tbVJQda33347eXl5zJgxg8bGRiZOnCgZy+PHj0cmkxEUFMTll18ubTWLkMlk3Hnnnfj6+qJQKMjIyEAul2M2m7nrrrsQBIErr7yS3NxcvL29pS2W+fPnu+Tl7+/P9ddfj8lkYurUqXh6etLe3s60adOw2WzU19eTmZlJeno6+fn59PX10dzczF133UVAQABPPvkk9fX1TJkyhbCwMJYuXUpZWRmTJk1Co9Fw3333SQtGgOTkZP7yl7/g6enJkiVLyMvLo6+vjzvuuENKIy6CrVYrGRkZxMbGYjKZpDJfffVV9u/f7+J1FQSBjIwM6R5LSUnhpZdeoqysjPHjx+Pr68sdd9wB/B+xqb+/P7fffjs7d+7Ex8eHc845h+bmZtra2pg/fz5paWmn/cz+1nFW8dDYbDby8vKIiIigs7MTtVpNXFwcx44dkwJpHQ4HkZGR5OXlERQUhKenJwEBATQ0NBAeHk5PTw8ymQy1Wk1LSwsajYYNGzYQFxdHT08PJpOJ6dOns337dq677jrJC2Cz2aipqUEQBCIiIlCr1dTV1aHX64mJiRkQSArHAzptNht+fn50dHRIL2utVkt1dTUZGRloNBqXDh3hogABAABJREFU64xGI/n5+QQEBODl5UVsbCx1dXU0NzeTmZmJUqmksrISjUaD1Wqlt7eXqKgo6urqJAVppVIp7V8DDBs2zIWuXhAEqquraWtrk2j7dTodvb29hIaG0tTURGRkJGFhYbS0tFBRUUFaWhqhoaG0tLRQWVnJ0KFDCQoKora2FpPJJK04kpOTkcvlVFZWSuKIXV1dNDU1ERsbS319vbTFJLZ7w4YN7N69m6uvvpr4+Hg8PT2prKwkMDCQmJgYOjo6KC0tJSUlhYiICLq7uykqKiI8PByVSoWPjw/19fXExsbi7e1NXl4e4eHhJCYmotfrKS8vJzExkcbGRsLDwwkNDZW2JBISEqQ4nOjoaNrb2/Hy8iIhIUF6wYiGclxcHP7+/lRWVuJwOEhMTEQQjotByuVyYmNjaWtrw2QySe5sZ89VdXU1DoeDpKQkLBYLRUVFxMbGYrfbUalUUh81Njai0WgIDg6W7uOUlBRMJhNFRUUkJSWh1WoJDg4mMjJS2jKIi4vDy8uL8vJyNBqNdPIiJSVFmsxMJhM1NTXYbDa8vLwwm81oNBoUCgWfffYZ06dPZ//+/Vx++eW8+eab3HjjjXz00UfExMQQGhoqTRZu/DQIgsDatWv54YcfePrpp89aJnJBEFi1ahWlpaXceOON0nP1a8Fut/P444+TmpoqeTDc+P3grDJofgmUlJTw/PPPM3PmTHx9fUlLSyMqKkraE/09ETj9HuFwOFi1ahUFBQU8+eSTkgq5G/9b2Gw2KXi2t7eXjIwM3nrrLebPn09sbKwkhvlLn4r6o8DhcEj6cyNGjJAOBpxtEE/32O12UlNTf/X3qcVi4ciRI3h6ejJ8+HC3cf47g9ugOQUEQaC5uZmuri4iIyMH9bS48cvB4XBQV1eH2WwmPj7+jGI03Ph54fyqOHr0KGVlZfzpT38aELvjhhtuuPFrwG3QuOGGG2cEkW9HDLz+tYwYkUSsvb2dsLCwQU/biIGxMpkMX19fif8pMDBQYr8ODAz8xVfidrudlpYWjEYjCQkJJ9xCEgnPxBOFg/WtGHyuVCrPKKjcjR8HQRDQ6XQAUkDvzw2Hw0F3dzdyudyljI6ODhQKxYDYIpHvqq+vj5CQkAH3iclkoru722Vb+48At2/YDTfcOCOIR2ZPdvT3dGGz2X4SQ3djYyO33XbbCfOw2WwsXbpUoslfsmQJb7zxhkSctnTp0p901Px0IAjHWau1Wi3btm3j2LFjJ02fn58vBer2z0cMCH/22WfJycn5BWt95hDJL8822O12VqxYccZM6WcCkan866+/lsZZPDlWUFAwIL3D4WD9+vUsXbp00Py0Wi0bNmz4xer7W4XboHHDDTd+FTgcDj755BM2btwo6Tc1Nzej0+kwmUxUV1ej0+mw2+20t7fT1dVFTU2NRPEu8gPBiTlkvLy8uPzyyykoKEClUrFw4UIOHjwoEQGee+65+Pj4UFdXR0tLy6D52O12tFotWq1WYrWur6+no6ND0gnr7u6mqqpqUHmN9vZ2vvrqK9RqNXPnziUkJASbzUZrayttbW00NDRIFPYtLS0olcpB9bO0Wi3//ve/0ev1BAcHS8rz4vFk5z7r3w6xD9va2qiursZms0nl1dbWYrfb6evro6Ojg7q6OqxWq8Q+K0ogtLe3U1NTI51KbG9vlxiAbTYbb731Fnv37sVqtdLV1SVtFYts1ceOHaOzs1OSN2hsbBxUWkP0vGm1WpqbmyVvYFtbm1Q3UV5ALKO5uZne3l66urok3pfOzk7a2tqkQPz6+no6Ozsl9mnxiLnJZKKurk469CFeL0KhUBAeHo7RaKSqqgqdTofZbKaurk76u7m52eUa8XRje3u71I6enh7psEFnZyft7e0YjUYaGhrw8PAgOjoaq9VKVVUVzz33HN3d3UyfPp3MzEysVittbW1otVrq6+uRyWTEx8djsVgwGo0ShYPBYKCtrY3Y2FguvPBCHA6Hy7Mjyip0dXVJh2V+De2rXwpnVei86IYzm80SpbtIc69WqyVCI1FewA033Pj14HA4/h977x0eVZn+/7+mpU167ySBQCghoXdBmoAIKiprWXXdVXf97Loqa1/sbnF33WJbZV1RFukgICA9QIAUEgIJKSSk10mZTKbX8/uD3znfTBIQXF0V53VduWBmznnaeWbOfZ7nvt83BQUFDBkyhNLSUl544QXmzJnDkCFD0Ol0REREsHHjRl588UUef/xxli5dSnZ2Nvfccw9BQUGcPn2a1NRUN/HJvshkMoYNG4ZKpaKgoACDwUBDQwNlZWVUVlYyc+ZMtm7dSlRUFKtXr+aZZ55xC2N1uVxs2bKF+Ph4vvjiCxYuXMiRI0eYM2cO69at45lnnuG1114jMTERnU5HXFwcDz/8sNvKVU9PjyQfsGPHDsaNG8f48eN59NFHue+++9i2bRuvvfYaVVVVkv7TQCsdGo2Gs2fPSjmDTp8+TU1NDQaDgRUrVrB9+3bCw8PZtGkTf//736XtKEEQaGxs5Be/+AW33norR48eZd68eSQnJ2MwGDh+/DgjR46UQvYTExOZPXs2dXV1hISE8Mknn7Bo0SKampooLy8nMDCQlJQU/vWvf7F06VL27NnD3//+d/Ly8oiKiqK2tpZNmzYxZswYdu7cyU9/+lPeeustpk6dyp///GeeeOIJDAYDvr6+bNu2jTvuuENKGSDyn//8h9jYWI4ePcr9998vSWs0NTXh5+dHamqqJJqo1+uxWq1ERUWRnp7OX/7yF1577TV++ctfMmPGDMLCwmhoaGDhwoWsXbuWF154QRLo+89//sP999/P73//e5566ikiIiLIzs7mRz/6Ub/xLywspKenh7Nnz/LXv/6VF154gfvvv5/hw4ezZ88e7r33XrfjDx8+zH/+8x+WLVvG2rVrufvuu8nNzWXKlCmEh4ezdetWXnjhBZ588km3hKWNjY2Ulpai1WrZsGED0dHRjBgxgldeeYXbbruN7du389RTT0nz22Aw8Mtf/pK//e1vGI1Gamtr8fHxYevWrbz++us88sgjLFiwgPz8fG655RaGDx/Opk2bSEhIYMuWLfzhD38gKSnpcl/V7w3X1ArN+fPnWb16NXv27OHzzz+nqqqKPXv2kJWVRUtLC59//jk7d+78tpvpwYMHLgqLRUZGkpyczJAhQ/D29pa0Wc6ePYsgCJKkgK+vL9dffz3Tp0+nvLyctWvXSgqvYWFhl63H29ubG2+8kTVr1qDVavnRj37EmjVrpPQKe/bswWw2M378eMmgENFqtWRnZzNx4kR+/etfM2LECBYsWIBSqaSurg6LxUJISAgZGRnccsstlJaW9qs/KiqK4OBgMjIyGDRokKS/4+Pjw8KFC0lJSaG6upr169czffp0xo8fP+BDV1hYGGFhYQwaNAiZTMbEiRO58847qaqq4tSpU5SUlAAwatQot37IZDIiIyPx8/Pj1ltv5Te/+Q3bt29n7dq1mEwm0tLSUCqVxMXFERMTw/PPP8+hQ4cYM2YM8+bN4/7772fz5s0YDAaSkpLw8/MjOjqa4OBg7rjjDmnVSkzlEhMTIymkl5SUoNfryc/PJyMjg/j4eBITE9myZQtms5nMzMwBlXOnTp1KeHg4nZ2d1NXVsWnTJqZNm8bdd9/N3Llz2bBhA5MnT2bZsmXccsstqNVqrFYroaGhktSAj48Py5Yt45577mHRokXIZDJqamooKSnhwoULjBs3TjJgb775Zvbu3Ut1dbWkMN6X8ePHS7pGlZWVkl5VdXU1kyZN6ifkGRsbS2BgIEuWLMHf35+MjAzmzJlDSUkJgYGBWCwWKb9b79WdiIgIwsPDSU5OJj4+HqvVSlxcHIGBgfz4xz/mnnvu4fPPP5eOj4yMZPbs2Rw+fJi6ujomTZpEcnIyVquVgIAA/Pz8mDFjBrNmzaK8vJwLFy7Q2trKjBkzCA0NJT4+/nJfn+8V15RBExQUxIQJExAEQXoa2759OzExMYSFhbFkyRKWLl36bTfTgwcPvRC3R0QlZJvNhlwuZ/78+Tz88MOSmrTou+NyuST5+d6I2yh9EfMQZWVlER8fz80338znn39OcnIyTqcTo9EopesQEwX2LrOurk5Sb7VYLGzZsoXAwEApGaxYh6hLdKX0VikXt6D6GlR9j+9dvuiQLToSi2P20EMPDejkLNalVqslbamRI0dy1113MW3aNARBkK5Bd3c39fX1bilYBg8ezO233878+fMlhWGR3u2qqanh6NGjJCcnS6J5s2bNYt++fdxyyy2Ehoai0+mYMGECP/nJTxgxYoRbOwVBYPfu3dhsNiIjI6WtMY1GIyV01Wg0tLa24u3tLTlYi9t2YltEVW2bzcaWLVsIDQ3F398fq9VKdXU1NpuN4OBg7HY7CxcupLCwkPLychITEwccf3FsoqOjJYXk2tpa8vLyJOG7S425aOzI5XJpi0dsZ985e6moQbGMwMBAQkND3c657bbb2LVrF1qttp+TcO82CILAmDFj8PPzo7CwkAcffHBAZ/rvK9fUlpOYyC85OZnKykqGDBnCfffdx+uvv84zzzwjJffy4MHDd4O4uDiOHz9OWFgYJpOJjo4OMjIyeOedd1i5ciWDBw9m0aJFWK1WSTHZZrNx8803s2rVKm644QYMBgONjY3s2bOHOXPmkJmZ2a+exMRE7rjjDiZOnEhQUBC33347o0ePRq1WExcXx4oVKxg5ciRLlixxOy8sLIzRo0fzq1/9igkTJkgJAY8ePYper6eoqEhSXRZ9RWw2m5tekph6QKvVSv4tOp0Om83mlppgwYIF/PWvf2X06NH09PTQ2dlJVFSUVI6YF+no0aN0d3fT0dFBR0cHZrOZtLQ0PvjgA1auXMmQIUO4/fbb+42B0WiksLCQhoYG7rvvPqqqqnj22WeZPn06kyZNorOzk+7ubiwWC7fddhsvv/wyBQUFTJ8+ncWLF/Pqq68yc+ZMxo4di0qlkvpkNpvp6ekhJiaGQ4cOkZSURGFhIeHh4bS2tlJXV8fZs2fp7OwkLCyMoUOHMmnSJJ544gnGjx/P3LlzJVVlQEq3Ifr9lJWVsWjRIp588klmzJjBggULWLZsGc8//zwzZ85k7ty5DB48mE8++URSEW5sbMRkMqHRaPD39+fs2bMcPXpUUk0XE2eOGTNGUiieN2/eJaPlwsPDycvLo6CggJSUFEmjZunSpdhstn4GpOjjZbFY6O7uxmw2097eTldXl5Tlvr29nQ0bNmCz2WhsbKSzs1PKgdXV1UVOTg7t7e1YLBYpXc3p06cpKyvj1ltvpbi4WHKxiI2NZezYsaSkpCCTyWhra8NoNEptEL87er2ehoYGamtrUalUdHZ2MnTo0GtG5+iaCtv+6KOPOHXqFCNHjiQ3N5e77roLf39/SVlVfPpZvHjxt9xSDx48wMXw0tbWVkJCQtDr9QQHB+Pv749Op0Oj0RAfH49SqUSj0RAQECA5mYaFhdHc3IxMJsPHx4eQkBA6OzsJDAwccLtGzAIt/nD3/r/o4BkRETGgzpT4eXh4OMHBwZLTqPjE7nA4UKlUyOVyKU9Y77Bso9GIVquVcuvI5XL8/Pzo7u4mKCgIvV6Pt7c3gYGB1NfX4+fnh5eXl+QH2LsPTU1N+Pr6SukexJxXERERmM1maczElYze4/zjH/+Yl19+mYiICCIiIiSNJy8vL0ld3WazERERgUqloqWlBYfDQVxcHDKZTEoEGhcXJ6UJCQoKkvohk8no7OwkJiaGhoYGIiIi6OrqwmAwUFxczLBhw6ScRQsWLKCuro7AwEAiIiL6rfZ0d3ej0+kICgrCYrEQERFBY2MjPj4+REVFSQrcKpWK6OhoKT2KmOjW19eXrq4u1Go1wcHBtLa2SnWICT8bGxuJjY3Fz88Pm83Grl27mDVr1oBbmA6HQ8oBGBsbi1KpxGq1cuDAAcaOHStl9e7bB5PJJKXy8PPzw+l0YrfbiYiIoL29HZlMhkqlQq1W093dDVw0npqbm6UVQbgYQfbkk09KaTkCAwOl6xUWFiblzFqyZAl+fn5otVpMJpOUskatVmOz2SS/NR8fH2m+paamMnr06C/5pn4/uKYMGqPRSFlZGYMGDaK5uZn4+Hi0Wi1KpZL4+Hg6OjoA3J4GPHjw4OFaRhQHfeihh3jjjTdIS0v7n2qT5OTk8NFHH5GRkUFERATTp0//TumjNDQ08Mwzz3Dbbbdx8803X1G72tvbWbFiBTfccAN33nnnN6qOLQgCZ86c4eWXX+af//wnkZGRbm08d+4cr7/+Og8++CCzZs26bPsFQWDVqlVUVVUxdOhQYmNjmTlzpmeFxoMHDx48fPcRQ7q7u7tRq9UEBgb+T40JcYtNjD5VqVTfGWMGLoa0d3d3S0k7rwQxXD8oKOgbz7Ml6tKIflx9jQ+Hw4FOp+u3oncpxES4on/TteRD4zFoPHjw4MGDBw/fe64pp2APHjx8fxDFz8SoJj8/P/z9/ZHJZJITMICXlxdBQUGSMrEoXjdQolK73Y5GoyEqKuqKn5wFQZCinQIDA6Vy2traCAgIkGTnu7q6MJvNREVFSZFXl+ubyWSSwnL7ftbR0YGvr6+kFeN0OqWM8H2fmG02G+3t7URHR/f7TPQNEiO+vLy8CAgIQKVSSeJzdrsdhUJBQECA5Ftzufb9kBBXjiIjI7/0WFHPRwy5/yax2WxoNBpiYmKuegXFbre7ORjrdDpUKhWhoaEYjUaMRiNhYWFfOoe/j1xTYdsePHj4/iAIAllZWdx3332cPXuWDRs28PLLL9Pc3IzFYuHxxx/n448/Zvfu3axYsQKNRiOlEdi4ceOAIdKtra386le/wmg0XnE7bDYbq1ev5h//+AdwcQn/3XffpampidWrV1NVVUVJSQnr1q2jpqaGf/3rX/3SEvSlurqaF198ccDj7HY7r7/+Onl5eW5t2LJly4CaLF/WJ61Wy913380XX3zBkSNHeOWVVzhw4ABOp5Pdu3fz8MMPc+rUKVauXMm+ffsQBIHKykpefvnlL+3HtU5FRQU5OTlXdKwgCOzatUvyxbwaxG2jKx3vlpYWHn30UQwGw1XXZbPZeO211/jnP/+JyWTiN7/5DevXr5cMsvfff39AJeprAY9B48GDh28FhUJBQkIC3t7eXH/99fz0pz8lNTWVFStW4OPjg1qtZtiwYdx11110dHRw6NAhLBYL5eXlbNq0acD8TeHh4SgUCtra2qiqqpKiUxobG3E4HLS2tvY7z8vLi5SUFMlgqKqqoqysjAkTJpCens7HH3/MRx99xMSJE5k8eTI5OTnU19dfsl+CIEghvuXl5f0+FyOYjEYjFRUVGI1GvLy8uPvuu1GpVLhcLpqbm6mtrcVgMBAaGopMJqO9vZ0LFy643RRlMhlRUVHI5XLGjx/PLbfcwkMPPcTrr79OaWkpiYmJeHt7s3DhQubMmcPbb7+Nw+EgLy+P/Px8qqqqBmy/1WqloaGBlpYWSca/sbGRjo4OSTq/sbERo9FIS0sLOp0OrVZLU1OTFGLc2toqhR1fuHBBOlev19PY2EhPT49bvXa7ndbWVjo6OiTV5PPnz0vKyGazmZqaGinayGKx0N7eTk1NDVarlba2Nim1QG90Op00ljU1Neh0Onp6eujo6GDUqFFMnTpVmhsajYaamhpcLpeURqCmpobOzk7got5LWFgYFouFtrY2GhoapPQMDoeD+vp6GhoapHQUvdvw0ksvUV9fL5VbXV2N1WqlqamJhoYGKaVHe3u7FM0HFw2U6upqKTWH0WiksbFRkgroi5+fH8uXL6egoICAgAAWLVpETk6OpKS/cOFCFAoFFy5ckB4SrhU8Bo0HDx6+E8jlchYsWEBlZSUtLS0AdHd3U1ZWRk9PD8OHD+fcuXPccMMNhISEcPz48QF/jPV6PQcOHODFF19k3bp1HD9+nD/+8Y+SYFvflY7ewmdwUXE8MDAQuVxOVFQUZ86coby8nPDwcJRKJd7e3tTU1FyyH1qtFoVCwT333MPatWsHFE4TUxds376dN998k/r6en7+85/T09PD0aNHOXz4MBs2bOCFF17AYDBgMBjIzc3l1Vdf5dSpU5esWyaTkZCQwJgxYzh48CCAdNPMy8tj8uTJ6HQ6vL29Wb58OZ9++umAYoQffPABdXV1vPnmmzQ1NbF27Vo6Ozs5dOgQXV1dfPzxx2g0Gk6fPs3nn3/Oxo0b6ezs5KmnnqKuro7777+fDRs2cOjQITZs2EBDQwO//vWvaWxs5IMPPkCv13Ps2DG369fU1MRPf/pTsrOz+d3vfsdHH33EkSNH+OMf/4jRaOQf//gHPT09fPjhh+zZs4fPPvuMxx9/nHXr1rFnzx7y8/PZtGkT//73v93KbWlpYcWKFZhMJlauXElNTQ0XLlygsLCQjRs38u6779La2soDDzzA4cOHWblyJSUlJezcuZOnn36ao0eP8vTTT9PV1cWKFSsoKysjNzeXX/ziF5w8eZLHHnuMnp4eNm7cSGlpKW+88Qb//Oc/3VbbjEYjZ86cQafTsXr1an7729+ybt06Pv30U3Jzc1m7di2FhYUUFhaya9cuOjs70el0OBwONm7cSGNjIytWrKC6uppVq1ah1+vJysq65BwYM2YMer2e0tJSzGYzZWVl1NbWUlFRQWpqKhs3bqSpqYknnniChoaGS86n7xseg8aDBw/fGRQKBUqlUrrpV1dXU1NTw1/+8hdGjRrFwYMH6ezsJCkpiY8++mjAxHr+/v7ceeedksT/7NmzaWtro7q6msDAQMLDwy/bBlFZtff/B3pvIARBIDs7m56eHvz8/Ni/f79knPWtY/Lkydx+++1UVVURGRkp+b2cPn0ahULB9OnTJcl7tVrNjTfeyIQJEwZcVemLqIsDF1cHioqKmDt3LitWrODYsWPo9XoCAgLYs2cPGo2m3/nz5s3D19dXykFlNBr597//zejRo1EqleTl5VFYWMiIESMk9V1RdTcsLAw/Pz/uvPNO0tPTyc3NxWq1MmXKFKxWKzU1NezcuZMxY8a41RkZGSmt1mVmZhISEsJNN91ETU0NZ86cobGxkdGjR0s5mWJjY4mOjubpp5/m+PHjGI1GhgwZ4qbgDDBkyBAiIiJobm4mPDyc/Px8mpubmThxopTgMTQ0FF9fX+bPn8+oUaOoq6sjOjqa6Ohobr/9drq7u1EqlYSEhOB0OomJiSE4OJibb74ZlUqFTqfj5MmThISEMGHCBJKTk918vAIDAwkODmbIkCHEx8eTlJTEc889x5w5c0hISECn01FTU8PatWuZMGECGRkZhISE0NzczJEjRzCbzUybNg2r1UpLSwsbN25k7Nixl7z+arWaOXPmsHr1ahQKBfPnz2f16tWoVCq6uro4duwYFouFqVOnDphQ9fuKx6Dx4MHDdwJBEDh58iSpqanExMRIT5qLFy8mNTWVuro6EhISyMzM5K677qK1tZXS0tJ+qzSiwREYGEhUVBRBQUHccMMNvPnmmwwfPvxL25GamorBYMDlctHe3k5GRgbDhg2jq6tLEkZLTk6Wsj73RlwNmT9/PtOnT2fKlCls3brV7bjLpUuQyWTcdNNNXLhwgZ6eHilJYl/5+suNYWdnJ6WlpcyZMwe4aCjceOONTJ8+XdKkmTt3Ltdddx3jxo1j+/btbmW6XC42b96Ml5cX4eHhCILAsmXLmDhxIr/97W+xWCy8+OKLlJWV8eGHH0r1in/w/9IOOBwOrFartKWoVqtZsWIFgiDwhz/8oZ9PyUCGpMvlQi6X09bWhsPhQK1WExAQgEwmkxzFe3p6SElJYdmyZVx33XVu/VEoFNx000289957zJgxg0OHDmE2myVn7951970ml3OY7vvZj370I3JycggPD5fGvu+16d0mgL1796LVaomLi5O28nQ6nXScw+HAaDQyc+ZMHnjgAUJCQnjkkUcICQnhlVdewW63Y7PZBpwTixcv5vPPP2f48OEsW7aMjRs3Mnz4cJxOJ2azWSozMDDwmtl28kQ5efDg4VvB6XRKeZIOHTok+Yj8/ve/l+T06+vrsdlsyGQy1q9fz6xZs4iPj8dutzNq1Cg+/vhjXn/9dUkdWC6XEx4eTlFREa2trfzsZz9DLpezdOlS8vPzB8y543A4pGzO4lP+4MGDOX36NGfOnOGee+7BZDJx8uRJbDYb48aNIzY2ltdee4377ruPlJQU4OIN69ChQ9jtduLj45HJZFx33XW8++673HrrrcTFxUn1GQwGOjo6CAkJwWw209LSgsVioaOjg5KSEhoaGnC5XFJ9VquVjo4OtFotTqdTusmLBorNZiMvL4/a2lqKior45S9/SVpaGmvXrpV8NiIjI9m/fz8ul8utff/+979ZsmQJMTEx0nWprKzk8OHD9PT0cPr0aVwuFzNmzKClpYWenh5yc3O5/fbbOX/+PIMHD+bzzz8nKCgIh8NBbW2tpAA9ePBgXC4XTz75JGlpaSxevJh9+/Zx4403cujQIbdr0d3djcPhoLu7m56eHuRyOZ2dndhsNpKSkkhNTSUrK4v29nZ+8pOf0Nraik6nw2KxsGTJEl5++WXJiBT7AhcNj2nTprFnzx7mzZtHeXk5I0eORBAENBqNlGbCZrPR1dWFTqeT2mI2m+ns7MRqtaLRaDAYDLS3t+N0OrFarW6pBfLy8iQfGbVazXXXXSf1z8vLC39/f/bt24dGo6Grqwur1UpdXZ3kP6PRaFi6dCkffPAB8+fPx2QyIZfLCQ4O5vHHH2fkyJHcdNNNfPHFF8yfP1/y+3n33Xd55ZVX3NIviFnmf/SjH5Geno5CoWD58uWkpqZK+byeeOIJRowYwW233fYNfLu/HTw6NP9jxKeYyy1bX2v8EPvs4csRk0za7Xbkcrkk/y/m49FqtcjlckmKXgx1DgwMxOl00tnZKaVBEENQRcdH0VgICgrCZrPR3NxMeXk5CxYs6DcHnU4nWq0Wh8NBSEgIXl5e2Gw2Wltb8ff3lxIBdnR0YDKZiImJQaVSodFopOPFuru6unC5XISFhSGTydBqtVgsFoKCgiRBNDFHkVKpRKlUYjKZ8PPzk/7dsmULY8aMweVyUVpayvXXXy+FXZvNZmQymVS+2F+9Xo9MJkOpVBIYGIiXl5db2HZQUBB+fn6XbF9wcDB+fn5SP0TZflGuX6lUYjab8ff3l9IKOJ1OyQm7qamJgIAABEHAy8tLktsPCAjAZDLR2tpKdHS0dK7dbickJARvb2/pephMJnQ6nVs/vb293RyjW1tbpVD67u5uabtIqVTS0tKCTCYjOjq6n8Bc79QXJpMJHx8fqf/iqo+4DWcymdwSpYqfied6eXlJ1y0gIICenh4UCgVbt25l+vTpmM1mSktL+dGPfiRtO4nXQvwddDqdhIaGYrPZ6OzslFJ/REVF0dLSIn0fgoODpfkbGRmJv7+/JHUQHByMSqWio6ODqKiofvO6b7oP8XoCWCwWtzKvld9lj0HzXyB6/4v7+F5eXgNqY/Q+vrq6mvz8fBYvXoy/v///qqnfGmKo4LFjx7jxxhv/5yqlHn7YCILAX//6VxobG1m5ciUhISHfdpMui91u5+WXX0alUpGQkMCwYcOYPHnyNaXmei2i0+l4/vnnSUhIICYmhtGjR5ORkeH5rfsf4zForhJRhEv8gTl37pwkABYfH09ycvIlz3U6nVRUVBAeHo7ZbGbQoEHS+3K5/Komf+92fF1fmm+iTJfLRUVFBREREeh0Oikb7NeBIAjS0vv/+ofjmxirb5LLtdflciEIwjV50xQzYMPFcNbv+rUSw3+7u7vx9vaWfEU8fLcRQ93Flanvw1y7FvlBGTQ2m42CggIMBgNxcXGkpaVdVVIxQRBoa2ujpaWFxMRECgsL8fb2ZuLEiXh7e3Pu3Dk0Gg3jx4+XFEfFek+dOkVPTw9jx46VMq0WFRUxbNgwnE4nNpuNYcOGXfGXwOl0snfvXhYuXPi1fHHEbLUGg4GAgABKS0sJCwuTPOnz8/NxOp1MmDDBba/2SsrtzZe11Ww2k5+fj8ViYcKECYSEhNDY2EhpaSmjRo2SnEXh4o04JyeHUaNGuY33N40gCHR0dEhOqmfOnEGtVjNhwgRUKhVnzpyhu7ubCRMmfKNJ3+x2OwUFBXR3d5OZmUlUVBSdnZ0UFhaSmppKUlKSNFaCIFBYWEh8fHy/5HbV1dWYTCZGjRr1pf02GAz4+PjQ3d0tKfKGhYWh1WoJDg7+RpP0efDgwcPl+EH9+iiVSiorKzl27JjkFHc1OJ1ODh8+zLBhw/D396e4uJg9e/ZIYlhVVVWo1WrJQbF3vUqlkjfffFPakmpra6Onp4fQ0FAGDRpEUVFRP6GpyyEmfPu6sFqtHD9+nNTUVMLCwti6dSsVFRVSpIK4ynK1idgGCnu9HF5eXlitVt577z28vb2lLStxz7hv2ampqezbt29ALY1vCpfLxcGDB0lLSyMoKIgTJ05w7NgxlEolTqeT8+fPExQUdNntx68DhUKBt7c3f/rTnyQjU6PRoNVqJb8TEdFJUFSQ7Y3VasVsNl+2LvE65OXl0dTUxAcffICPjw9//vOfaWlpobW1lYKCgquKlnC5XLS1tdHd3X3F53xTuFwuyVlzIET/mGtFiMxisVBbW/uV+iJutdfV1f1Pv3dfld5ifn37q9VqaW1tvaIyOjs7aW1txWq14nQ6aW5upr6+nq6uLvR6Pc3NzVelUO3h6+cHZdDIZDLKy8u5/vrrv5IvR01NDSqVCl9fX7y9vbn55pspLi7GZDJRXFxMeno6EydO7JcjQy6Xk5mZSWhoKFVVVbS2tmIwGLjlllsICAhALpeTnJxMfn5+v/DOS/1dCtFbvqurSzpWr9ej0Wgwm83YbDba2tqwWCy0trZKN7KysjJCQ0NRqVSS0uSRI0ew2+2cOnWKRYsWkZqa+o0/gSsUCqZOnSr9YNTV1aFSqVi0aFG/ZVyZTEZ4eDgdHR1f6aYo5tRpb2+Xtl1MJhNtbW2YTCYpikBUBdXr9dKN3el0olar8fHx4fbbbycvLw+r1crp06eZPHkymZmZ33gWXrlcTnp6OrGxsZSXl6PRaOju7ua2224bcH6LeXu+ipCW3W7nP//5D+PGjeP48eOSum54eDh79+4lLS2N48ePD6i5cimcTid//etfOXLkyFW35+vG6XTy3nvvsXfvXrf3xe+Pw+Fg8+bN/Oc///mWWvjVEB1D+6ZU6OzsZP/+/Vdt0NhsNkwmE9XV1Tz55JOXldB3OBxSjqlvm5MnT/L73/++3/tVVVWXFSoUEQSBN954g7KyMhwOB2VlZbz00kv87ne/o6GhAYvFwpYtW/j888/7nTfQ+Hv4ZvhBhW1brVaqq6sH1KJwOBxUV1eTnJzsZpCYTCZaWlpISUmhoqKCuLg46UaRmJhIZGQkmzdvZvz48Zf1D1GpVCxevJg1a9Zwyy23MHXqVMlnQVT33Lp1q6RfIAgCtbW1lJSU9Ctr8ODBDB06tN/7giBw9OhR1Go1lZWVLFq0CK1WS1ZWFmPGjGHr1q3cfPPNHD58mOTkZFJSUvjss894+OGHKSsrc3NimzBhAn/961/ZvXs3Y8eOJTw8nObmZpxOJw6HA6VSiVqtllYCeus2iL4tNptNiibofcylEI/z8/Pj+uuvZ82aNdxwww1MmjQJvV5Pd3e3FHEBEBcXJ0XB1NbWStEol6un9/UpKiqiu7ub7u5uxo8fj4+PD1u3bmXGjBm8+eab/OhHPyIvLw+lUsm0adNYs2YNTzzxBJWVlcTGxkplpaam4uXlxbZt2xg7dizx8fH9xsrPz88tMqX3WIn7719lrBQKBUuWLGHt2rXcfvvtTJ06VQr/Ff2LAGJjY1GpVMTHx1NRUXFZX6+BECXeg4KC8Pf3p6amBofDIUVZyGQyIiMjOXjwIPfcc88VPSwolUrCw8MxGo2cO3eO+Ph4AgMD0el0dHR0EB4ejkwmQ6PREB8fj9lsxuFwEBwcTH19Pd7e3m4rraKOh8lkIjQ0lJaWFuLj46XwXC8vL6KiojAYDDQ2NhIdHU1wcLAUNu3r69tv9aqnp4eXXnqJn//850RGRtLR0UFFRQWhoaFERERgs9moq6vD39/fbTsULhrMra2tOJ1OoqOjJQPZ29ub6OhotFqtJHUfFRWF2WzGYrGQnJyMy+WSDM9Bgwa5PUiICTi9vLxob28nKSmJuro6oqKipDDwhoYGQkJC8Pf357XXXmPZsmUMHz4cvV6PwWCQtGlkMhkOh4Pm5mZcLhcJCQmYTCa0Wq1bhBdcNPr+9a9/ERkZydSpUyXjXhAEUlJSpDQTMpmMuLg4Nm7ciNls5q677nJbtRaPczqdJCQkYLFY0Ov19PT0kJSURFdXF93d3SQnJ+Pl5SWNqc1mo7GxkdjYWDo6OvDx8SEkJIS2tjZJ7E6cq+K4uVwumpqaUCgU/a6tIAgMHz6cQYMGSWOqUCjQ6/WkpKT08ymzWq0kJyfj6+vLpk2byMzMZMmSJdI9ISYmpl9aDbPZzKuvvsrtt99OZmYmXV1daLVaIiMjCQoKcpu7NpuNjo4O5HI5BoOBmJgY6urqSExMxN/fH71eT1NTEzExMQQFBdHZ2Ul3dzfR0dGo1Wo6Ojowm82EhYXR3d2Nj48PcrkcnU5HfHy89Jsgk8kICQmhvr6egIAAoqOjrxl/nx/UCk1LSwsqlYqwsDDph0QUJRIEgfXr12OxWNxWQvR6PZs3bwYuLk+K2XHh4s1k3Lhx6HQ6Ro4c+aUiTOPGjaOlpUXyteiNj49Pv0RkMTExTJs2rd9fUlLSJeuprKxkzZo1qNVq1Go177//PpMnT5Yc1eLj4+nu7mbixIn4+PhIT1Dd3d1uUVcBAQEkJSXh6+tLQkICWq2WU6dOUVtby+rVq9Hr9eTm5krHFxYWSsvPTqeTY8eO8emnn7q1zeVykZWVxaZNm9z+Tp482W+sJk2aRFNTE+PHj0cul3P8+HH0ej3vvPMOnZ2d5OTkYLFYkMlk+Pn5uW3X2Ww2du3a5VbH5s2b3UTYxB/jTz/9FLPZTGRkJJ988glpaWmSj0lcXBx6vZ5x48YREhIiOX/3nQcqlYrRo0djtVoZNmwYOp2O/Px86uvr+fe//y3J1oucPXtWemITlWH/+c9/9jNi8vLy+o3VoUOH3Jb5RfG51tZWxo8fj0qlIj8/n87OTj7++GNJl0Rsu5h992ppb2+XfuBFddPVq1dz8uRJhg8fLv1I1tXVXXGZokFXXFzMoUOHeP3119Hr9bzxxhvo9Xpef/11dDodjz32GFqtlpKSEs6fP8+mTZtobW3lt7/9LefOnXMrs6WlRUoK+eqrr1JbW8v69evp6Ojg6NGjdHR0sGXLFurq6njsscfo7Ozkvffew2QyUV9fP6BQ3rlz56SVzMrKSvLz83nyyScxmUysX78ejUbDk08+2S8dwpkzZzhy5AilpaW0trby1ltv0dHRwZ/+9CdaW1t55513eOuttzh58iS/+tWvKC4u5pVXXuHChQvs2LGD+vp6/vSnP/VLEdDR0cFDDz1EdnY277zzDu+99x6nT5/mpZdewmg0smHDBlpaWvj1r39Nc3OztIr8xRdf8Oijj7J161Z2797NM888g91uZ+PGjbS0tPDJJ5+QnZ3NBx98gNlsJjs7u9+KcWlpKXq9HpfLhVar5fTp06xcuZKKigo++eQTqqqqeP/996msrOT8+fOSbk7vMrZv305+fj4bNmwgLy+PdevW8dxzz7F161aOHDlCbm4uX3zxBW+99ZbbeAqCwCuvvEJdXR3Hjh3jww8/pKCggOPHj1NcXEx7ezvbt2+noaGBP/7xjxw/fpwNGzZw4cIFWlpa+q0miQKC7733Hm1tbfz0pz/l2LFjvPLKKwM+SPZux+zZs6murua+++67bF4vh8NBSUkJJpOJs2fPSkbec889R3Nzs9ux1dXVPPjggxQWFvLSSy+xceNGsrKy+Mc//kFnZydbt26lpqaGxx9/nNraWv7xj3/Q1tbGX//6VzQaDevWrUOj0VBaWkpWVhZr1qzBYDDw9NNP09DQwM9+9jM2b97MoUOHWL9+PW1tbaxYseKy7f++8YMyaM6ePcvIkSNRKpVYrVZyc3ORy+WcOXOGnJwcVCoVBoOB7du3U1BQIIlzDRkyBLi4cmCxWKTyxDDsWbNmuT1Bif4tvZcZBeFihtvJkyf387GBixO/r89FfX09e/fu7fdXUVExYP+am5sZNWoU9913H8eOHcNut9PT00NiYiIFBQWMHj0aPz8/zGYzUVFRnDhxgoyMDLq6uvDz83OTwLZareh0OsaNG4dMJkOlUjFv3jwMBgNjx44lLS2NKVOmuNUt/vgpFAqGDRvWzy9DLpczbdo0lixZ4vY3YcIEt+MEQaCiooKZM2dKcvCiTLlKpWLUqFHMnTsXX19fBEHAbrfj4+Mjne/l5cW8efPc6rjpppsYNmyYdEx3dzcBAQE8/vjj5OXlSVoPw4YN49y5c6SmphISEiKJg506dYpRo0ah0WhQq9Vu80B8mp4+fbo0VvPnz8doNDJmzBiGDRvmNlai4in8P9+WgeTHx44d22+sZsyY0W+uVVdXuzkgjxw5klGjRmE2mxk3bhyzZs2SwpVtNtuA8+/LCAwMlNqsVqt55plnmDlzJiEhIZIyq8ViueqwaJlMxvjx41m2bBkNDQ2oVCpuueUWDAYDVVVVhIeHM23aNHJyctBqtSQkJHDgwAFMJhMzZ87sZ9wFBARIK4Oib5vL5eK9994jLS2NgoICLly4gEwmY8qUKZw5c4b29nZGjRo14EOJqHmSlJSEXC4nLS2NZcuWYTKZqKqq4tixY5jNZmbNmtVvBcDLy4uNGzficDiIiIhg8eLFyGQyWlpa0Ov1REZGkpKSwk033YTNZuP6669n0KBBVFRUsGPHDsxmM1OnTu23dRkaGoqfnx8zZsxg0qRJ+Pn5sWjRIlpbWykpKaGwsBC73S59f4KDg0lOTiYpKYnIyEh+85vfMGPGDEl/58SJE4wbN47HHnuMzMxMNBoNa9euJTMz061ehUJBaGgoiYmJ+Pj4EBwczOLFi0lLS6OhoUHayu/s7KStrY3w8HDi4uL6yVNMmjRJerBqaGggPj6exMREnn76abKysjAajSQlJREfH+92fVUqFUqlEkEQJH0Wb29v1q1bh1wux8vLi507d2IymZg2bRoWi4UvvviCSZMmkZGR0e8hUqFQSKszoaGhqNVq5s6dy8iRIy97k1coFMycOZM33niDm2666bJbpn5+ftL82bZtG4MHD2b06NHExMRIubbEuRsVFYWvry+zZ89m6NChxMbGsnDhQqqqqjhx4gS1tbUoFArpmt9444309PRIvo7Z2dkUFxeTmpoqfQ9EzabQ0FACAgK49dZbmTx5MidOnMBisTBr1qwB04d8X/lBbDkJgkBdXR1ZWVkkJydz+PBhjh8/zuTJk6mvryc/P1/KvyEIAiUlJUyfPp0PPviApKQk6YudkpJCQ0ODtHqTn59PS0sLZrNZEkwSWb16NSNGjGDevHlS/dnZ2QwePFjaXuhNW1tbv5WX1NRUBg8e3K8/ohx4X0TFy4SEBG666SZ8fX2ZN28eZ8+eZd++fbz00kt0dHQwdOhQ6cstl8tRKBSkpKTQ3NxMamoqNpuNw4cPS0v4YWFhkmhWSUkJy5YtQ6FQEBwcTH5+PhqNRtqHDg4Odrt59233l0VICYIg7WtPnz5d2tYIDw+nuLiYuLg4lEql2/ZSR0cHEydOdKvnyxxyTSaTpJ66cOFC1Go1CxYs4Pz582zdupX77rsPvV4vLedaLBbphzk5OZmSkhIpFDonJ4fu7m7Jx8bf319aeVi6dCkKhYKQkBBKSkqor6/n9OnTGAwGgoKCmD59+iXHSqVS9fsR7ktjYyNZWVnExMRgsVjw9fWV8tb4+flJfjPiWLW0tAy4XfllJCUloVarMZvNKJVKzpw5Q25uLs8995y0WtXU1MTMmTOvePm699abeE5nZyebN2/mV7/6lXQNb731VlauXMnDDz+Ml5cXer2eyZMn4+PjI2Vw7l2n+P0UvyM33HADoaGhvPDCC9xyyy24XC7mzp2LXq/nxIkTtLW1SceK516qD73bKsrST58+nRkzZkjZj8XPY2JiWLlyJX/605+kVcb77ruPkJAQt773/lcst6enhzFjxhAeHn7JXFC92yOudjkcDiwWCzNnzmTq1Kn9fFj6hu27XC7q6uro7u4mMDCQ7u5uHnnkET7//HNeeeUV3n333X7f2b7XTRSK27x5M7NnzyYqKkoah4HG8/PPP2fw4MHEx8e7tQsuJnGMiYlh1qxZ0jZY760f8bdPNB7j4+N57rnneOONN3A6nVIkaWhoKOXl5dI2zJXQdywvRe8+paamXrEfklKppLGxEUDKzzVQGwC3hxaXyyWtLs2bNw+dTkdVVRV79uzhnnvukRKmvvrqq7z77ruYTCYSExMlY7739ZLL5TidToxGIzNmzBhw3n6f+UEYNHDRh+DVV1+VXk+cOBFfX1/Onj1LREQEZ86cITAwUJoEdXV10pOH6LmelpbGmTNnJD2P0aNHs2rVqgEdZe+++24p6ZtMJiM2NpYXX3wRYMAfiKqqKiZNmtTvB+5S2iADGTRpaWkkJCQgCAJqtZrOzk7Gjh2Lr68vkZGRREdHo1AouP/++wG45557cDqd+Pr6kpGRwc6dOxEEAZVKxXXXXceMGTPcFFg1Gg319fVuyd8yMzOl/d85c+ZISfHEG7x4k70aEhMTeeONNyTlU/Hp/+TJkwwaNMjtyycqd0ZFRV1VHbGxsSxevBiHw8Hw4cMxGAzSE+zu3bsZMmQIAQEB/N///R8ymYwlS5ZgtVpRq9WEhISQk5OD3W5HpVIxduxY/vWvf0nXStzOrKurIzg4WBqr4cOHS6H5U6ZMQa1WSz4oBoMBk8nkZoBcCVFRUTz//PMAkgFgs9nIz88nJibG7YYgOmmKK45Xg6+vL8uXL+fs2bOMHTuWlJQUxowZI5Xd3NxMXFwcI0aMuOIyxRtQZ2cn7e3tku/PuXPnOHz4MEajkcrKSkaOHMnQoUMZMWIEERERDB06lF//+tdkZGSwbNkytzKDgoJwOp188sknUiK/nJwcJkyYwMyZM5k2bRo7duzg2WefZcSIESxcuJDNmzfz7rvvUl1dTUxMjJufllKpJCgoiL1790r+Vh0dHdhsNvz9/YmIiOCxxx5j1KhR3HHHHW5tKS0txWQycfPNN+Pv7091dTX79+/HZDJJ4fai4rHoP2EymbDb7UybNo3HHnuMcePGcdNNN7mV29PTg8PhkGT69Xo9HR0dOBwOYmJicDgcrFixgpEjR3LHHXcQFRXFnj17CAsLQ6/Xo9frpcAAlUrF1KlT+eUvf8nkyZO55ZZb2L17t/Tk3ve3LSkpiaysLLy9vaU29/T00NXVRX19PYcOHUKr1VJUVERiYiJZWVlMnz6dxMREqYzm5mYaGxuRyWQ0NzeTkpIiKRbfcsstvPHGGxw/fnzAFAYjR45kzZo1BAcHo9frKSgoQKlUsnTpUmJiYpgyZYo0bosXL2bWrFn8+c9/Ji4uDoPBgE6nk6Ilxci23uPX2dlJT0/PZW/yLS0tvPXWWyxevBij0cjChQsvOcflcjnR0dHs3buXm2++mU8++YTCwkLkcrlbvicxik5cNRPVrsPCwrDb7WRmZvKf//yHZ599lpEjRzJ48GDOnDlDSkoKXV1dFBQU0NLSwq233kpHRwdJSUmsXbtW+k2vrq7GarVKq/jBwcHSvF2+fPkl2/994welQzMQDoeDuro66cmlqamJyspKZsyYQUJCAm1tbfj4+EjOr2VlZTidTkaNGnXZG09DQ4O0jPllNDU1UVNTw7Rp0674ZuZwONi2bRu33XbbJc8pKiqitLSUoKAgBg8efFmdG9GXw9/fX8r30RuXy0V9fb2USC0iIsLtmEOHDjFz5kwUCgUul4va2lr0ej1JSUn9ksBdLeIWXl1dHb6+vqSkpEiGzsmTJ0lISCAhIeG/qqO2tpYDBw4waNAgQkNDGTt27GXHqrKyEqPRSGZm5oBj1dDQQFdX14BjlZ2dzbhx46QtM/HY+Ph4yXH4v8FgMFBTU4NCoWDIkCGSDP6ZM2fw8fHpNw8qKiowGAyMGzfusuWKjuoxMTFuK4zi+3FxcVelUST6YYgrdyaTiaCgIGkL1GKxEBgYiI+PDyaTCV9fX2QyGXa7nZaWFkJCQvoJz4mrik6nEy8vL6kcUfbd398fg8FAV1cX0dHRUr0dHR0EBgZKEYy9V7W6u7slB2un04mfn5+k1ySTyWhrayMsLKyfMSpu24ordDqdTjKCez/li23w9/fHZDJJDvdi6oXg4OABy1Wr1VitVgRBwMfHB6PRKD2UiauPop+cxWKRUjoEBwdLofpBQUHI5XJpPP39/aWUAqL0QO+6LRYLXV1dBAYG9muzQqHAYDCgVqul7Q7R4Vk0EAVBwGw2o9VqCQoKwmAw4OXlhdPpJDg4GIVCIUUdRkVF9TOorFYr7e3t0pgoFAop9UBoaKjkcBwYGCjllmppaZEeHnrPF/HaOhwO/Pz8pDQB4ipkb6ddl8vFY489xiOPPMLgwYNpbGxEoVAQExMjyTVs2LABp9PJvffe6zYfxfGPiIjAYDDQ09NDWFhYv3lmNpsxGAzSmIqZ581mM8HBwVgsFskJWKFQSE76er2e4OBgjEYjLpeL0NBQKZmnmOpAXNn08fEhICAAq9UqbQteSyKAP3iDpjeCIHDixAkCAgJIT08f8CILgiBNustNgr5LypfDaDTi4+NzVUqt4k2+95duoGN6qxB/WVtcLhdGo/Er5faw2+0olcr/6Rfjv2lvX8SxEpdlv6y8K50HA/FtjNXl2ivqaog/fl9WDtDPiOj7ngcP1xKCcDHxqFKpJCMjw00TSwwwOHv2LCkpKVeU0d3DN4PHoPHgwYMHDx48fO/5QUU5efDgwYMHDx6uTTwGjQcPHr5VRP+B6urqb0xKX3S6PH369HdGvXYgxK1BMZy87wK60+mkvr7+G++Dw+Ggtra2n5S/GBzQV0PFg4fvAh6DxoMHD98aNptNUu79zW9+c1kp/f+Gnp4ePv74YxoaGjh//vw3UsfXgcPhYNWqVXR1dZGfny+9LxpkdrudV155hdLS0m+0HRaLheeff76fWKDT6WTbtm188skn32j9Hjx8FX4wYdsePHj4buF0Ovn3v/+Nv78/s2fPRi6X09DQgN1uZ+jQoVLUoV6vJzU11U1CQIx2sVqtuFwuhgwZIq0qKJVKBg0ahFarxW63093dTVNTEyUlJSxfvpzIyEjq6uowm82SWKModihG6Xl7e9PW1kZKSgp1dXVEREQQERFBT08PGo2GoKAgwsLCaG1tRalU0tbWxtChQ/H29qajo4POzk5iY2MJCAigvb1dKqt3RIkgXMzz09TUhI+PD/Hx8VRVVXH06FGWLFki6VKJ+k+rV6/m2WefJTg4GJ1OR1FREUOGDEGtVtPd3U1jYyPx8fFuUVFOp5PW1la8vLxobm4mOTmZ1tZWfHx8SEhIwOFw0NDQgMPhYNCgQXh5edHV1UVXV5cUHedwOKivr8fhcDB48GASEhIk/R8xWaMoCufBw7eJx6Dx4MHDt4KovySKk+l0OoqLi9m8eTNPPfUUnZ2d2O12cnNziYuL44EHHpBu1KdOneK1117jvvvuY8OGDaxYsYKGhgYGDRrEhx9+yE9/+lM2b95MYGAgISEhJCUlYbPZcDgcbNq0iaCgIFQqFR9++CEPP/ww//d//8eSJUvw9/dn06ZN/OQnPyE7O5vIyEhSU1N56623+Mtf/sIbb7zBnXfeyauvvspLL73E008/zXXXXUd1dTXp6elkZmZSVFREXFwca9eu5c4776S4uBiLxcKaNWv4/e9/L0UzGgwG/vGPf3DrrbfyySefkJmZSUJCAlar1U1JWlydaWpqwul04nK5OHv2rKQTc++99/LFF18QFhbGm2++ybvvvisZFxaLheeee47p06fT1dVFc3MzCxYs4D//+Q/vvPMO69evZ8SIEbS3t7N+/Xruv/9+Nm/ezI033khbWxuCILBz505CQ0PZsmULixYtkq5ffX09u3fvZsyYMRiNRkaOHPk/nkEePLjj2XLy4MHDt4KoHRIfH4+vry9BQUEsXLiQ9PR0amtr2bp1K2azmfT0dKKjo6XzxDxbkZGR3HHHHdx5551s3ryZvXv3YjKZmD17Nmq1mqioKMaNG8evf/1rBg0aREhICJGRkWzZsoWxY8cyc+ZMLly4gNFoJCAggOXLl3PXXXfh7+/PlClTmDJlCj4+Ptxwww10dnbidDq5/fbb0ev1kjR+cHAwEyZMYM6cOVRVVbF161bS09OZMWMGv/71r9mzZw86nY7IyEhGjx7t5iN09uxZ2tvbSUtLY9GiRaxbt47o6GgCAwNJTk52S14bGhoqtV9MIXLTTTdx4cIFjh49SnNzM97e3syYMcNNyt7Pz4/g4GDGjBnDvHnzcLlczJs3D6VSKaVWGTNmDDfccAMnTpxgzZo1xMbGMnToUBITE7FarVIahmnTprmlMfD29ubAgQOcP3/eTTjPg4dvC49B48GDh2+NvtnHe0vPGwwGBg0axNKlSxk1atQlHYa9vb2JioqSco/dcccdkhBmX00hUaa/vb0duVxOcHCwJNjXV6up92tRIO2TTz5h0KBBbpo9vdtsNpupqKiQBADNZjP+/v7ccMMNzJ492y3fk7e3t5SV3dvb+0uzxfd9X6xTLHPOnDksWbJkwPMv1T+TyeQm/udwOGhra3OrU6fTMWLECJYtW0ZiYqJUvp+fH6+//jo5OTls3boVh8MhJfv14OHbwLPl5MGDh2+NpKQkjhw5QlBQEHa7HY1Gg06nQ6fTccstt/Diiy8yadIk5s6d228VoLW1lZMnT1JXV8f999+PIAg8+uijUlLPnp4eWltbsdlsNDU1YTAYMJvN/OIXv5BWTmbNmoWfnx82m43W1lbJ4Onq6nJLdeByuejo6ODChQvs3bsXm83GqVOnsFqtUuoGk8nE8uXL+e1vf0tRURELFixg6dKlPPPMM5SVlTFp0iQ3mXwxIeaBAwfo6OjgwQcfpK2tDavVikajIS4uTjJAQkND0Wq1HD16FKPRKKWKsFqtTJw4kY0bN/Lss8+SmZnplg5CPKarqwtAUkUWjY/777+fnTt3kpqaym233UZmZibPP/+8lNepvb2dG264gRUrVjBhwgRuvPFG2tvbMRgMVFdXU1lZydKlS5HJZJw4cYLi4mIeeeSR/8HM8eChPz8IYb3eXfSomX41RKn2gfJWXen54hP2lSjxflk7rkT5+HJlXAnf1Fzp2wfxNXBV43ulY9E3QeDlruP/+rtis9kkKX2z2SylKlAoFPj7+9PR0SFtTfVuT01NDStXruSNN96QVlkcDgcajYbAwED8/PzQ6XTI5XICAgIwGo3Y7XYCAgJQqVRSeoTQ0FDsdruk1q1SqdDr9fj6+mK323G5XFLqBbVajdFolKTuVSoVdrsdLy8vKWFiYGAgOp0Oh8NBeHg4MpmM7u5urFYrERERbnNfdLjt7OxErVZLkvcWiwU/Pz98fHzcjm1vb5fSHahUKmQyGVarlaCgIMxmM3q9noiICDcVajHRpehQbbfb8fX1lVJJ+Pj40NXVhUwmIyQkBJlMhk6nk3KwiekcNBoNfn5+BAQEoNfrcblcUroAQRAICQnBZrNhtVoJDAz0/M56+Fb4QRg0ZrOZt99+m8WLF3tkqflqUvWFhYVUVFRw5513fqX66urqOH36NBcuXODRRx+9qpw/YhlwMbv09u3beeSRR76ycWW1WlmzZg16vZ4777yTDRs24OXlxb333ktPTw/r169n+fLlxMbGfqXyL4eYM+vw4cM89thjKBQKBEFg//79yOVy5s6de8Xl1NXVsXr1ap5++unLRpg0NDTw/vvv8/jjj5OVlYXVauXOO+8cMH2B2Wzmrbfe4pZbbvlKWbn/FwiCQG5uLn/729/4+9//ftWJST148HBtcs340IhPoWK6dPFfQRDw9vampaVFSjInvi/uP/d+3ftz8bOB6uqdmn2g133LG6itfdvb9/zLvb5cuy9XlsvlYv/+/fT09PQ79nLjOWzYMObNmyd91rf+S/VVZO3atSQlJXH33XdLN/Er7a/L5eLgwYN0d3cTFRXFsmXLpJWNS/X1cm3x8vIiLS2N/fv3ExQURGJiIkeOHMHLywtfX18CAwOJjo6+bDlXOgd6vyeuUAUFBXHmzBnpM7iY/X38+PFXfE3Fcs6ePevml9H72onXKDw8nPLycimzemVlZb/jTSYTX3zxBV5eXjQ2NqLT6dz63veaX67v/wtGjx7Ne++955ZTx4MHDz9srikfmmPHjpGbm8uyZcv44IMP+OUvf0l9fT3e3t7I5XJyc3P5/PPPmTt3LpMnT+bYsWNcuHCBhIQEhg4dyqZNm/D29mby5Mk4HA5qampQKpXcddddbokjNRoNBQUFWCwWZs6ciVar5cyZM1RWVpKcnMzo0aPZuHEjgwcPpry8nKeeeorAwEDpfKPRyL/+9S/8/f2xWCzMnj2b8vJyKisrufvuu9m0aRMBAQGcO3eOuXPnMmTIEDZv3oyXlxdTpkyR2qZSqViyZAlHjhxBLpczYsQIfHx8OHXqFKWlpSxdupSmpiaKi4uBi058Cxcu5LXXXuOZZ54hNTWV2tpauru7Wbx4sdtTvtVq5cCBA6hUKuLj49FoNNTX13Pvvfdy/Phx2tvbKS4uZsaMGchkMgoKCvDy8kImk/GLX/zCbbyamprIzc0lNDSUcePG8dFHH/HII4/wwQcfsGTJEmpra6msrMRutxMcHMx9991Hbm4u3d3dVFZWsnjxYl599VWeeOIJ4uPjyc7O5v/+7/8oKCigra2N5uZmZs6cicvlYtu2bcTFxdHQ0MBTTz3llhVaRCaTMXbsWJxOJ+fOnSM+Pp6ioiJqamro7Oxk3LhxlJSUUFlZSXV1NQ8//LDb9RMEgcLCQtra2qioqOC+++6jpqaGpqYmWltbmTJlCoIgsGvXLgYNGkRpaSmzZs3i2LFjLFy4kOjoaLRaLX/7299ob29nxYoVHD9+HC8vL0aOHMknn3zCkCFDKC0t5emnn0av11NQUEBpaSm33XYboaGhHDx4EIVCMaBibE9PD7t370ahUBAfH8+4ceOkbPJi1uO+HDhwgH/9618kJSWhUCg4efIk27ZtY8GCBdJ3paamhsDAQObPn092djaCIEjZxgsLCyktLeWOO+4gOTn5G91ykMlk+Pn5XVEyTQ8ePPxwuGZWaABSU1MpKCggNDRUulHrdDpJoGro0KEsXbqUzZs3s3//fsrLy4mMjGT9+vWSU9sdd9yB3W7nyJEjxMTEsH37dsmhTqSkpIS8vDxGjx6NUqmUbgQRERF0d3cTFhZGfn4+N9xwAzqdjoqKCrfzfX19KS4uZtCgQdx4442sWrWK4OBgGhoaKCwspLy8nEGDBvHggw/y9ttv43K5OH78OMuXL8dms3H06FGpbWfPnmXv3r2kpaXh6+vLqlWrJB+Jbdu24XK5uHDhAnfffTd79+7Fx8eHqKgoJk6cyMmTJ6mqqiIjI6PfDai9vZ0dO3YwZMgQKaP3mTNnsFqtvP/++8ycOZO2tjYCAgJQKBRUVFRw9913k5WVRU9Pj1tZcXFxJCQkMGXKFFJSUjh9+jS+vr5oNBo6OztxuVzU1dVx5513snv3bs6fP8/Ro0eZP38+iYmJ+Pr6Eh0dzeTJkwkODiY3N5euri5WrVolGaevvPIKKpWKgoICbr75Zi5cuEBjY+Ml54qvry/z589n06ZNlJaWsmDBArZt28b58+eJjo7m3//+N6GhoVRUVFBYWOh2bkdHB9u3b2fevHkMGzaM1tZW3n33XWbPns306dN59dVXUalUnDp1isWLF9Pa2opKpWLevHls27YNAJVKxQMPPICvry/r16/HZDJRWVmJv78/+fn5zJ8/H71eT1lZGatWrUIulyMIAp999hkfffQRCQkJXH/99QPe1Lu6uqTIny1btlz+S/P/k5SUJBn2giAwfPhwFi9ezNatW8nJySEvL4+YmBg2bdrEmTNnOHjwICNHjsTLy4v3338fhUKBw+Fg586dV1SfBw8ePHzdXDMGjUwmIyoqiqSkJHbv3i2tathsNsnZLSQkhMDAQOx2O2VlZQwaNIi5c+fy1ltv4e/vj7+/P9HR0bS3txMeHs6UKVNYtWqV29M5wIQJEwD47W9/i06nY/To0Rw9ehRvb28mTpyIUqnEx8cHf39/fH19sdlsbufL5XJ8fX0JDw/Hz8+Pzs5OMjIyeO2117j++uvx8fEhJCSElJQUAgICAAgICCAqKgqNRiO17YMPPiAjI4O0tDSeeOIJGhoaqK6uZtSoUTz00EP84he/QKVSSX2TyWTSFodMJmPBggVUVFTwpz/9CZPJ5NbG6OhopkyZwmOPPUZdXZ3koKhUKklMTCQ7O5uhQ4eSnJws1aFWq5HL5f22QHpfo4Feq1QqAgICJI2LmpoaTCYTCoWCpUuXEhISIh0vtqO7u5u2tjYUCgWDBg2SttD8/PxQq9V4e3tjt9vdtmj61n3jjTeye/duDAYD999/Pxs2bEAmk2GxWDAajWRmZvLnP/9Z2goSEbcvFQoFCxcuxM/Pj5aWFhQKBXFxcZjNZhwOB76+vlJ7xBUFcZwDAgIIDg5m6tSptLe3SytJveeOePyFCxcYNWoUjzzyCA888ABFRUUEBQWhVColZ8/eBAQEYDAYMJlM0hhcjt5OxTKZDLlcLn1XrFYrlZWVxMTEMH36dD744APGjBlDQkICK1asoKWlhZqaGtLT0/nVr37Ffffdd9m6PHjw4OGb4poxaODij/HNN9/Mpk2beOihhygqKiI1NRW46N1vs9mkv2HDhrF+/XoqKyvJycnBZDJJSqJDhgxh9+7dlJaWkpubi8FgcKunqqqKBx54gFmzZnHmzBkCAgIICQlh8ODBpKSk4HQ6cTgcuFwuqczeNxVBELDZbFLUhV6vZ9++fZw7d44LFy7gcrnQ6XQ0NTWRlpaGv78/NptNknjftWuX1Lby8nKuv/56HnroIfLy8oiIiGDDhg1UVlZSVFSE3W7H4XC4aUTIZDKam5spKSnh+eefJz4+ntraWrc+trW1kZGRwVNPPcXRo0el8YOLKy5Op5NZs2ahVqul8p1OJzabrZ/vkcvlwmq1YrPZpMiQyspKKUS3dxvtdjuDBg3i5MmT7N27l4MHD0rRKk1NTRiNRmw2GxEREfj7+1NZWUlzczOTJk1CpVK5tcPhcHDy5EkuXLgw4HxJSUlhxIgRTJw4keHDhzN48GAyMzMJDQ2lubmZrKwszp49K4moicTFxVFRUcH27ds5dOgQLpeLsLAwysrKaGtrY+zYsVLUjcvlwm63u/URLkafWCwWWlpamD9/vjQv+46lIAiEhYWxceNGzp8/T3FxsTQ/m5qa6O7u7rcidujQIcrKyhCEi9opomEjju9A+ZLE6B6NRuP2PbHb7aSmprJt2zbKy8vJycmhrKyMG2+8kXvuuYdTp04RHBzMpk2bpPZ58ODBw7fBNRflZLFYOHfuHGPHjqWwsJD09HRcLhc7duwgISEBHx8fzp07x6JFi9i3bx+dnZ3ceOON2O12jh07xrx584iNjWXv3r1UVFQwf/58RowY4ba6UF1dTW1tLSqVioyMDLZt20ZtbS02mw21Ws0dd9xBdnY2s2fPJicnh+TkZMaPHy+Vodfr2bFjBykpKUycOJGqqip27tzJyJEjmTNnDitWrGDEiBHSeV1dXWRnZzN//nyio6PZu3cvlZWVzJ8/n7CwMM6cOYOfnx9paWkIgsD69esJCwtj8eLFlJWV0djYyPjx4zl06BDz58+npKQEp9NJTEwMBoMBuVzO+PHj8fb2lvqo1WrJy8vD39+fpKQk6uvrqa2tZd68eaxatUoSDUtLS2PkyJFUV1czefJkDh48yJw5c0hISJD6q9Vq2bVrF4MHD5ba0dnZSXh4OPHx8Wi1WjQaDWPGjOHw4cMsWLCAmpoaTpw4wdy5cxk9ejSHDh3CZDKRmJhIUVERixcvxmQySasVmZmZtLe3c+LECWbPnk1WVhaZmZm0tbURGxs7YHSbIAiUlpYyePBgvL29KS0tZciQIXh5eVFSUsLevXvJzMxk1qxZbr4ngiBQVFTEwYMHmT59OhMnTqSlpYWCggICAwPJyMigra2N/Px85syZw9GjR0lJSUGhUFBZWckNN9xAdnY2CoWCmJgYRowYwdGjRzGZTIwaNYpjx465zZ2EhAQ2btxIZGQkixcvxm63s3nzZsLCwnC5XMyaNYvw8HCpfa2trWzevFnyBZo2bRq5ublMmTKF1tZWOjs7WbRokZvPlNls5tNPPyUzM1PyBVOpVJSXl7N48WIOHTpEXV0dCxcuxN/fn3PnzuHr68vIkSOxWCxs3LiR6OhobrzxRinU14MHDx7+l1xzBk3vkOQvC0/u3fW+x1zpZzqdjieffJKHHnoILy8vOjo6uO666wbcCviyNov/f+SRR/jxj3/M1KlTB2z7lV6yK+n3pY7te4z4eWFhIZ9++in33nuvtKIwadKkq7qBXarsSx1zqWt5uWvU95gr+Xyg/19t+y5X1+XOv9pjr3R+/jd1fFmZA+ExZDx48PBtcc0ZNP9rXC4XRUVFlJWVkZiYyLhx4yQp9a+CwWBg9+7dxMfHS/443yVEp+SOjg6GDx/OyJEjv3Nt9ODBgwcPPzw8Bs3XxJetBFxL/JD66sGDBw8evh9cUwaNGNHy38jif111CcJFWXNR/+NK29NbyOyrKuFeaRuvtH5ROO2/SVlwqbZdrdT/1Yzppfr/v5wnl2oT/D+D0GMYevDgwcN/zzUT5SQ6eL7wwgsDRnF83bS1tfHCCy/Q3d3d7zO73c6+ffvYv38/77777iWzBA+E1Wrlb3/7GwUFBf9V+wRBoKmpiZUrVw4ovnYpHA4H//znPzl+/LhbWf/4xz9oaWn5r9okluV0Otm2bRubNm264vPEMT1w4ADvvPPOFY1pT08Pv/vd7ygrK3Orv6qqipUrV2I0Gr9SH74qLpeLkydP8vbbb/Ppp5/y/vvvs27duv+Zuq4HDx48XMtcU84P0dHRlJWV9ZOa773SAP/vidjlcrk9pYurEOKx4gpC3ydol8tFUFAQFRUVUmht77KKi4spLS3ll7/8JW1tbW4JCHuX2XflQxAElEolPT09dHR0XLKfvc/r3a++qypiG8VQ4b7n9+672DalUonJZKK1tRWn0ym17a677iI0NLRfP/o60vYdy95j53Q62bNnDzNnzkSpVEpy/KL2yaVWggRB4Ny5c5SUlPDoo4+Snp5+RWMaGBhIZ2dnv7DmmJgYysrKpPDygeZJ77HsPVZ92zXQOIr0fT87O5v33nuPv//974SHh1NbW8tnn3024Fz04MGDBw9XxzVj0Ig3RfEmVFhYSGNjI62trYwbN44hQ4Zw9OhRADIzM2loaMDlctHZ2cnChQs5evQotbW1qNVqli5dyoEDB/D29mbQoEFuIb89PT3s2bMHf39/Ojs7ASguLqayspKKigoefPBB9u/fL4ViHzt2jJUrV7J161YEQZBCqOfPny/V6e/vz2233SYlKCwvL2fixIkD9tPhcLB161bpZj1nzhzeeecdRowYwalTp1ixYgV+fn7s3bsXX1/ffirHgiCQn5+PRqOR1Hm1Wi2lpaXU1NQwd+5cFAqFZJQNHTqURYsW8dFHH7F8+XL8/PzIy8vj3LlzzJs3j9OnT6NQKLjrrrv49NNPWbBgAWfOnKG6uprY2FiWLl0q3aSbmpr4wx/+ICWmrK2t5c0338Tb25tf/OIX5OXlUV9fT1tbG7/4xS+kMHKHwyEpO+/cuZOjR4/y0ksvcejQIRoaGjCbzcTExPCjH/2IrVu3EhQUhMFg4Pbbb3dLwdB7nohtElMcaLVa0tLSyMjI4ODBg6hUKoYMGUJ3dzdms5nOzk6WLl0qlScIAseOHUOn01FfX8+cOXMwm83s2LGD2NhYWlpaePrpp6U+uFwuVq1axcKFC4mIiAAu6tncc889OBwO/vGPf3DzzTczePDgrzL9PXjw4OEHzzWz5dQbg8HA22+/zfXXX8/cuXN5/fXXKS8v58CBA4wYMQJfX1+2b9+OyWRi5MiRHD9+nIKCAqKjo9m5cycXLlzg888/Z8iQIW7J7wRBYMOGDYSEhDBnzhyCg4PR6/V8+OGHUuoCUcwvLS2NBQsWSPL74irB7bffzmeffUZ2drZU544dO8jNzeXUqVPMnz+fIUOGXLJvov7L0KFD2b59OyqViuLiYqZMmUJ4eDj5+fmsXbuWyMhIZs+e3U/lGC7qlKSmplJTU8OZM2f48MMPmTFjBrNmzcJsNgOQmJjIvffey44dO/D29ubChQv09PTw0UcfYbVaUavV7Nixg3HjxnH8+HFMJhNhYWHk5eVRXl4uye6L5QGEh4cTHR3NxIkTUSgUhIWF8bOf/YwjR45QWlrK1q1biYyMJC8vz03oT6lUMnToUIYNG8bChQtpamqSxAu7urr48Y9/zK5duzAYDNjtdoYMGcK2bdv6KTT3xWKx8Le//Y3rrruOBQsW8Ic//IHz58+ze/duhg4dSmBgIHv27EGr1ZKenu52rkajYc2aNcyfP5+JEyfy+uuv4+3tTUlJCcuXL6eiosJti87pdNLY2EhMTIyk2Lxr1y6efPJJuru7Wbp0KTExMZdtrwcPHjx4uDTXrEHT3NyMXC4nKioKp9NJQkICqamprFixgtbWVu677z7WrFnDunXrqKioIDY2llmzZvHee+8xePBgpk2bxmOPPUZNTY1b2adPn5byF4nqqj09PYwbN44//vGPTJkypZ+zp0wmk+T9/fz8cDgcVFVVERcXJ9XZ2NgorRyIT/UDyfaLq1AajUba4vHy8kKtVqNWq7FYLJw5c4bAwEAUCoW0GtKbkJAQqqurkcvl9PT0cOHCBVQqFenp6YwePRqA4OBg1Gq1VIdKpUIQBCorK0lPT+f+++/n8ccfZ+TIkchkMnbv3s2IESMkUbb58+fz1ltvDVi/SEBAAL6+vsjlchobG/Hy8mLChAm89957JCYmDhhN1XvrR6lUSikCem/Bifmhvsw3xWKx0NDQgFwuJyIiArlcTmhoKBMmTODxxx+nvr6eu+++m88++4wPP/zQzTdLq9XS0dGBXC5n0KBB6PV6KfWCn58fXl5eblt9SqWSpKQkqqqqEARBStlgMBgkleneQncePHjw4OHquGYMGjECxmazERgYSHx8PGfPnpUk/LVaLXPmzOH++++Xtjb+/ve/09bWRkJCAps3b+bs2bMcP36c6upqRo8ezZNPPsmJEyfc6hk9ejQ7duygvr6erq4uXC4XXV1d7Nmzh8LCQmpqatxk7EX5+N7/2mw2hg4d6lZnQkIC+fn5VFVVodFoaG9vp6ioiD179rjdmBsaGtiyZQteXl4YjUa6urokWXur1YrD4WDkyJHs2LGDhoYGurq60Gq10vlOp5P3338fpVKJ2WzGZDLh5+fH+++/z9GjRyW/IKvVit1u7/fvoEGD+PjjjyktLeXUqVMoFAqWLFlCdna2tD23du1aSkpKOH78uNsqiWhwNDQ0uKWasFqtxMTEcOrUKU6ePElOTo6bD5GYKqLvmPb+s9lsnD9/np07d6JUKjEYDHR3d0spF3r7GYkpAHx9fRk6dCiFhYVoNBqGDx+OxWJhwoQJPProo1LyzjfffBOTyYROp5PaFBcXh7e3N1VVVTQ2NjJ16lQUCkW/a9HbKHv00Uc5cOAARUVFGI1GOjs78fX1RRAEdu/eTXt7+9f4jfDgwYOHHxbXTNi2IAjU1NRw/Phx5syZg1KpJC8vj8DAQNLT07FarZw7dw4fHx9GjBjBuXPnUKlU0uu9e/dSX1/PDTfcQEhICIWFhfj5+TFkyBAiIyOleoxGI5999hk+Pj7I5XKmTJlCd3c3e/fuJT09nenTp5OTk4Ner2f06NEcOXKEqVOnUllZiY+PDxEREZw5c4YFCxZw/Phxqc6kpCSysrKor68nPDycpKQkKWXDs88+K4nXiTLz8fHxtLW1kZyczIULF5g8eTLnz59HrVaTmZnJ9u3bpWSRU6dOlfogCALHjx+ntraW+Ph4jEYjGRkZfPbZZwwZMoSZM2eSlZWFj48P8fHx5ObmMm3aNPLy8khNTSU5OZktW7bg5+fHjTfeSGBgIAaDgcbGRtLS0rDb7ezcuZOOjg4WLlzolgJBEAT2798PgLe3NzqdjjFjxpCVlcWsWbNoamqS6hs7dqybU+7x48fR6XRkZmaSlZXF9OnTaWhoQK/Xk5aWxokTJ5g5cyYHDhwgJSWFxsZGRo8eLWUuHz9+vOR43NjYSFZWFjNnzkStVpOTk4O/vz8jR44EoKioCD8/P1JTU6U8UF5eXmRmZro5Djc0NFBcXCylXmhubqawsJAZM2Zw9OhRxo4dy9ChQ93639DQQE5ODjKZDF9fX0aNGkViYiKff/45kyZNIioq6pv+qnjw4MHDNck1Y9D05XLS9H25ks+vtOyvmgJgoDZotVq0Wi0pKSkDRtFcrs6rlca/XFmXKrtvm65Glv9q23wlfBXJ/0udNxBXkx7jSuu71Bh68PDf4HK5aGxspKuri7CwMIKCgqipqcHPz4/k5GTgYlZ7lUpFYmLigHpQVquVCxcuYLfb8fLyIiwsjPDw8P9aH8uDh2+Ka3Zm9g2z7f2679+VfH6lZX+dbQwJCXEzZi7Vzisp+1KfXUlZlzr/y15fru6rbfPVtOur9ufL/r6u9g50nidk28PXiUwmQ6fT8cADD6DX65HL5fztb3/j008/lebavn37qK2tveS8k8vlbNmyhbfffhutVstzzz3Hrl27EAQBs9mMVqulu7ub7u5u7HY7giDQ1dVFV1cXFosFh8NBd3c3RqMRQbiY+d1oNErnitvRdrud7u5uqbyenh4EQcBoNKLRaDAYDB6tJg9XxDVr0FwLeG5yHjx4+CrIZDIiIyPx9fUlMjIStVrNz3/+cyki0el0ArgFMfRFpVIRGRlJeHg4U6ZMYfz48ezevRuA8vJyHnnkEc6dO8eBAwcoLy/nzJkzbNmyhcOHD/PFF19gsVhYsWIFp0+fxuVysXPnTs6cOUNNTQ0///nPJZ+0oqIiPvvsM55//nk2bdrERx99RHt7O++88w5lZWW8//77Uns9eLgc14wOjQcPHjx4GBiZTMaYMWMICgoiKyuLwYMHk5CQcNkoRJGmpiY2bNjA6dOnue+++wCIiIhAqVSSmZlJVFQUsbGxHDp0iPLycp577jkMBoMU7RcREYFCoWDIkCHExMTg7e2NUqkkKCgIQNJ/qqioID09nREjRmA0GsnLy2Pp0qUsW7bM82Dn4YrwrNB48ODBww8AlUrFfffdx8cff8yxY8eYNGnSFRkKvr6+ZGVlMWjQILdztFotR48eZceOHcjlcubMmUNcXBy/+MUvriqtSEBAACqVSnodGBhIVFQUDz30EM8++ywnT570GDQergiPQePBgwcP1yh9HdBnzZpFV1cXra2tREREIAgC3d3dl8yNJggCISEhvPzyy2RnZ0s+NABhYWHMmzePBx98kM7OTmpra/nVr37Fvffey7/+9S/gojFktVoRBEFyLu6Ny+UaMEdcW1sb48aN46233mLjxo0D5szz4KEvni0nDx48eLjGcLlcVFZW4uvry/nz5wkNDUUul6NWq3n44YcZPHgwcrkck8nEyy+/zG9/+1vCwsLcyhBzuun1ery8vHj99df53e9+R2JioiRceezYMRobGwkPD8dgMFBZWYnRaGTu3LnIZDJuvfVWSWPJbrcTFhZGRUUFgiBw5MgR2tvbCQgIYM6cOdhsNtrb23G5XOj1evbt28eoUaOYOXMm/v7+39JIevg+cc2GbXvw4MHDDxVBELBarZKCtre3t7Rt43A4JMVxQRAkcc2+2zoulwuLxQJc1GFSKBRYLBZJ0by3EraXl5dbnQEBAZIUhdFoxG63ExgYiFwul4Qte58rl8ulsr29vXG5XJLAp7+/PwqFwrPt5OFL8Rg0Hjx48ODBg4fvPR4fGg9XzLVg+w6UH+ta4FrskwcPHjxcDR6D5geA6JDX0dGBXq/HYrFc9Q3QarVy5MiR75TIlSAIuFwu6e/L2iUIAkVFRdTX139n+tAX0eDq3afLtVUQBHQ6HceOHXNbxvfgwYOHHxqeLaevwH8r0X+5Mr/ufWIxd1F2djbDhw/HYDBQUlLCQw89dFkJ897tEQSB4uJiaV97woQJbqka/pd7273r1Ol05OXlSdmrJ0+ejFqtvuS5bW1t1NTUYLfbGT9+vJQYUizv26LvV7CoqEhKVBkfH8/w4cMv2T6Xy0Vubi4+Pj74+PiQlpb2rV0bDx48ePg2+cEbNIIg0NHRgcViQaFQEBkZKSWCvBRWq5XKykopmWF7e7vkvBYQEEBraytyuZyoqCgUCoVUT3d3NwaDAS8vLyIjIxEEgba2NgCCgoKorq5mxIgRXzlXitPppK2tDZfLJQlamc1mXnvtNR577DHpdXFxMRMnTrzkzU4QBDQaDVarFbVajclkwtvbm4iICBwOBxqNBoVCgb+/Pw0NDdJN9Gqx2WxoNBoA1Go1wcHBly1HEAQuXLhAaGiolHE7ICCAwMBAKULCz8+P0NBQtxxJYuZthUJBVFQUMpmM9vZ2HA4HYWFhnD9/nhEjRnzpdR+Igcb8asfCbDZTXV3NoEGD0Gq10tyRy+V0dHRgs9mIjIx0C3l1Op1oNBocDgdBQUEEBARgMpno6urC398fo9GIQqEgOjr6Kxs1YmZylUqFy+VyC7t1Op2SeJpKpcLhcKBUKj0GlAcPHr41PFtOIOkniDemyyGGG4oql4Ig0NTUxP33309NTQ0Wi4UPPviAqqqqfmX19PTw6KOPcvLkSelmsXbtWioqKvDy8kKn01FeXv6Vt0NcLhf79u3jD3/4A3q9XspALmb51mq15OXlERsbe9lyrFYrx44dIzIyks7OTh566CFKSkqAixESH3/8MbW1tfj6+tLc3Extbe2AbRbDLy+Fw+Fg1apVrF279oqEuNrb26moqCAwMJDz58/zk5/8hM7OTmQyGVqtlo8++ojOzs5+55lMJl566SV27NgBXLxmn332GYWFhVJ6ifz8/AH7IAiClFtmIARBICsri9deew2dTnfV104QBA4fPkxYWBhWq5U//elPfPrpp1I5e/bsIScnxy2iRDyvqKiIhx56CIPBAEBrayuffPIJOp2OiIgIjh07JkWpXC0ul4tTp07R1taG1Wrl008/5Z///CdwMffOv/71L86cOcP777+PTqfj7Nmz3+mtPA8ePFz7eAwawGKxMHToUDIzM79UClyn01FfX09cXBwymQy5XE5GRgZTpkyhrKyMoqIifv7znzNjxgw39UuZTEZiYiILFiygtLQUgJKSEm699VZmzpyJUqkkIyODo0ePfuWbglKppL29neuvv56UlBTkcjkhISEolUqsVitKpZJt27bh6+t72XJOnz5NdHQ03t7epKamMnPmTM6fP48gCJw9e5Z7772XSZMmoVAoyMzM5NChQwOWY7fb2b9//yXr8fHxQaPRcMMNNxAfH/+lqzOHDx8mPT0dpVLJlClTiImJoa2tDZvNRkVFBb/+9a9JTU3tl/QxMjKSW2+9VTLKzp8/z/Tp01m8eDHe3t6kpaVRUFBwSR+UvXv3XjKXjEKhoKOjg+uuu44hQ4Zc9epaZ2cnGo2GqKgoQkNDuf322zl37hwymYzq6moyMjJYtmwZfn5+bucplUquv/56fH196e7uxmw2U19fz+OPP86gQYPw8vIiLi6OgoKCq2qPyJkzZ2hqaiIuLg6VSkVYWJi0mnb69Glqa2uZMWMGvr6+7Ny5k/T0dHbt2nVZA9aDBw8evkk8Bg2Ql5fHhAkT3N4TnTH1ej09PT3Y7XZcLhcXLlwgPDzc7aYpl8u54447+PDDD4mOjiYmJuaS2ZkXLFjAiRMnKCgowNvbm+TkZOlYPz8/DAaD9MQNF42CsrIyiouL3f7Ky8v7PbU7nU5KSkrIzMyUyoyJiWHJkiWcOHGCiooK7r//frctmb6IRovYLplMxk033cS+ffs4deoUoaGhkjEHEBwcTHt7+4ArAaIz8qXQ6/V0dHSQnJzcrx9GoxGTySSpjDqdTmpqaoiKigIu3tCXLFnCli1byM/PZ+TIkQQGBl5y3KdOnUp1dTX5+fkYjUY3vxSlUomXl5e0/deXy/XB5XJx5swZxo4d269uUYOj70qd2D9BECgvL5fmi5hvp6uri5ycHDo6Ohg9evQlr5W3tzfz589n27Zt5OfnM2bMGHx9faWyUlJSOHv27FUbyC6Xi40bNzJq1CjJaO9tnBcWFkpbWYmJieTl5aFSqfD29v6vDHIPHjx4+G/4wRs0TqeT4uJiyQiw2+1YLBbMZjOfffYZRUVFVFVV8c4772AymWhrayMkJMTtJuNyuTCbzfj7+w+45dGbuLg4goKCaG1t7efsKZPJ8PLyoqenx+0cl8uFw+Fw+3M6nf1uHF1dXVgsFmlLyWw243K5GDlyJNdffz3jx48nMzMTuVyO0+mkvb2938qDIAiSH4bIkCFDcDqdmEwmhgwZ0s+YAySDRhAEKisryc7Oloyo7OxssrOzpSd8kZqaGiIiIggICJBu/mLUziuvvEJxcTE7d+4kJycHu92O3W6XbqwymYzp06dz5swZYmNj+xmZfQkODiYtLY2zZ88yZswYt5UUmUyGWq2mq6tLeq+lpUVqd0VFBcePHyc7O5uamhq3ce/u7kav15OQkCCNuWhoOp1O3nnnnX6y7TqdjrfffhuXy4VGoyE4OFj6TK1WM27cOHJzcxk3btxlV3xkMhmzZ8/m2LFjDB06lKCgILcx8Pf3p6ur66oNDIfDQXV1NYGBgQN+brPZJH8ZpVIpGXwxMTEUFxdfVV0ePHjw8HXxgzdotFotZrOZuLg4BEGgpKSE7u5uVq9eTXR0NNOnT2fMmDGkp6fj6+uLQqFwe+IWzwkMDOShhx5i/fr10g1EXFnofUMRVTqnTJlyyZtV7/ddLhdtbW20tra6/Wk0mn5P/uXl5SQlJeHj44PL5SIvLw+n0yk9sYt/cPFm+9577/VbWRGfyHu32Waz4e/vz7hx4y65AtL7/eDgYGJjY4mNjSUkJITY2Fji4uLctk0EQeDUqVNSmXa7XfJjEZ19x40bh1qtpra21s3JV6Snp4ehQ4eSlJTkVv9A4y4anddff73kqN37eEEQ3MZdrVZL7Rb7EBsb63aTF403sW/imDscDtrb2zl//jx2u12aI1VVVWg0GgwGA+Hh4ZJaa9/51NPTw5w5c9xWRcQ+9DVOtFotY8aMuaTzb9++XgmioXIpf7Jhw4ah1WoRBIHOzk6GDh0qte+rOFZ78ODBw9fBD9qgMZvNZGVl4e/vT1NTE/v372ffvn34+flRXFzMmDFjsNlsFBUVERMTg1wuJy4ujvb2dilqaePGjZSVlZGWlsaIESMoLCyksLBQikJ55ZVXsNlswMWti1OnTl3SUVMQBGw2m+RwDBcz5GZmZjJ+/Hi3v4yMDOmGJ66q7N27l7i4OM6fP8+aNWvQaDSoVCpqa2ullSYxUqioqAhvb29MJhO5ubl0dHRQXl4OQGRkJFqtFrjoIHzy5ElpVacvosEkGisymYyIiAhSUlJISkoiMjKSlJQUkpOTpVUfQRBobW3l6NGjhIeHU1ZWxjvvvIOXlxcymYyKigoUCgUFBQVERERw6623olKp8PHxwWq1Ahe3q/bv309kZGS/LaGOjg5efvll6Vin00lpaSnNzc2SRHtf9Ho9ERER0uvAwECp3ZGRkSQnJ5OSkkJYWJhkOHR3d0tjXllZybp162hqasJsNrN582bsdjshISG0t7ezadMmFAoFGzZsoLy8nGHDhgEXw7LF+STm36mqqsLPz8+tnYIg8Mknn3DkyBHpdXd3N/v37ycmJmbAbTGtVktkZORVRx4plUpGjx5Na2urNH7t7e0YDAZsNhuTJ0/GYrFQX1/P+fPnufnmm4GLTsl9t249ePDg4X/FD/pxShAERo4cyfDhwxEEgZiYGEaPHo23tzfh4eHSNtL+/fuZPn06ACkpKZw8eVJ6eh05cqRkgAQGBvLWW29JN+7Q0FCmT5+Ow+HA29sbQRCIiorixRdfHHB1pru7m6ioKLeVDLlcTmho6BX1Zfny5ZKRM2bMGJKTk2ltbeWLL75g6tSpZGdn4+XlxeHDh8nIyCAxMRGtVktWVhaJiYk0NTWRlpbGhAkTqKysJCkpCUEQGDRoEM8888yAN0aNRkNSUlK/1QSx7b23U/q2d8WKFfj4+CAIAnPnzmXo0KEIgsDp06dZtGiRm/+IeK3q6+sZNmwYLpeL2bNn91tNgosrRNdddx12ux0fHx/g4orL66+/Lr3ujcViQaVSER4ePmBbLxVOLggCt956K0qlEkEQSE9PJykpiZqaGkJDQ2lubiY6Ohq1Wk1AQAA2mw21Ws2FCxeYPHkygiAwdOhQNz8Xb29vXn/9dQICAvrVN3fuXBoaGqTXLpeLW265ZcCxFwSBqqoqSTPoapDJZNx9993k5OSQnp6OIAhMmDCBjIwMnE4nQUFBPPHEEzQ2NnL//fcTGxsr+X1NnjzZE7rtwYOHb4UfvA7NQAiCQH19PaWlpYSEhOBwOBgzZgxqtRpBEMjPz8ff3/+ygmfw/xx6R40a9aXRLy6Xi8OHD5OWlkZcXNzX1pfCwkLq6uok/5OgoCAsFgudnZ2EhYWRkZHBjh07mDt3LgEBAURHR+NwONizZw/XXXfdJR1t4eKT+759+5gwYcKAxoC49XM12xCdnZ1SmYMHD3arW6/Xc/jwYRYsWHDZaDSHw0FpaekVjbsgCOTm5hISEsLQoUMHdOy9Wo0Vk8nE6dOnpW0sMcR67NixpKenU15ejlqtJi0tDYATJ04QGRnZzz+pbzsaGxvx9vYmMjLyS/uk1+vJyspi0aJFX2kbSBAEqqurUalUJCYmfumxxcXFJCQkfKmW0NXicrkoLy/HarXidDqlLbqkpKRLGqDfBC0tLXR1dUnaU18Fq9VKcXExaWlpXzl7tLglWVJSQnBwMJGRkZw7d46kpCQGDRr0tYy9qN1UUlKCWq0mPT29X7kOh4OysjKioqK+dD7+t23R6XRUVlYyZswYt7lst9spLy8nLCzsS6UoLofVaqWkpITU1NRL+o193bhcLurq6rBYLF9Zx+tKEKU7BEEgJSXlmn/Y+EFvOV2K3iHWkyZNYtq0aZICrUwmY9y4cQM+QfdFqVQycuTIK5pEdrudtLS0/+qLORAjR44kOTmZYcOGkZqayqRJk4iNjZVWEyIjI5k8ebIk9gcX/S5mz54tbdlcCpvNxpgxYwgLCxvwc9EX42oICwvjzjvvHPDm7u/vz9SpU7+0XQqF4orH3el0EhcXN6AxI/ZBpVJd1Q+Br68vU6dOZdq0aUyePJmmpiYmTpzI5MmTJV8k8UdMJpMxadKkAVeO+hIbG+u2LXY5rFYrs2fP/ko+NIAUJSU6O38Z6enpX7sxAxfn2MaNGzGZTKxcuZKCggIqKio4ceLE11rPl9HT03PJKLgvw+Fw0NzcjN1u5/e//72kAv1VEASB999/H7Vazfnz53nvvfcIDAx0W7kbCKfTSXNz8xU5iAuCwKpVq/Dx8aGurq7fZ01NTbhcLj744IP/iRN4V1cXr732mrR137sta9asIT8//78q3+Fw8Kc//Ynm5uZ+5ev1+n5O/V8XOTk5rF279hspW9zWt9vt7Nu3j507d34j9XzX+EFvOV2Oy/0wKxSKK/qhl8lkV3xD8fb2/lpXZnqXm5mZ6fZe39djxoxxey1G/VwujQBcvHF/mabN14lMJruip/KrGXelUnnFN+0rpe/cWbZsWb9otqttw9X0SfRj+m+5UuPkm3zqUyqV/PznPycyMhK1Ws2QIUOYPn06ra2tnD17FpPJxPDhw/utJAqCQEtLC1VVVSQnJ+Pl5UVFRQVDhgyhpaWFkSNHYrPZqK2txeVyMWrUKJqbmzGbzWg0GknxuaWlhSlTpkiRjWazmYqKCnx8fGhtbWXq1KkoFAoqKyvRaDSMGTPG7WFHEAR2797N0aNHefLJJ1Gr1dTX11NZWcmUKVNQq9XU1dVRX1/PyJEj3Xy0AEkqorm5mYyMDDo6Ojh8+DAzZswgKiqKDz/8kFtuuYVJkyZRVlZGZ2cnY8eOxdfXl5qaGtrb2xk8eDBnzpxh+/btPP/880RHR7u1T6/XU1VVhdPpJD09ndbWVg4ePMiECROYP3++27ZvUVERf//733n55ZcJCQmRVlQzMjKIjIyksbGR6upqhg4d6uaoLkYxVlZW4ufnR3NzM6NGjeLs2bPSg1xHRwd1dXWoVCppJez8+fN0dXVJ5Wi1WsrKyggODmb48OGS4npvxO1WuLhdXF1dTXp6OtXV1cjlcurq6pg8eTLBwcHU19fT0dEhld/V1UVDQwMKhYLU1FT+8Ic/MGzYMG6//XY6Ozu5cOECqamp/aQ5RNX3hoYGQkND8fHxoaamhszMTHQ6HTabjdjYWCorKzGZTKSnpxMdHS1pkrW0tEjb/AkJCf3K1mg01NfX4+Pjw/Dhw9FoNHR3d9PZ2Ul0dHS/B0CNRsOKFStYsWKFpIaelZVFfHw8Q4YMoaenh9LSUgICAv4rdfrvGtdGLzx4+I4iGiJyufyaX+79JlAqlW43YLj4QLF161YMBgMOh4Nnn32236pdY2MjO3bswGaz8dhjj+F0Ovnwww/ZvXs3ra2tKJVKVq9ejdVqZd26dVRUVLBnzx7eeecd7HY7TzzxBO3t7Wzbto2cnBwOHz7MBx98gNFo5KmnnqK6upr169dz8uRJDhw4QFVVFadPn+bNN9/sd4O12WyS47zdbqelpYXc3Fw2bNjAmTNnOHHiBK2trTz77LNu2lKCIHD8+HHOnj3L+fPnee2116RIy9DQUAIDA1EqlYSGhrJnzx4aGho4ceIE7777LkePHqWkpERaSXE6nTidzn4CjTabjb/85S/4+/tz4cIF/vznP7vVMZARbTabJamF+vp6mpub+ctf/kJFRQX79++np6eH3/zmN/2uSWdnJ08++SQajYYNGzawceNGDAYDb7zxBp2dnfz9738nOjqaffv2sW7dOnbu3MmFCxfw9/fHarViNBpZs2YNCoWC5557jtra2kvOm+rqalavXo1SqeSPf/wj7e3tPPfcc+Tn53P27FlefPFFysvL2bp1K1FRUej1elwuF++++y5eXl689957tLW10dPTQ0BAAHV1dXzxxRcYjUaeeOKJfoEdjY2NbN26lfDwcP7xj39gNBr5/e9/j8lkory8nO7ubtasWYPL5eL06dNkZWVJ5zY1NbF9+3bsdjuPP/54P3FKjUbD22+/TVxcHJ999hmfffYZR48e5Y033sBoNPLCCy/0W71SKpV0d3dLPoyNjY1YrVZWrlyJTqfj448/RqFQ8MILL3D+/PlLjuP3DY9B48GDh+8VZrOZQ4cOMWzYMCZOnEhjYyMtLS1ux+Tm5tLd3Y2vry8PPPAAQUFBPProo6xfv15aVVm6dCkWi4Wuri50Oh1RUVHEx8czadIkfH19SU9PJzU1lebmZmkFLSAggKCgICZMmMCIESNoaWlhz549CIJARkYGs2fPdjNoZDIZAQEBBAQEoFarUalU0vZjY2MjBw4cwG63Exsby+23396vr2LZQ4cOZdGiRQQFBeHr60tUVBTh4eH4+voSFhbG3r17EQSBiRMnMnXqVEm9efLkyTz++OMEBARIueZ609LSQmlpKcnJycyZM4djx46hVCqlOnpvGctkMvz9/d3yro0ZM4brrrsOjUbDkSNHMJlMhISEcM899/RblQwJCSEoKIixY8cybNgwoqKimDhxIu3t7Zw9exa9Xk9cXBzz589nx44dfPbZZ2RkZJCUlIRaraa6uprKykqcTic/+9nP+hlnvesSE8/6+PigUChQq9WEh4czbdo0fvazn3HhwgU+++wzBg0aRGxsrBTFunz5cjo7O+ns7MRkMknn5eTkYDQaCQwM5N577+33cHL06FEiIiJISkrid7/7HUOGDGHGjBns3buXjo4O4uPjyc3NZdiwYfzkJz9h9uzZbnNVq9Xi4+PDAw880G+15PTp09jtdmJiYpg3bx47d+4kIiKChIQEpk+fjt1u72fQ+Pj44OvrK+lzDR06lBkzZuB0OqmqqqK0tBSHw8FPf/pTAgMDrxkxTI9B48GDh+8Fol6Q6KDd2tqKIAiEhob2c7JVqVR0dnYyfvx4Ro8ejcVioa2tjfHjx/Ppp5/idDp5++23CQ0NJTw8vN8PunjDkslkX5rfzeVyIZfLmTZtGpGRkf2OF8vo+77YF6PRyOTJk0lJSekXfi+TyXA4HEydOpX4+Ph+6uDiMXa7HW9vb6ZNm0ZoaCgul4v8/Hzsdrs0TqKOUe+++vn50dPTI+Uri4mJQaVSXfYGN5AekiAIqFQqdDodEydOJC0trd9Ntm+be58bGBhIY2Oj1P/4+HjsdrvkryNqHGk0GkaNGsWUKVMkjaeBkMvlOBwOSZCzd7+dTifx8fH4+PhQXV0tlW82m/nb3/5GYmIiQUFBbserVCq0Wi0TJ06Utit74+PjQ05OjmQgW61WfvSjH7Fu3ToCAwPx9fWltbWV2tpa7Ha7JDIqCALe3t50dHRIchx9yw4JCaG+vl6Szei7JXW58b3UnNNoNIwYMYKpU6cOOKe+r3gMGg8ePHznaWxsRKFQcOHCBQB+85vfsG/fPrKysli+fHk/36rrrrsOrVbLihUryMnJob6+np6eHn71q19x/PhxysvL8fLyYuvWrchkMk6dOkVnZ6eUPd3Ly4uWlhbMZjMGg4Hm5macTictLS1SDjKTyYTRaOTHP/4x//73v3n55ZfR6/X9tmkGDRpEfX09J0+eRKlU0traSmdnJw6Hg8WLF3Ps2DGeffZZKZFsb+6++242b97MypUr0Wg0dHR0oFQqaWhooKGhAblcTltbG/fffz9vvfUWr732GlarlQcffJANGzbw4osvotfriY+Pp6Ojg8LCQrfyw8PD+elPf8qOHTs4fvw4//d//4dWq0WpVA6YbFQMADhw4AA2m42Ojg5aW1tRKBTMmDGD8vJynnrqKUpKSvoJaXZ2duLn54dGo5Fu/G1tbfj4+BAfH8/s2bPZtWsXJSUl/PKXv+Thhx/m3//+N5s2bcLf35/AwEBSU1N57LHH2Lx5M76+vphMJrRabb+bckpKCm1tbWzcuJGIiAg6OjpwOBxkZWWxf/9+fvazn3HHHXdQXl7OP//5T2QymeT0vWnTJnx8fCgoKCA9PZ0DBw6QmZlJTU0NTz75JKdPn+63OjRnzhx6enp4/PHHKS4uxsvLi8TERKZOncro0aNRq9Xce++9PPvss7zzzjuoVCo0Gg0Oh4Nx48bR09PDihUrOHHiRL9VtDFjxjBx4kR2795NZWUlDz74IFqtFpfLRWtrK97e3m4q5wBeXl4kJCTw+eef09HRgdFopKWlBS8vL9RqNWPHjuXXv/4169at66cw/n3GE7btwYOH7zxiyg9AEmC0WCy4XC78/PwGDLe32+1YrVbUarX0dKtQKLDZbCgUCreQfJfLJZWhUChwOBwDqoIrlcp+n3l5eWE2mwGkXFp922I0GvHx8ZHOFVcMvLy8sFqtOBwO1Gr1gOdaLBYEQcDX1xen0ymVAUiyCAqFApPJhFwul4wi0c9DfG0ymaQtmL519D5XrEPM4dV3NcVsNqNSqaQweplMhtPpxMvLS1oREa9Jb6dgsVylUildD1GwU6zHaDTi7e0taSuJ7VIoFKhUKlwuF0ajURor0T9JnBN9x03sq0Kh4Gc/+xkrVqxg8ODB0pjYbDZJ0sLLywubzSatuMFFXxSLxSL5P4nzqe+2kDiXrFYrfn5+knN5Tk4Oy5cvl+QbjEYjXl5eqFQqbDabNAecTicWi+WSZYvnimMj5hYU56NSqXTbHhQEQfJh6juvxRVOo9GIn58fCoXCY9B48ODBgwcP3we6urp45plnWL58Oddff/03HtWTnZ3N6tWref755/ulZvHwzeExaDx48ODBwzWNy+XCbrcjl8uvSiTzv6nP5XJdU6sf3wc8Bo0HDx48ePDg4XuPxyn4KyLua35X7MHe7RkoAuG71t7L8U2293LlXmmdX7Vd37frAP372vv1t9UPl8tFWVkZNTU19PT0cPLkycu2w2KxcOzYsQGTq36diOkivvjii35Omt9XTCYTJ06cuGRC3W8Tp9PJ2bNnaWxs/Labgslk4uTJk5hMJrf3XS4X58+fp6ys7GurSxRDPHHiBEaj8YqOr6+vp7Cw8Hvzu/NVuWYMGkEQOHv2LNu3b5eEq3bu3El7eztWq5VDhw5RUVHxtVxQUVZaVAD9tieJ0+nk5MmTUnu0Wq006Q8cOMDOnTs5fPiw9IU7dOgQO3fu5MiRIxQWFmIwGL71Poj9EBVN9+7dy5EjR6Ss318X3d3dfP755wP2t6enhx07dnzpWBQXF1NUVHTVdet0Ov7zn/+wZs2aK5qLLpeLhoYGSSTuwIEDmEwmmpub2bVrF3V1df3KMJlM7N27l127dqHT6XA4HOTk5JCdnU1jY+MVX2tBEKirq2Pfvn1uBsypU6coKytDp9Px2WefodPprnoc/lsEQWDnzp0cOnRICoEdyIjv6uqip6cHp9MphS5/kzQ3N7N161b8/Py+9nn7bWGxWPj973/fT+ztqyIIAu3t7Vd0IzYYDHR0dFy2rA0bNpCXl/e1tO2/wWaz8cc//rHfdRcEgRMnTrBly5avtT673T5gfZfi/PnzrFq16jvxO/9Ncs0YNHDRm/93v/sdDocDmUzG73//e2nP9OzZs5JWxZc9nV/qCVR8LSpKjhgxQpL+H+i8y/0NxOXKGOgYkfz8fHbs2MH48eMpKCigurqa7u5uHn/8cYxGI9OmTSMwMJBPPvkEpVLJ6dOn2bNnD8nJyTQ0NPB///d/nD9//pL1uFwuSktLBxyXL+vXpdrf91iXy8WqVavYsmUL48ePJz09ne3bt7s9fQmCQEVFhaTTcCUrLX3rUyqVhISEDHicQqG45Ge9y/Dz87ukqNflxiIwMJDS0lIiIyNJTU0d8Pze5Zw4cYKuri7CwsKoqKjg73//OwCtra3U19cPmN7A29sbrVbLSy+9hMViQSaTkZubi1qtJjAwkP3791/x03ZPTw9r1qxx649arcbX1xelUsmnn34q/aB+ldUal8tFZWUlxcXF7Ny5k6qqKsxmM6dPn+bUqVOUl5fT0tJCVlYWlZWVCIIghR6LERwKhYJBgwYBF5OXFhQUkJeXh16vZ+XKlRw4cACr1UpKSgoGg4FDhw7R0tKCVqulqKgIi8VCUVERWVlZUqSS2B+xrqysLI4fP47T6aSnp4cTJ06Qn5+PzWajrKyM4uJisrKyOHjwIBUVFcTHxxMfH8+ZM2fIzs5Go9FgNBopLCwkNzeX0tJSCgoKOH/+vKRcfOjQIek7Vl9fT35+PhcuXMBqtVJQUEBZWRl79uzBaDTicrmoqanhxIkT1NfX43K5qK6u5uDBg5c03KxWK0eOHKGhoYHCwkIKCwuxWCwUFBRIKSAOHjxIU1OT2/ni9a6srJTqt9vtlJSUcPLkSXQ6HcXFxeTl5WG32ykqKkKr1dLY2MihQ4ekxIgiBoOBp59+muzsbGw2GzU1NRw8eFDKdVVaWkpRURF6vZ6//OUvbN68GYPBIJVXW1uLIFxMMyAKzg00ryoqKjh+/DjV1dVYLBaOHz+OVqulpKSEsrIyuru7OX36NEeOHJE0YcRx6nttDh8+TElJCYJwMVFmfn6+9HAorsBUVFS49TE3N5ecnBwEQSA6OhpBuBjlVVFRQVFRUb8HAafTSWlpKcePH6e+vl4ay/Lycnbv3u2WT0omkxEUFCSlprFarRQVFZGdnU1HRweCcDFVQm5uLqdOncJutxMVFYVMJpPmb1dX1zVp3FwzBo1MJmPw4MFkZGSQnZ3NiBEjEASBgoICTCYTERERhIWFkZeXx+bNm92++EajUdK0WLVqFS0tLVitVnJzc9m8eTNtbW3SF2/9+vVkZ2dz4sQJqqqqSE9Pp7i4mIMHD5KdnY3FYuHEiRMcO3aMzZs3k5+fz4kTJ9i0aRMnTpzgxIkTrFu3rt+P5/nz5zl8+DAHDx7EbDZTWFhIdnY2n376KWfPnpWemHft2kV2drZbOKnZbCY/Px+DwcDPfvYz4uLi2LRpE0qlkptuuomwsDDGjRvHihUrUCqVknJpQkICS5YsYebMmbz++utSmWJ78vPzOXfuHLm5uTz//PNUVFSg1Wo5fPgwO3bswGg0kpeXx5EjR/jwww/Jz893M3S6u7s5evQou3fvprm5me7ubnbt2sWxY8f6jUFDQwOrV6/m5z//OfHx/x977x1eVZW+f39OTe+9h1QCCaETegcBFRB7GcuMvfdRR0cd7GAZsSECKkhHpQUIEEpCCBDSeyG915OT5PTz/pF3718aTXG+DpP7unLBOWevstdae+1nPetZ9+2Lt7c3r776ai+xzvPnz/Pcc8+RmpoqusL379/fy9thNpvJy8vj6NGj7N69m4MHD5KamsoPP/xAa2srHR0daDQatFothw4dIjk5mfXr11NXV0dnZydarVac/M+cOcOPP/5IRkaGKPCm1+vp6uoS+S8OHTrEkSNH2Lp1K11dXRQVFfHrr7+SkpLSb8LQ6/WUlpYSHR19yWDBpqYmDh06xPDhw7GysuLxxx/H3t6e1atXk56ezn333TfgEWGZTMayZcuYNm0aq1atIj09neDgYEaOHCn2+759+8S6dXV1sW3bNoqLizl8+LDI9hofH4+FhQV6vZ7Y2Fh+/vlndDodGo1GPP4rTKhqtZqEhAR27tyJSqW6zCcWUY5g69ataLVannnmGUpLS3n++ecpKCggJyeH7du3I5VKef755ykvL2fVqlV4e3uLZGnFxcW88847dHZ28u233+Lq6srBgwcpLCyko6MDV1dXGhoaeOONNwDYsmUL2dnZtLa2UllZyY4dO2htbSUuLo4NGzb06rO0tDRef/11TCYTK1as4OjRo6xbtw6FQsGKFSs4ffo0K1eu5ODBgzQ2NmJnZ4e1tTVOTk58//33NDY2Ym9vz0svvURRURHPP/88paWl1NXV8eabb1JQUCBqIwmLsebmZr788ktcXFz44IMP6OjoYPny5WRnZ3Py5El27dpFamoqSUlJODg48PXXX5OamkpCQgJtbW289NJLA77k5XI5O3fuJDc3l9LSUnbs2IHJZCIrK4u8vDyOHDlCZ2cnL774Yj9iN61WS0tLi/gsp6SkkJSURG1tLZs2baKlpYUvv/wSs9lMeno6NTU17Nq1C4Dnnnuul3dHoVCgVqtxdXUlLy+PY8eOoVareemllzh58iSnT59Gq9XS1tZGV1cXjo6O1NbWsnv3bsxmM8899xwVFRV8/fXX+Pr69pNXMJvNHD9+nGPHjhEQEMDy5cs5f/48GzZsoKSkhKqqKr7//nvy8vJ4+eWXqa2tRa1Wi+l1Oh1vv/02ubm5ZGdns3LlSqysrHj33XdpbW1lxYoVODk5UV1dzXvvvcfp06c5evQo3t7edHZ2YjQaWb9+PRKJhK+//prjx4+LeZ8/f549e/agUCh6CWGazWb27t1Leno6Pj4+/OMf/6CsrIyPP/6YpKQkcnNz+43NnmnXrl1La2srtra2vPjii1RWVvLJJ5/g6+tLUlIS69atE5+39PR0zGbzZYkr/zfimjFooJvTYOnSpezYsYO8vDzGjh3L1q1bycjIYNiwYezevZva2loaGhpYsWKFOEAMBgOffvqpyAuwfPlyduzYQUNDA3V1daxcuZLc3Fy++eYbPDw88PPzw87OjuHDh5OdnU1sbCwxMTHs2bOH/fv3s2nTJtLT00XK7NjYWEJDQ/nXv/6Fs7Mz586dIyUlRax3eXk569atY9y4cWRmZvLDDz9w6tQp9u3bh7+/P5988gl1dXVs3LgRNzc33nvvvV4quBMnTmTGjBncfvvt7NixAxcXF86ePcuwYcOQSCSiYbd9+/Z++icSiYTx48dTUFAgGhg6nY7Vq1fj4+ODSqXC0dERa2trfHx8WLNmDVZWVsTGxrJv3z727t3LyZMnCQ8P5/XXXxeVhI1GIytXrsTd3R0/Pz9efvll2tra+PTTT7G2tiYxMbEXyVdJSQlKpVIk7mpoaOCnn37i9OnTYuyDg4MDCoWC4OBg1q9fj0ajITo6mjfeeKMX9X1RURE//PAD4eHhfPrppxgMBhobG4mNjaWpqYmvv/4ao9HITz/9RFlZGQaDgY0bN4oTs9Fo5JdffiEzMxNPT0/eeecdQkJC2L59OxUVFSQnJ3Pw4EH0ej0hISFkZWWRmZlJVVUVu3btwtnZmTfeeIOmpqZebS0QkAkkcD29OQ0NDeTn51NXVyd6D3pyhlhZWfHSSy+xatUqhg4diqWl5QUNIrlczksvvURCQgK//PILs2fPFjlBgoODiYuL68WhkpycTElJCWq1mri4OIxGIw0NDeLLJzg4mB07dlBSUkJ8fDwJCQliWWazmR9++AGDwSAajpe78lMoFHh6ehIVFcXSpUtxdnamo6MDFxcX5s2bh7W1NU1NTSiVSh5//HGKi4tRq9V4e3uLR2F9fHyQy+VUVVVRWVmJv78/L774ItHR0VhZWeHs7IyXlxeWlpYoFAruvvtuYmNjKSwsZMyYMRw6dAiAuXPnMm7cuF7PhZeXF76+vsyYMYMbb7yRvXv3kpGRIVLvBwQE4O7uztixY7n55ptxdnYWRV33799PWFgYw4cPR6/X09raiouLC4sWLWLatGk4ODgQGRnJiBEjsLOzY+TIkWi1WiQSCbfccgvnz5+nvr4euVyOo6Mj0dHRjBw5UpRaCA0NZdiwYbz++uucOHECg8GAp6cn995774DjQiqVcsstt3Do0CGsrKzIysoiIyOD6OhoEhMT0el0uLq68sADD/RLb2lpyfjx4xk1apQo7BkWFkZdXR21tbVMmDABs9lMamoqvr6+5OTk0N7ejrW1NY8++mgv3huFQoGlpSXOzs4cO3YMvV6Pm5sb999/Pz4+PuzatYszZ87g5OQkGocpKSmo1WpsbGx45JFHyMzMFD0fgneuJ2JjYwkICMDHx4fw8HBOnjyJUqkEur2rZnM3I7KHhwfLli0jKChITCvIOkRGRhIdHY2trS0jR47EYDBQVlZGdnY2AQEBTJ8+nbNnz7JlyxbCw8Px9PTEzc2N1tZWzpw5g9Fo5O677+5VP2dnZ1JTU/nll1/6CYTu2bOH4OBgUYohMzMTJycnIiMjGTt2bD8lcAE6nY6DBw8SGhpKVFQUHR0dHDhwgJaWFry9vZk7dy4HDhzAaDSSlpbGiRMniImJQaPR8OWXX7Jy5co/RQzS1cI1ZdBIJBJiYmKoqalh586dPPHEE+Tl5XH48GFCQ0NJSEjAzs6OqVOn8re//U1MZ2Njg5OTE8OHD2fJkiVUVVVx5MgR7OzsmD59Og888ABubm74+Pgwc+ZMPD09kcvlWFtbc/r0aVGnJSYmhtOnT+Pi4kJERAQzZ87ExcUFLy8vgoKCsLW1xd/fHw8Pj14vO0FEztbWlilTpoirL19fX8LCwtBoNBQUFNDW1oatrS3Lly/v9VI0Go28+uqrrFq1iq+++or9+/fj6OjYa3/Vy8uL/fv3iwaDAMGT4uzsLD70CoWCkJAQnn76aZFcS6FQYDAYyMjIECeWSZMm4eLiQmhoKJMnT8bLy4va2lqgO5YjJSUFLy8vwsPDaW9vR61W4+DgQGhoKD4+Pr3cqIIAnbA6dHV1FV8eAmeEQqFALpejVCo5duyYqMHi6OhIYWGhOAacnJxwc3MjKCgIBwcH/Pz88PPzo76+HldXV+RyOQqFAgcHB4KDgwkNDaWlpQVXV1exDAcHB/z9/QkODsbGxoaAgABsbW1Rq9V4eHgA4Obmhlar5fjx4zz11FNkZWVhMBhwcXHhnXfe6aVWbjZ3x3hFREQgl8sxm7sVgU0mEwcPHuT06dPY29uTkpLC0aNHaWpq6sUaazQayczM5C9/+QsrV668aAyCRCLB3d2d8ePHiy9zARYWFrS1tYkGjVQqZdmyZezfv190e585c4bo6GgkEonYv66urqKnsyf0ej3JycnY2dlxyy23sGjRogvWayD03Ma1s7MT85dIJFhaWlJTU8OIESMYPnw4RqOR8vJydDodBoMBo9EoGk8WFhbk5+eLK24h9qLnNQBjx46lpqaG6upqXF1dRdK0mJgYlEplP2NM+Gw0Ghk6dCj19fUMGTKEcePGDegJEbYu7ezsxO0Re3t7nJ2dL9oOwsKjtraW1atXExkZecH6WFhYkJiYiFarpaqqCqlUSktLC2PGjMHf3x+1Wk1dXV2vtBKJhNGjR1NeXk5FRQWTJk1ix44dhIWFYWFhIap0BwYGDuj16Pn/vXv3UlxcTHBwMNBtFC9cuJAvv/yS4cOHi0rk0dHRhIeH98sPureFhHLHjBlDQEAAZrOZjz/+mMLCQg4dOiQS9llYWFBbW8vIkSMJCwsTt9wEssW+fezu7i5uAclkMnx8fJBKpSIx3kDSDZeC2WzuxUwsGEXW1tYUFRVhNpvFUIfm5ma8vb2ZMGFCr7qZzWbRE75p06ZefePu7i6KRCqVyn6CrBeCTCbD2tpaZHV2cHAgMDCQ6upqkXgyICAAiUTC8OHDqa+vJzY2FisrK2688UZuvvnmfizb/824pgwa6JaLnzx5MmazmZCQEMaOHSuKk1laWlJeXi4+9D1PPQiGQXt7O+Hh4djZ2VFRUdHr2r4PNnRTbGdnZ2MymdBoNAwbNqzXNQOl6fu9r68v58+fR6vV0tXVJXpWeqazt7enoKAAFxcXHBwcehkDu3fvprq6mvHjx7Nw4UKqq6u55ZZbSEpKoqKiQrxOLpcjlUp7abo0NTWxceNG/va3v4nsnHq9nunTp/P888+L3gzh/oVAUx8fH3EfVtBMsbW1xcvLC+h+wdjZ2YkTj5ubmxif0rMtBAwbNoyAgABxNQHdD3pPplJh0hfKz83NxWQyYWVlha+v7yXHRs/2H6g/BooD6fld3+u0Wi0rV67kkUceEdsuJycHT09PLCwsRFe2kD45OVlczZaUlFBeXi5u7V133XV4eXkxYcIEYmJi8PHxEV31JpOJffv24evry0svvYSlpaXYL5WVlf1eXgPdswCNRoOrq6u4apZIJIwcOZKqqira2tqYMmUKsbGxBAUFXfTehe8EzZyWlhb8/PzEyf5KcPbsWfbt28d1110nstVWVlYyduxYLC0tee6558SV5dChQ/nggw+oqKhAo9FQVlaGhYUFFhYWLFiwgOeee45t27aJXo99+/ZRUFCAlZUV9fX1WFpacssttzB69GjkcjkPPPAAH374IR9++GE/Vlzo9p4ePHgQiUTCHXfcwbx583j++ef54YcfsLCwwGAwiFvUNTU16PV62tvbef7550lKSiI+Pp4FCxZgYWGBpaWl2M5WVlY0NTXR1taGTqejtrYWGxsbVCoVarWaPXv2YGVlxZkzZ0StH+HZu+WWWzhz5gyvvPIKTU1N3HTTTaSlpYnyCZWVlTz11FP9tv+srKxYsmQJMTExLF68mHHjxmFlZcXChQspLCzk5ZdfpqCgoFd8mFBXYctYYMI9efIkpaWl1NTU0Nraypw5cxg/fjzu7u5MnToVjUbDCy+8wJkzZ3ptb0ilUqKiotizZw8zZ84kJyeHv//97xQVFdHU1ERaWhpTp05l6NChjBw5kmPHjhEeHo5areaFF14gJSWF6dOn4+npyUcffURTUxMqlUo00CUSCffccw8qlYr4+Hg8PT2ZPn0606dPZ+PGjRQUFGBpaUlpaSkKhUKMORGgUqmwtLSksbGR1tZWUQtLWJw88sgj7N69m4SEBB577DEeeOABkpKSWLNmDba2tnR1dbF06VL+/ve/880332BrayvKLlRVVZGYmMjYsWMZM2ZMr755+OGHKS0tJT4+noiICIYPH45MJqO5uVlc/AoGtBDfJRiEL7zwAgkJCcTHx7N48WKmT5/OjTfeyJ49e0hPT+exxx6jvr4ee3t77r//frZu3Up2dja+vr4EBAT0k9v4b8Y1yUOTnp6OVColMjKS9PR0FAoFw4YNo6SkhPfffx8vLy/uuOMOhg4dKoq/3XbbbVx//fUolUqmTJkiRq37+Phw++23U1ZWxokTJ3jhhRdoamrik08+4eGHHyYsLIwNGzbg4eGBTqdj+vTprF69mqCgIJYsWcLmzZvRarXMmDGDVatW8dRTT7F37168vb1ZtmyZSEe9fft2URRu+vTpxMbG0tnZycSJE1m/fj0vvvgi27ZtIycnh1mzZrF06VLRo3LgwAEKCwvx9PTk/Pnz3H///Tg7O5OcnMzx48fFvWalUskNN9zAmjVrqKmpYcyYMXR0dBAREUFMTIz4ktNqtWzcuJGwsDDUajWTJ0/mzTff5Prrr0er1bJ582ZGjhzJPffcw48//khVVRUTJ07E1dWVqVOnioZHdnY2x48fx9/fH09PT3ESeuyxx9i1axfBwcEsWbJEpAWvr69n06ZNYgBrVVUVd9xxh2gkGQwGli9fTmRkJOPGjePXX38lLCwMhUIhsn+aTCZiY2NJT0/n7rvv5pNPPuGee+4hJyeHtrY2UZzwqaee4scff2Tq1Kmo1WpycnKYM2cOP/74I0899RQ7duwgIiICd3d3tm/fzlNPPcXq1auZNWsW1dXV1NbWipPkddddJ37+8ccfqa+v57rrrmPBggUizf25c+fYsGEDMTExWFhYcO7cOR566CESExOxsLBg8eLFlJSUcPz4ca6//nosLS357LPPeOmllygoKOCHH37gsccew8fHh59++onDhw/z5JNPkpWVRXV1Na+++movmvm6ujo+//xzHBwcePDBB3FycsJsNouqzNddd12v6/ft28ewYcPE1e/s2bNJSUlh06ZNPP3006xevZopU6aIAYtLlizhk08+4eabbxbjBcLDw7n77rvx8fG5LDIxs9nMp59+ioeHB0uWLMHKygqTySRS+8tkMoxGI1qtFisrK5Emv6urC6VSKUoWCPT/EomErq4uLC0txWsFWvueEgFGoxGJRCKOu66uLmQyWT/6/OzsbD777DM+/vhjsXyz2SzGEAmrfsHwFuJ6BMNIp9NhNBpFyQKj0YhMJhPT9ZRAEOorl8vR6XSiPIAgsdDzWqEso9Eovoz0ej0GgwErKyvRazhjxgxxjhAg3LvQ/kK+PdP3bAOBkK6n1INcLhfbWafTiXUQ6is8qzqdTmy3nn0ueFWEGC2hXKEs6F4QmUwmse975id4bjQajTgO+pYh/G5tbS3OC8L1wvwktLcg39D3fgfqG6lUSldXF1KpFAsLCwBRLkGQHxDGlFA34Z5kMpk4RvpuGQvt0vMehbEl/C6MK8Gj2fM56TnWBDFUQaJC8K73lEr4TxEM/qdxTRo0wi0Jnd/z/z1d7UJnGgwG7rnnHt5//318fX3Fh6PntUK+PQdYz8EgaGRcqvy+3/XMTxigF0rXs5y+kw4gTt5CvsIDKTygwmTT1xXd81/hd6GthDSC1o0wmQjt99lnn+Hl5SUaZ33zEbxBPcvue0990wgTgOD16NlGPevRs449rxmonAv1Q9/+vNzf+vZvz88Xq1NPSCQScnJyOHHiBLfffjv19fVs3LiRN954A6lUSk5ODh0dHYwZM6bfWBHSV1dX09TURFRU1AXLE9I2NTVx8uRJ5s+f3+9F11PL6EruXbiu55i43ElSq9Xy8ccf4+joyAMPPCC+IP4MMJvNHDlyhB07dvDmm2/i5ub2XzP5CwZDXwNtEIO41nFNGjRXirq6Oj766CNuuukmYmJi/nCdj2sFWq1WdKveeeed15Tr8j8Fs9nM+fPnaW5uxtHREQcHB1xdXUWDobGxEWdn536CggIEY/VyxmxLSws2Njb9jJn/KwhGPPCnWy0KK3hhVTs4JwxiEH9+DBo0DLxyHsSlMdhug/i96Onx+U+kGSjdhfK7VDl9f/8t9boYrrTOv6U+V7vOgxjE/yWuuWXHQIGdl4LgJu85EVwNO+9q5XOxvC8WDHo1yx6ovJ7tdqUvl99Sv9+armfaq4nfU58rLeP3XnM10lxt6HQ6tm7dekW08A0NDXz99de9guIvBYPBQGJiIl999VWvIHnoboeSkhISExPFOm3YsIFPPvmEo0ePDpif2dzNCRUXF8e+fftob2/HZDJx8uRJ4uLiBuQgulKYzd18Sn35R4T6fvDBB6xZs4bOzk7M5m5W5F27domnBFtbW9m/fz8HDhygoaFhwProdDp27tz5m1ivBzGIPyOuGYPGbO5mkb0Qrf3lQphk+5JLXSkMBoPIHPpHwGAwsG3btl7kdAJaWlrYvn37VdNfMZu7Ccx27NjRTw+ksbGRn3/+WZRXyMjIQKPRDNgHZnP3KalTp06xcePGXtwxl1OHoqKiXqRwfWEymTh06BB79+5l27Zt7N69m19++YWMjAx27NhxRaRvl4LJZCItLY3Dhw9ftTz7orOzkxMnTrBx48YL9mV9fT07d+4c8AjxQDCbzdTU1PDzzz+L2z3/V5DL5WRmZooSDj1jf/oaXMJnW1tbjhw5ctFj633T5+fnk5SUxPXXXy+yhQvo6upizZo1ohGSlZVFZ2cnQ4cOvSCbs0aj4YMPPmDUqFFIpVK++uor0tPTOXz4MFOnTmXr1q0UFxdfdv0uZBhLpVL279/fi0RTo9Fw5swZ5s2bR2JiItu2bRP79OuvvxYDVAVyvsDAQFauXDmgjpVcLic/P1+s66WM3D+DETyIQVwM14xBA/+P9bTvQ3ehybHn/4U/qVSKt7e3eKLhYpPrQDAajWRlZSGRSEhMTBTpzC81eV0OeqYTiL96Btv2/O3HH3/sZVh0dHSIPAdXuurv6uri22+/Zfz48f3iZOzt7dm2bRsNDQ0MGTKE06dP8+yzz15QnG/t2rW4u7szevToi76EB6pje3v7gP0rQK/Xk52dTXh4OJs2bUKlUuHq6kpubi5btmyhra3tsvvxUvWSSCQ0NDSwd+/ey76HK31h7NmzB41G0+/FKqzIBUPghx9+EE9PXE5ZRqORH3/8sR8VwYXqc7nPzpVC4JrJyMhg+fLlHDt2DKPRyPHjx4mLi+PYsWOYTCbS09PZv38/O3fuRCKRoFQqMRgM/Prrr6SlpfV64RuNRhITE4mNjeWXX35BpVIRFxdHVlYWKpWqF3WAyWQiKSmJoUOHit8dPnxYPE3o7Ow8oOexubmZ0tJSbG1tGTp0KCdOnGDXrl2EhYVhaWmJj48P8fHxA96zRqMRWaezs7Oprq5m7dq17Nq1i1WrVtHV1SWSQAoEbT0hk8lYtGgRo0aNYtGiRXR0dCCRSPDz8xNlWNRqNadOnSI0NBR/f39yc3N7aSKZzd0keHFxcZSXlwPdC5ODBw+ybds2amtrOXHiBGvXrqWrq4vjx49TVFTEmTNnSExMvCoeqEEM4o/ANWPQSCQS7O3tkUqltLW1sXv3bmpqaqipqWH//v3s3btXFLI7ePAgv/76KyqVioMHDxIfH88XX3xBQUEBHR0dtLa2otPpSExM5MSJE6xbt468vDzR3Suk7yswaDabOXr0KG+99RbFxcXY2tpSWVnJ2rVrycjIEFWCd+7cOaDyqUajYcuWLSIV+cGDB+ns7CQuLo7Ozk4yMzPZvn07mZmZdHZ2olKp0Ol0dHV1ceLECXbu3CmqCisUCpEfobm5mbVr1/L1119TW1tLRkYGSUlJFBUV9WtHwdN1+PBh9u/fT0dHB2fPnuXcuXNUVlYSHh7ea5JXKBRYWVnh6OhIYGAg999/P3K5fEDG2PLycvbv309mZiYZGRnU1NTQ0NAgijUKmiWtra0cOXJE1HFpa2sjPj5e7IMLQaFQcP/99xMUFISVlRUuLi5MnDiRefPmYWFhQVpaGqtXr6aurg6dTsfp06fZvHkzlZWV/fJVqVRs2LCBoqIiGhsbOXLkCBUVFZw4cYL9+/ej0+lwcHBAKpVSW1vLzp07RcHG8vJyNBoNSUlJbN68mdraWlQqlXiMvq9XzWw2o1arOX78OHv37qWgoICmpiYOHDggMpP2Jdn76KOP+Omnn9DpdEgkEo4ePcr3339PR0cHXV1dnDx5ks2bN/fSqRGeExsbG8xmM4cOHeLHH3+kvb2dPXv2cOzYMbRaLYcPHxYFMePi4vjqq69EaYiEhAS2bNlCY2MjeXl5xMXF8cMPP/xmkUqzuZtjaeHChSxfvpyysjLOnDnDsGHDWLNmjfgszJgxg6qqKtRqNSaTiXPnzuHh4UFkZGSv8ZicnExiYiJz584VNaKGDh1KQEBAL8NF2NJxcnISiRIBnnrqKVatWkVaWtoFjWdHR0csLS05ffq0qKVUXV2Nvb09Ekm3zs6FvI/V1dVUVlbi5eXFunXrkEgkbNu2jaioKLKyssjOzuaLL74gNDSUkSNH9jOolEoltra2GI1GampquP766/tdI0hzWFhYiEfbe8oPCJIPM2bMEO89PT0dmUxGR0cHsbGx+Pv7i3NmVVUV9vb2rF+/Hn9/f/H04SAG8WfDNWPQCNDr9aSkpBAYGIidnR1r1qzB1dWVH374gZMnT7J69WqcnJzYuXMnx44d44cffqChoQF/f39ee+01VCoVX3/9NV1dXcTHx5OcnIyfnx+rVq1Co9GwYsUKRowYwZYtW/q9LKCb4dbe3h5/f3/MZjM6nY6wsDA+++wzsrOzOXjwIE5OTvzzn//s9xJQKpXk5OSQl5cn8qmYzd3kd9nZ2SQkJGBra8vrr79Oe3s7q1evRqVSsXv3bqqqqlAoFPz0009IpVJ0Oh0uLi7U1taKZXp7e6NQKPj+++8JCAgYUEG3qqqK7777jpiYGMrKyvj2228JCAjA0dFRJHu6GKRSKRMmTCA1NbXfy8DV1RVbW1tGjRqFVqslKSkJJycnKioq2LBhA0qlEhsbG9asWYOFhQUJCQns2LGDL774Am9vb4YMGXJRg0YqlWJvb9/rO4G11Wg0Ym1tjdFoZOvWrezevZuKigo6Ojp4//33e63yoZs9Ojs7m6ysLNrb2zGbu+nJBcr27Oxs8Vpra2s2bNiAVCoVRRV37NhBU1MTjY2NfPrppxw6dIjS0lIcHBwGZE796quvUCqVREdH849//EMUlAsLC+vH5CncU1BQEHZ2dnR1dYliiKdOnWLr1q20trZSW1vL559/PmCbdXV1ERgYSHJyMhkZGej1evbs2YPJZKKyshKFQsGqVauwsbHBYDDw0UcfsXHjRjo7OykrK+Orr74iMzOTzZs3ExgYKPKUXKm3RiKREBgYSHR0NB4eHnR2djJ8+HCKiorQaDRkZGSIdPlPPPEEDg4O6PV61q9fj62tbT+qgIyMDJH/Y8yYMWRkZIiUBT2vFeQuDhw4wLZt2zhy5AhZWVlYWFgQFhbGc889108mpGd/r1y5koaGBs6cOcPYsWPx9fUV41kGYlQW4O7ujouLC9XV1SKfjZ2dHZ6enjg6OtLc3ExaWhru7u7Y2dmJVBA9IXhYpk6dip+fX7/2trS0xMrKSuSqEYhFBeTk5CCVSlEqlTg6OgLdxJYdHR2oVCo6Ozvx8/Nj2LBh/PLLL9jY2ODq6srkyZN54oknBpz3BjGIPwOuOYOmuLiY3bt3ExISQmNjI+Xl5dja2vKPf/wDV1dX0XPy4osvMn78eJycnAgPD2fOnDniEViBjEmgzQ8JCUGtVmM2dxMm1dbWEhERQXBwcK/JVCDYEuj5JRIJQUFBhISE0NXVRUpKCjKZDG9vb9577z3RRSxAoKE/ePAgXV1dnD9/nqSkJCIjI0lJScHCwoLAwECWL18u6sZAtyFUVFSEs7Mzo0ePxsrKCktLS8LCwggKCqKlpQWFQoGFhQUODg64u7vzwgsviER+PZGfn49Go8HGxkaUYVAqlchksgHFEAdCY2Mjvr6+/a4ViKcEvRuJRIJcLue+++4TCfgE5ViBEG78+PGkpKTg6+sryhb8FsjlcsLCwggJCaG5uZmEhAQcHR2ZNGkSTz75ZL+6SqVS7rjjDmJjY8nNzWX06NHMmzePmpoaVCpVrxgOqVQqcrAIbZqYmIijoyOzZ8/mb3/7G6NGjWLHjh3s27ev3wpXr9dz8uRJPD098fHxwcbGhoqKCvFF3veehXIsLCyQyWTY2toSEhKCn58fzc3NnDx5EicnJ+bPn8+99947YHvY2dkREhKCt7c3bW1tzJkzh/Lyck6ePEloaCgODg44OTkREREhkv4J93T99ddz11134eLigq+vL1OnTuXnn3/m9ddfJy4u7or7RiBB8/LyQqvVsmnTJvz9/TEYDLi7u3PgwAHy8/M5e/YsKpUKhULBQw89xPvvv98v4DUyMpLMzExRP2ncuHEiqV1PyGQynnjiCe644w6mT5/OqFGj8PPzo6ioSEw7fvx4mpubSUpK6mfwBgYGMm7cOKqrq3nwwQeZNWsWRUVFInvz1KlTSUtLo6Kiolf94uPjSU1Nxc3NDZ1OJ8a8mM1m8Yi4UqkkPT1dNC56bs2azWaSkpI4f/48FhYWJCUlodfrRYJB6JYRiYyMpLKykoaGBgICAnoZWB4eHmRlZdHc3CwyE69ZswadToeNjY24VX3XXXfxww8/MGzYMLRaLREREbzzzjts27ZN9NwOFJsziEH8X+GaM2hCQ0Px9vYWV/mVlZXIZDIcHR3p6Oigrq4Oo9GIs7OzOBkaDAY6Ozvx9/fH1ta2l+q0AJPJhEKhYP78+VRVVXH77bcPqLchUMH3nKiE9C4uLmRkZODq6orZbB4wsHHo0KGo1Wqqq6tZsGABv/76KyEhIWJaDw8PUfFZWA1HRUXh7u6OXC7njjvu6BfbYDZ3E9tptVpUKhVLly7l/vvv59tvv+1n0Pj4+FBRUUFXVxdqtZqoqKh+bdEXwn0aDAYKCgpITEzknnvuGdD46Vsvk8lEamoq999/P5999pnIcNnW1oa7u7tIP37+/HlRJVso50IxOGbz/+MQuVC8h52dHfn5+Xh5eYmU8j0hkUiIiIjAYDBQW1uLnZ0dH3zwAZ6entja2ooxK0I8jdFoFFe4AvNpSUkJXl5eNDc309nZyb///W8aGxt7CZNCt7Hl4+NDTk4OBoMBGxsbvL29xfYZqO2lUilarbZXPYTrLCwsOH/+PJ6enrS0tPR7GQ80PmxtbbnhhhtYu3Ytw4cPF/vVaDTS2trKiBEjsLCwoKKiAi8vL5qamsS6SSQSbr75Zl5++WWmTZt2wXEyEBYtWkRtbS1Hjx7lkUceISQkhCFDhpCamsq4ceNwdnbmlltu4bvvvhOfqxEjRuDr68vMmTM5fPhwL49XTEwMS5YsYf/+/bi5ubFgwQI0Gg1ubm69PKJSqRQvLy/RMJnx/7Pqbtq0iS1btmBtbc2cOXNobGxk3bp1vRSZzWYzmZmZnD17lhdffBEfHx/Gjx/P8OHDOXToENOnTyciIoKUlBR2797d634jIyPR6XSiVMT58+cZPXo0dXV1+Pv7I5FIeO211zh8+DCpqanMmDGj1/11dXWRmJjI2bNn+f7776msrEQul1NRUcHo0aNFbacnnniC3Nxczp49yzPPPNPLKI6MjGTJkiV89913uLu74+XlxfTp08nOzsbNzU1k6A0KCuL6668XtYAKCgpobW3lgQceENmIBw2aQfyZ8NuWu39CmM3ddO8BAQEsXbqU1atXU1payn333cdHH33ExIkTue2223jsscf47LPPGDt2LHfccQcmk4mEhASKi4t57LHHaG1txd/fn8rKSpF+u76+Hl9fX1pbWzl9+jTt7e1IpVJuuOEGFi1a1OvF7eXlhZOTE6dOncLCwgKtVkttbS1+fn6MHDmS1NRUli9fzvXXX09kZGS/+1AoFNx3331ikF9QUBAWFhbMnz+ftLQ03nnnHZYsWUJXVxc+Pj7U19fT2NgoTnKenp7cfffd+Pv7U11djV6vRy6XM3z4cJKTkzl//jxZWVkEBgZy66239jM6QkNDuffee9m7dy8ymYyHH36Y8vJyfHx8qKqqElWOBdTV1Ymery1btqDX63nvvfdE4bqeENq2trZWZDUuLCykvb2dZcuWkZeXR2ZmJs888wzr169n+PDh3HPPPTz//PN8//33hIWFMXToUFpbW9m4cSNPP/10P9E/s9lMZWUlnp6e1NbWotFo0Gg0+Pj4UFdXh1qtxs7OjiVLlvDxxx/zySefcPvttw8YFyCXy7n//vtxd3dHKpUyatQodu3ahb+/P/n5+fj6+uLk5ITRaGTOnDmsX7+eIUOGIJPJePTRR1mxYgWlpaXcdddd1NbWUl5eztSpU0WDQYBEIuHll19m165dHD58mNtvvx07OzusrKxoaWlBr9f3I8ObMmUKu3fvFvVYampqRIr/xx57jE8++YSSkhLuuuuufrTwjY2NYhqBKt9sNjNr1iza29uxt7dHq9Wi0WhEj8ujjz5KU1MTn376Kfn5+dx5551UVlZiZWUlprlSSCTdOlIjR47s9f2bb77Z63N0dLRoOEkkEl555RWAfumEPrvhhht6Xb9kyZKL1iMiIoKIiAjMZjOvvfaaWDch2PbBBx/sdTpKIukW+uvJzqxQKLj++ut7lTtz5sx+W2KBgYG8++67ACxcuBDoNsIAnnzySaC7j3rGBvVMb21tzYsvvtjvHqKiosTFB4C3tzc33XRTr3vp2Ub33HOPWFcBU6dOFf+flZVFTk4OY8aMET2rt912W680zz777EXbdRCD+E/jmiHWu9ht9NQYge6Vp3CK6cknn+Svf/0r0dHRl2QDTUtL49ixY9x9992o1WqKi4uZNWvWgHT/F8tL+P1ShF0CesoYmM3mXnkbDAbefvtt7r33Xuzt7cnKyiIqKmpABdWe9y3kcyEvSk8q+74YiMzrYtdc6tqB0FPi4EKfB6rflZRxqXYYKL+eEgF90/Tt94vdw0D1Hqh/BQx0/cXG2aXK6vs5KyuLvLw8Ro0aRUhICFqtlrvvvpuvvvoKFxeXC/bDhep3reBifXIpXKit/uwQDjeUlpZy5513/qkkKQYxiIvhmvHQXGzS6BuDIBg37e3tuLu7U1tby4gRIy458QQEBGA0Glm7di2RkZEDutclEsklA2cv9fuFjIiBvpfJZEyePJmNGzfi6+vL9OnTRer8C5V7qfsUYlsuB1cyWV/JtX3b6FKf/6j69L3mYn33W+sslHOldb+SuvRN2xMmk4n4+Hi8vLwICgoSj6QHBQVRUVHRazxdauxea/g9Bsl/q1yCRCJhxowZ4v8HMYj/FlwzHprfgoE8IVfz+v8U/qz1GsR/B3qOn760+X2/H8QgBjGIPyv+pw2a/2sM9CL5rel/ax5/JIT6/d56DZTP1cq7Z14987ua+Q9iEIMYxCD+eFwzW07/aVyNF15jYyNnz56ls7OTpUuXXpaLWjgdlZiYSHt7O9bW1shkMmbNmvWnILzqaRzs2rWL8ePH4+Xl9Zvz6+zsZN++fUyePLlXPiUlJRQXFzN37tyL9kFaWholJSXiKSxLS0scHR2ZNm2aGJuUmZlJTU0N8+bNA7rlL44cOYK3tzcjRoy4YN6tra0cO3YMiUTC6NGjSU9PR6vVMm7cOHJyclCpVEybNg13d/dBw2gQgxjEIP5g/Hdu8l4EfY+w9vxuoN8v9NvF8jGbzbS3t4tMuxfKc6A8eua1Zs0aAgMDCQgIuGRdBHR0dPDSSy/R2trKnDlzCAsLY+fOnQOStV3JvV2sDS+n/YTPXV1dothgTzr2S6W70DVyuZzY2Fjq6up61dHOzg5vb+8L5iUgMzOTkJAQysrKOHbsGMOHDxfrJ1zf1NREbGys+J1MJiMjI4OMjIyL5m9nZ8fp06dF1tq8vDyOHj2Kh4cHVVVV7N69u98prEEMYhCDGMQfg2vGQ2M2dwsoCrIAADfffDPV1dWUl5fT0dHB9OnTOXnyJNCtxzJ+/HjKy8tRq9XMnj2btLQ0Ojo6qKmpwcPDQySsu+222zAajaSkpFBbW8t1113H6tWraW1t5fnnn6e1tZXMzEwcHR0ZN24c8fHx4lFHLy8v1Gq1SHYleAVKS0s5fvw4QUFBLFiwgJycHGpqatDpdMyYMYPMzExUKhWVlZXcdddd4kmDw4cPU1dXx0033YRcLsfBwYHXXnsNpVIpHqk0m800NDSIXoIZM2Zw/vx5qqurqaqqYsmSJeTn53P+/HmmT58ukuCZzd2sxPHx8eILesqUKSQnJzNixAhGjRpFdnY2ubm5BAYGEhoayqFDh3BxcaGkpITbbruNDRs2kJWVxYsvvkhDQwNubm5YW1tz+vRpVCoVzs7OREVFERcXh7OzM9nZ2dx00024uLiQmppKYWEhY8aMITQ0lNTUVFpbW/sxKpvN3VpG9fX1+Pv7c/DgQVxdXcnPz+f222/HwcFBvHbJkiXY2tpy5swZbGxsCA4OxsbGhmPHjqFWqxk7dix2dna0tbWxdu1a7OzsWLp0KXZ2dkD3KbKe9QoLCxO9LVKpFBsbG5RKJQqFAltbW/Gzra0t1tbWfwqv2SAGMYhB/C/gmvLQVFVVsWHDBsaOHcv+/fs5cuQIu3btwtfXl9jYWEpKSti2bRuVlZUEBQWxd+9evL29OXToEAUFBWRmZnLkyBHGjBnDxx9/jLe3N8XFxZw6dYq1a9cilUrJyspi8+bNODs7M2TIEFF/ydPTkxUrVlBeXs4333yDRCLB2dmZNWvW4Ofn18+D4u3tjbOzM9HR0dTX17NhwwYmT55MSUkJ3333HSdOnCAhIYGIiIheW1Hp6ekEBQWJp5BycnJYvXo1OTk5vfI/cuQIlpaW5OXlcfLkSTIyMti9ezfDhg1j9+7d1NXV0dHRwcqVK3t5HkwmE2vWrMHNzY2ysjLi4+OJjo7myy+/JCsrSzRg3n77bVpbW1mzZg1OTk7k5uZy5swZXF1d8fX1xd3dnV9++YXS0lJ27txJVVUVU6ZM4auvviInJ4fNmzfT2dmJ0Whk586dxMfHk52djVKp5O233yYlJYVjx44xceLEAY+NlpeXs3PnTqRSKevWrcPCwoLq6up+ooB2dnb9tntKSkpobGxEp9Oxfft2oNvzFRMTw7Zt2zh+/Lh47dGjR3vV63JVrQcxiEEMYhD/WVwzBo1EIsHJyQlPT0/CwsKYM2cOBQUFzJkzh/Pnz6NWqzEajTg5OREWFsbYsWOZM2cOpaWlqNVqNBoN9vb2uLu74+vri42NDV5eXri7u1NXV8e5c+dwdnbmL3/5CzfeeCNyuRyFQkFxcTE6nQ43NzdWrFghkq1FRUUxevRooqKiePLJJ3sZNMKxaJlMJhodJpMJKysrJk6cyLlz50Sxx4kTJ/Za5bu6utLU1CQaIX5+fiQmJuLi4tKrPSZPnkxLSwttbW10dHTg7OyMr68vEydOJDU1FRcXF2bMmCGSeQn1sra2xtbWFl9fX7y8vER6e4HqXKlU4u/vz0cffYSzszP29vYEBATg6emJSqUS20UQrDSbu2UAXF1dsbOzIzw8nNzcXGxtbQkICCAgIIDW1lZOnTqFo6MjI0aM4J///CdJSUm4u7tjaWnZSyFZqKeLiwtSqRQLCwvs7OwIDAzEx8fnskQShw0bhpWVFfX19ahUKqCbIXnYsGHMnTu3l3GYlJTUq17/a8eWBzGIQQzivwXXjEEjQNA1UalUDBs2jPfee4+AgAAsLCx6yREYDAbee+89/Pz8sLS0FCnkoX+8iVQqxWQyUVVVhZOTE7W1tUilUjQaDba2tmRmZmJlZYXZbKa1tVVMp9VqmTBhAq+99hpr167tRxMu1CUwMFCk9m9vbxc5cQaK21i4cCGVlZWkp6f3Ilfry0j6ySefYGtri52dndgmQl729vZkZ2fj4uJCQ0NDr3r1jWnpKTjo5uZGSkoK9vb2ojRC3zYTJBaEtoZuOQdBrNJoNBIeHi6mE/J3dXUlJSUFZ2dnWlpacHBwICUlBY1GQ1dXl6gv07NulxOXI3zfU8/np59+oqmpCXd3dwwGAyaTSfzT6XSMGjVKzEO4Z6FefT1tQr49yxDaTfi/oOI+iEEMYhCD+ONwzRk01dXV7N27F39/f6ZOncq4cePYvXs3ERERpKWl4ejoKHo4xo8fz+7duwkPDycvL09UZK6rqyM4OJjq6mqsrKyQyWQ888wz/Prrr2zcuBE/Pz8mTJhARUUFjo6OzJ07lw8++IDi4mIUCoUonWA2d6vidnV1cc8994ire7PZTH19PX5+flRWVhIeHs5dd93Fvn37aGtr45577gH+n2hfTwQEBLBq1SqOHj3Kjz/+yO7du7n//vv7eWhiYmI4dOgQnp6eVFRUoFKpsLS0pKOjg4ceeoi8vDy++OILXFxcxO0rITYlNDSUuro6LC0tMRqN1NfXExwczIgRIwgJCeGDDz6gra0NiURCSEgI9fX1WFtbYzKZiIyMpK2tjaysLJycnGhvb+euu+7C3t6eAwcOMGHCBMLDw/Hz8xN1muzt7Vm6dCk6nY6PP/4YpVIpxtV88803DBkypJfRZTKZaG5uxt3dnaqqKoKCgmhoaBDjlvpqF6nVarq6urCzs6O+vp6RI0eSnp4uagO5urri5+fHvn37iIiIYNSoUaIR27dePQVFm5ubRe9ZaWmpKO53/vx52tracHNzo6Kigj179lBbW3uVRvggBjGIQQxiIFxTPDTp6el8/fXXfP755722BnoGyw5EA3+h3/qi7zU9P1+Khr4vvf5AzX6hPC5EXW8wGHopPfe95lL3LeR9OVIGl1PPnnkM1JaXk06o00BpLkT6NhAut60vpw9/T1sJ1/03UuAPYhCDGMR/E66pU04tLS14e3vT1NSEh4eH+NuFtHcu9duFrh3o85XS0F9MmuBy63GxEzQ9X6CX0mO6nO974rdIN/zWdJfbdr8138vpw9/TVldy3SAGMYhBDOK345rx0PS9jcGXyCAGMYhBDGIQ/zu4Zjw0gwbMIAYxiEEMYhD/u7hmDJqrjZ4en8vR97mYh+hi6S6nDhdL91vz/q240viVK8nn9+goDdRfF7vuj2qvy63HH53flaa72vUexCAGMYj/NAYNmgsgLy+P/fv38+STT4qngKqrq3F0dMTGxqbf9V1dXcTGxqLVanFzc8POzo7Ro0ejVCqpqamhqqqKsWPHXvbLwmw2c/LkSYYOHdrvBFPPa9LS0sRTOv8JmEwmNmzYgEQioba2FicnJwwGAwEBAWRlZfHYY49ha2t7yXyqq6vZtGkTzs7OODg4YG1tTUxMDI6OjmRnZ5OTk8Mtt9xy2e3V0NDA9u3beeSRRy6Yxmw2s3PnTkJCQoiOjr5knr/F+GlpaWHjxo08+uij4rj5PWhubmbTpk08+uijV8SB09LSwtq1a7ntttsua2w0Njayfv16HnzwQRwdHX9HjQcxiEEM4v8G19yx7asFrVZLW1sbMplMfLF1dnZy7ty5Aa+3tLTk0KFD6HQ6xo0bR3JyMv/+978xmUwYDAa6urqAK9NJ6urqwmAw9ONc6XmNRqMROXQuxssicKPk5+eLx5ovdu1AfC7QfZRcoVBw4403kpKSgqurK3PnzkWn03Hq1Cl0Ot1l5eXh4cGpU6fw8PBg7ty5aDQaHn74YWpra/H392fy5MkXrVffPwcHBxYsWNCLv2eg8idOnEhgYOBl5VlRUUF2dvZlXSvAzs6ORYsWiePmQu14ub/Z2dmxcOFC8UTW5f7Z29uTm5uLSqW6aD8I39na2nL69Gm0Wu0VlXM5Y+Zy8FvSXmkdfm8dL1Tub0l3Nep2te5nEIO4VjDooRkAZrOZM2fO4O/vz4YNGxgxYgQjRozAzs6O/fv3M2XKlH6r9q6uLoqLi3n++edxdHRk0aJF3HXXXdx9992kpKQwfvx41Go1p0+fxmAwMGnSJPR6PcXFxbS2tjJx4kROnz6NnZ0dxcXFREdHo1KpsLKy4pdffiEyMpIzZ84wceJEsrKyRHbeiooKhg8fzv79+/H19aWwsFDkjKmrq6O0tJTGxkaGDx/OqVOniIuLY8WKFRgMBsrKymhra2PixIk0NjaSlpaGg4MDDg4OWFpa0tDQwNSpU3vdq0KhYNmyZSiVSuRyOZaWlgQHB+Pm5sb3339PbGws+fn5PPzwwzg4OHDy5EkKCgqYMmUK0dHRvXSQ5HI5SqUSe3t7Fi9ezLFjx9iwYQNTp06ltLSUZcuWiWrWHh4e+Pr6kpSURENDA0OHDkWn05Gbm0t9fT0zZ84kIyODe+65h82bN+Pq6kp2djYjRoygtbWVpqYm/vrXv5KQkEBgYCB2dnYcPHgQZ2dnGhsbefzxx8nLy6O6upra2lqWLFnCxx9/jIODA56enuTk5NDe3k5dXR033HADp0+fprGxkerqap566ilsbGxEIyg2NpYHH3yQlJQUDAYDBoOBGTNmiPeu1+vZv3+/yHQ8e/bsXtttiYmJmEwmJBIJ3t7eHDhwgHvvvZft27fj4OBAfn4+4eHhdHV1UVNTw0MPPcTu3bvR6/UUFRURHR3NkiVLxFNwTU1NnD59moKCAhYvXkxgYKBYnkaj4dChQ0ilUlpaWjCb/58mmqAjtn//fmQyGUVFRcyYMYPo6GgSExMxGo1ER0cjl8s5c+YMhYWF3HLLLfj4+FzxtpVKpSI9Pb3feLvYM1peXk56ejo33HADtbW1fPvtt+j1eu666y6GDh3aL41er2ft2rWMHj2a8ePHX1H9eqKtrY2ff/4ZZ2dn5s+fj6Wl5WWlU6lUfPnll9x7772isKoAtVrNF198wZ133om/v/9l1+Orr77i/vvvx9PT84rvYxCDuNYw6KEZACaTieTkZFxdXUUiO7O5m124oqJiwBVRSUkJSqUSX19f4P95TqRSKTt37gQgLS2NhIQEnJ2daWtr49///rdIMKfRaEhKSiIpKUlkpT169ChyuZwtW7aQn59PSkoKhw8fpqioSCTs27dvH0qlkj179pCenk59fT1nzpyhpqaGtWvXEh4ezsGDB4HuCXD+/Pmo1Wo+/vhjAgMDycrK4ujRo3R0dLB161YMBgPW1tYX5E2RSCT9tJUkEglKpRKTycS4ceNwcXHh8OHDbNu2jYaGBqytrVm9enU/wru+eURERFBcXIxWq+XkyZO0tLSwfft2goODsbKy4ocffiAgIIDJkydTWVlJS0sLKSkp3H777Tg7O3Po0CHkcrnoVZk9ezabNm3i+uuvJzU1lfr6ekpKSigpKUEulxMfH8/8+fM5e/Ys1dXVZGRkEBISQnp6OkVFRfj5+REVFUVDQwN79+5l/vz5KJVKvvzyS8rKyigvL+e2227r1R4WFhaip27Dhg04Ozvj7u7e617VajVVVVWEhYXx448/YjAYxN+EdO7u7ri6uqJUKomLi0MqlVJQUIBGo2HevHn89NNPLFiwgNzcXCorKykoKECr1XLffffxzTffiER+ZrOZtWvXiizIGzduFMsym838+uuvGI1GZsyYga2tLV1dXXzzzTe4uLiQkZHBqVOnKC4upr29nTlz5rBx40YKCgo4evQoERERSCQSVq9ejVwup729nW3btl2wjy8Es9nM9u3bWbdu3WV5G8zmbmLKDz74QCSejI+PZ+7cudx///0EBgYOmEYul5Ofny+Kl17Kw3MhD4i1tTW//vorTk5Ovfp+oOt7flYoFJSUlIjCpz2hVCo5f/489vb2A6Yf6E+pVFJSUnJZW7yDGMT/AgYNmgHQ0tJCZ2cnM2fORKPREBQUhEQiob6+fsAXvdncHe8yduxYLCwsMJvNHDx4kKlTp6LT6bCwsMDd3Z3IyEgyMzORyWScOHECDw8PPDw8uPnmm3FwcCA3N5eZM2cyd+5cCgoKiImJEen2x40bR21tLVOmTKGhoYExY8aQl5dHVFQUer2e6upqpkyZQn19PWPGjGHdunXExMSIbMOenp4UFhYyYcIENm/ezKhRo3BxcaG6uhoHBweys7MJDQ1l1qxZhIeHEx4ePqAn6mJQKpW4ubnh4uKCWq0mMzOToKAgrr/+et56661L5lVaWsrw4cOxsbERxT3Hjh3LU089RXV1NefOncPBwYGgoCAWL16Mra0trq6uBAUF4ejoiEwmQyqVivpPQmyOnZ0dFhYWaLVaMf5JUMR2dHTEysoKnU5HeHg4xcXFIoOwRCJBKpVSVlaGRqNBJpMxYsQICgsLsba2xsPDo5dQqEQiwcrKSmQsvummm/j73//OuXPner2krKyscHV1pbKyUtyi69mGN954Iy+88IIoqSGXy0VvjpOTE46OjqLmlqWlJRqNBmtra9zd3QkMDMTb21vUqDKZTOTk5BASEsJ9993Ho48+2qvNT506hZubm6i/1draSmNjI2FhYSxfvpwpU6ZgaWmJi4sLTk5OaLVawsLCsLS05JVXXqGtrY38/HxCQ0N55JFHuPfeey97vAioqqoSjRSdTnfRa83mbiX5c+fOiQa0wWAgPz+f999/n9bW1n4Gt8FgIC0tjaNHj9LY2Ii/vz/Nzc2cOnWK/fv3c/bsWXbt2kVWVhZVVVUkJiaiVquJi4tj//79/fTBmpqa6OjoECVKhDLS09M5efKkKPNRX1/P6dOnOXjwIO3t7VRVVeHs7ExRUREnTpzoxX5dXV2NQqEgMTGRbdu2odfr6ezsJDk5mePHj3P06FHMZjOFhYUkJydz8uRJysvLcXV1pbCwkISEhH7SKoMYxP8aBg2aAZCTk8Pw4cPR6XQkJSWJk3RqaioxMTH9rjcYDMTHxzN+/Hg6OjqIi4ujqKiIV155hdzcXPz8/KioqKCkpIQ777yT5ORkWlpasLCwoLOzk4aGBurq6rCysmLo0KGYzd2SCb6+vpw9e5Zp06ah1WpFgcna2lpkMhmnT58mODiY3NxcwsLC8PDwoLy8HLlcTk5ODu7u7sTFxWFlZUVtbS3Nzc1IpVLq6+sJCQmhrKwMOzs7xowZw+nTp1myZAkSiYS6ujrOnDlz0TYSVvw9Y2YMBgNGo1HUR/L19WXXrl20tLSQkZHRT75Ar9ej0WhQqVQcPHiQmpoabrvtNjGPlpYWxowZwyuvvEJsbCze3t6sW7eO7OzsXts5Qn5CuUajUdRm6vmdkG/P34V663Q6PvnkE5ydnTEajbS3tyOXy2lubsbX15fKykra2tqora1l6tSpYt59V9NGoxG9Xo9Op0OhUPDll19y7NixXhpQOTk57N27FxcXF1GrSoBWq8XBwYHPP/+cQ4cOodVq+92XcB9CecJ9qNVqVCoVzs7OeHt7i9c4OTlx4MABmpubRe+VgODgYA4dOkRTUxMqlQqJREJjYyOpqalUV1dTWlrar9yqqiruueceFi9ezIkTJ7C2tubIkSM0NjaSm5t7Rc+awWDg2LFjTJ8+Hb1ef0lx0aamJs6cOcOwYcPQarV4e3sjl8v55z//yUsvvcT333/f6/5MJhNbtmyhrq4OuVyOv78/Go2GVatWERQUxJkzZzCbzXh4eLBx40asrKxobm4mLy+Pw4cP4+Xl1e/EYlpaGkFBQdjb24tlbNiwgeLiYsLDw/n0008pLS1l5cqVBAUFcf78eQ4ePMi5c+cICAigqamJiIgIDAaDqDh/9uxZrKysiIqKYtu2bZSXl/PFF1/g4OAgHio4efIk8fHxhIWFUVVVRUpKSq/8LsbCPYhB/C9g8AkYAH5+foSFhZGTk8Ojjz6Kl5cXnZ2d1NTU9Ip3gO4Jrq6ujjFjxtDZ2UlSUhL29vasWLECDw8P3N3dGTZsGJaWlpSVlWFra8uSJUu46aab0Ov14kRmMpn4y1/+Iq72ha0bV1dXli5ditls5o477kAulzNp0iScnZ2JiorCy8sLKysr7rzzTqRSqZju9ttv58yZM4SFhTFu3DisrKwYN24c9vb2PPjggzQ0NFBfX88zzzyDQqFg8uTJDB06FIlEQltbG2VlZRdto+rqajEOSBDVnDFjBo2NjTg6OuLl5cVf/vIXXF1d2b9/P0OHDu116qeuro4pU6bQ1NQkxnB8+umnuLq6otPpiI6ORqfTUVZWhl6v569//SvPPvsscrmckydPEhoaikQiITQ0lPb2dlpaWpg2bRrV1dUMGzYMuVxOV1eX6NkaN24c7e3teHp6YmdnR3t7O1OmTKGuro7x48djNBq57bbbSElJYc6cObS3tzNv3jw6Ojpwd3fnpZde4tixY8jlcu644w4cHR1xdnbup7XV3NzMjBkzRKXz8vJyHnnkkV5xFkFBQURERFBdXc306dNpamoSf5NKpTQ1NVFdXc2jjz5Ke3s706dPp6amhtDQUCwsLOjo6CAmJoba2lpGjRolimAWFhZy/PhxHnvsMUwmE2PGjEGj0fDCCy+gVqs5fvw4kZGRvRik7733XlxcXDh48CA33XQTjo6OvPXWWyQlJXH+/HmCgoLEmKPOzk7R65ebm4uHhwdLlizhtddeo7a2llOnThEZGXnZz5nZbCYrK4uOjg4KCwsxm81UV1dfNHB2x44dxMXF8dprr1FcXEx2drboSQsPD8fZ2blXmsrKSg4fPsyMGTNIT09n2rRp/PzzzwQFBWFra0t5eTlhYWGMGDGCxsZGamtrGTt2LOHh4VRUVKBWq0XDRcCxY8eYNm2aGIBeWlrKnj17mDdvHm1tbUilUrZt20ZUVBSurq6i5ltycjLx8fH4+Pjg6OjIunXrePfddykvL+f06dPcdNNN2NraYm1tTX5+Pmq1mtDQUFJSUhg1ahTff/89s2fPxsnJiZtvvpkzZ85w+PBh/Pz8cHZ2HjxuP4j/eVwzTMFXEz2bRJi0kpOTcXNzE7effkteF8KFdJYu9Vvfa35LV16MU+f3TpB96/NbeXgGwh8xeV9u+/0Wnp1LaUBdqA8vt6z33nuPsLAwli1bdsnyLjXeLjb+LlT/K6mvkIdarWb16tU89thjWFpa8swzzzBp0iRmzpxJfn5+vy1Ps7lbwd5oNLJixQomTZrE1KlTSU9PJzQ0lJMnT+Ll5cXo0aPFdEePHmXnzp08++yzPPvss/zjH//g+++/Z+nSpRgMBjZs2MCHH36Ih4cHr7zyChMnTmTq1Knk5+ejUqnIycnhmWeeEdulo6ODxYsX89FHHxEaGkpGRgaFhYWcPHmSzz//nO+//55hw4YRGxvLokWL8Pb25scff+SRRx7h1VdfZcmSJaSnp3P//feLsVVdXV089thjfPbZZ/z66694e3tTXFyMVCplwoQJvP322yxfvpzly5fzr3/9C1tbW8xmM6+//jqLFy8mKyuLe++9F09Pz0GjZhD/0xg85TQABpoURowYgZWV1VXREfo9aX7rb1er/P9kPv/pyflqlvef7EOtVktwcDCWlpZizNaV5HkxjbKrVceBUF5ejq+vLxKJhM7OTsaPH4+FhQUqlWpAdXKJRIKlpSUmk4nx48czfvx4ZDIZubm5VFRUMHr0aIYMGdKrftHR0WRkZFBbW8vixYtxcnLizjvvJDMzk8mTJzNr1iwxSHfMmDHExMSI26729vbcdtttvfKrqqpi4cKFFBcXU15eLp568/DwIC0tjenTpxMaGoq3tzcFBQUoFAqeeeYZjEYjt956K9OmTaOhoUEUlpVIJCINQnZ2NjExMYSFhREcHMzevXuRyWTccMMNuLq68vTTT3Pq1ClGjBiBt7c3t912G1OmTKGxsXEwfmYQg2DQQzOIQQziT4ZLeYiu5PrLmd7y8/NpamrC29t7wBNSv6UOl1OnC3nR+qYfCFeDiXwQg7jWcE15aK50Ivw95ZjN5itSse6Z9lLXXmxL4krzuhL0vK8r2Tr4rXUQ0ppMJnG1+lvwW+p9sbwGqs+FXjB/hLzBHzWOr3Y7wcBbZf9pz9zv9XjJ5XJcXV178fNcKa7Ey3U588aV1GPQkBnEILpxzQQFm81m8eimXq//Q8spKiri559/FrlVdDod33//PceOHbtk2pKSEt59991ep176XlNVVcV7773Hd999x8aNG4mPj+/FVSKgtbWVjz766JIBvJdzT8IL6bvvvrus/Hqm2bNnD6dOnbricvV6PXFxcXz22WcUFxdfcfqedVi/fj0lJSVXnEff/E6dOsX3339PXFxcr5e0Xq9n48aNxMXFUV5ezrvvvotarf5d5QllJiQksH79evG71tZWPvzwQ+rr6393/j1RXl7OmjVrrkpeCQkJHDhwQGwjrVbLV199RWZm5lXJ/z+JkJAQwsPDBw2DQQzivxzXjEED4OLiwpkzZ8RjrJfzd6FrBQz029q1axk1apRISa9UKuno6KCsrOyiZQE4Oztz9uzZC0oaAHh6enLmzBn8/PyYO3cuH374IUlJSWJ9hDrb29tTVFREW1vbFd9Dz7/m5mYx/wULFuDh4XHJNFqtlsOHD2M2m4mJiSEiIqJf/QYqu2ed8vLySElJ4Z577rmsMvv+Cdw0APPnz8fLy+uS/XqhdjKbzbS3t7Nx40ZuuukmRo4cCSDmo1Ao0Gq1lJSU4OrqSkpKyoCSE1f6B92Ea6mpqWLb2NrakpGRQUdHx2XX/XL63cPDgwULFlzWmBiojJ79GhkZydixY8XflEolVVVVV90IG8QgBjGIy8U1s+UkkUiQy+Wihs6JEyfQaDTU1NSwdOlSvvvuO6Kiopg0aRK//PILCxcu5Ny5cxQUFBAcHExQUBC//PILMpmMSZMmMX78eMxmM7m5uRQVFdHY2CgGLZ45c4bIyEhuuOEGcVWnVCqBbobgxMRECgsLGTp0KOPHj2f//v1IJBI8PT2JjIxEJpPR2dnJ9u3bGTVqFIcOHWLUqFHExMTwyy+/sHjxYpGczdLSEplMhk6nw2AwcOrUKYqKirC2tmbZsmXI5XJMJhM7duzA2dkZT09P9u/fz1NPPYVcLhcNh+LiYurq6li6dCnQTahWX19PREQEiYmJ5OXl4erqSnx8PNOmTePMmTPI5XKWLVvG9u3bmThxIqmpqWi1WkaOHMn58+f5+uuvcXd3p6CgADc3NyZOnMihQ4cwmUw0NjZy0003cejQIdrb26mpqSEqKoqFCxcikUjQ6XQcOXKE7OxsysvLSUtLY+TIkRgMBlJTU7nhhhvYsGEDQ4YMITs7m2effRaz2SxKH4SGhrJ37160Wi2urq7ExcUxadIk/P39OXToENDtVZk7dy4//PAD3t7epKWliUfJBd2iSZMm4ebmhtncfZItOzubtLQ0Ro0axa+//gp0s/cuXboUpVKJXq8Xxxl0M0RnZmaiUqkYOnSoKIcwefJktmzZwpNPPklaWhq+vr4YDAZyc3OpqanhgQceoKamhuzsbJHrRYCw3bV7925KS0u588476erq4tSpUzzxxBOcOnUKLy8vkd6/paWF+Ph49Ho9w4cPJyoqCug2xuLj45HJZNja2mJhYUFSUhIPPvggubm5FBYWUlBQQHR0NL6+vsTGxhIQEEBOTg4zZswgISGBWbNmMXr0aNFj1dLSwrJly0hOTkav17Nw4UKOHTuGTqejpKSEGTNm/PEP+yAGMYhBDIBrykMjwGzu3n4aPnw4hw8fpq2tDV9fXxISEujo6MDa2prExESys7Nxd3dn7dq1GAwGDh06xA033EBYWBjQfZzyk08+YerUqUyZMoU333wTDw8PPD09mTx58oAU5gcOHKCwsBBXV1fWrl1LTU0NWq0WHx8fNm3aBHS/II8ePcqUKVMYMWIEfn5+nDhxArVajbW1tXiSY9++fdx777385S9/YerUqWRmZnLkyBG8vb3ZtGmTyF8ilUrRarXk5ubi5uZGUlJSrxdkfn4+gYGBlJeXk5KSwtq1awkNDWXq1Kk0NDQwZMgQQkNDCQoKIjs7m+bmZiIiIoiPj0er1WJtbU17ezuWlpbY2Njwyy+/EBAQgI+PD8OGDaOuro7CwkKOHTtGfn4+CxYsoLq6mh07doikgUuXLmXz5s1ivZRKJZGRkQQFBTFq1CiampooLy/HxcWFpKQkrKysOHbsGBMnTqSrq4u0tDTWr19PUFAQ06ZNExlfhw8fjo+PD3l5edTV1bF582bMZjOLFi3i0KFDnDt3jnPnzuHh4cHw4cPZvXs32dnZImeKYJhIJBIiIyPx8PBg4sSJ7NixA4PBwKJFi4iPjychIaFfX+v1elauXMm4ceOYPXs277zzDs7OziQmJuLl5UVqaiqdnZ20t7djY2PD2rVrcXNz4+zZs5w6dYovvviCyZMnM2bMmH6kaGazmQkTJnDjjTfy7rvvEhwczKlTp+jo6KChoUE8HQRQW1uLQqHAzs5OlNkA6OjoYPPmzfj5+eHk5IRcLhcZaj///HNGjhxJZ2cnGo0GGxsbTpw4wZw5c2hoaKCzs5O5c+eydetWDh48SEVFBQsXLqSoqIjdu3fT0dFBeno6ubm5JCUlMXfuXPz8/Aa3bQYxiEH8n+GaNGgkEgmBgYEUFRWJjK7z5s0jJyeH+Ph4oqOjycnJwc/Pj+nTp/P5559jb2+Pvb09fn5+ODo6AtDe3k5FRQUKhQIfHx90Oh2dnZ0AFwxiFYQjZ82aJTLPQjfhmsCq29XVxfbt28U08+fPJzs7m6NHjxIdHS3mP2fOHPz8/CgtLUUqlVJUVISTkxNjxozh22+/FTVphHu+EIYMGcL58+cxGo10dnaSkZGBnZ0dQUFB4haERCJBJpOJBHDR0dEYjUYOHjxIREQEHh4etLe309HRgVarFQNLpVKpeJw9Pz9f/C4qKoqioiKUSqUoQ9CTWbdnYOpA/5fL5VhZWeHo6IitrS2dnZ2kpaVha2tLYGAgixYtEvtAJpNhZWUFQHZ2tiicGR4eTllZWS/JgK6uLtET9Prrr4txMEK5wl/ffAaKz9FoNBQXF4vSFlKpFDc3N5RKJceOHSMgIIAdO3agVCrR6XS0t7czbNgwVq5cyZAhQygrK8PGxgZra+t+Bo1cLsfDw4NRo0ZhNnerbk+ZMoWffvpJlD0Q4OrqSkdHB2q1uldslo2NDfPnz+eZZ56hsLAQS0tLpFIpUqmU4OBgEhMT8fDwIDIyEgsLC6ysrLC1tcXW1hY7Ozvs7Ozo6OggNzdXTBcZGSl6CHu2k1Qqxdra+oJjcBCDGMQg/mhcMwaN2WwWqdlVKhWrVq0SWWfb2tqwt7dnwYIFHD9+HF9fX0JCQti5cyc1NTWcO3eOzs5OkRZfeOk6Ojri7u5Obm4uzc3NDB06FHt7e/R6fa+Xs9lsFr8LCQlhx44dYr5Hjhzh3LlzWFlZ0dnZSVdXFzY2Nvztb3/jX//6F01NTdjb2zN//nxOnDiBn58fJpNJNBpef/11jh07xs6dOwkICODgwYOiyrBKpRLrbG1tTU1NDeXl5bS3t4svNrPZzKpVq7CyssJoNKJWq/Hy8mL16tWkpaWRkpKCUqmktbWVtrY28T4UCgU33HADBw8eJCgoiJ9//pmamhpkMhlqtRqzuZsYrbm5Gb1ej16vZ+zYsaSnp9PV1UV9fT1Tp07tJTkgxJz0bDPhOxsbGyoqKigrKxPrL1DuC20dEBDAt99+S3p6OmfPnkWhUNDc3Ex7e7tY74kTJ5KUlCQan2PGjBGlAwSphNLSUh566CFmzZrF6dOnxfoIdTSZTEyaNIlTp06h0+no6OgQNYN6yjtYWFgQHBxMWloabW1tBAYG4uXlxaJFi9i+fTtPP/00cXFxDB8+HCcnJyoqKkhNTaWoqIjW1lZMJhMpKSnU1dXR2traS8dIIpGg1Wqpra1lwoQJ2NjYcOutt7J3715RFFLArl27qKioQC6X09HRIebT3t6On58f77//PnFxcb36wM/PD4VCQUxMDN7e3mI/9Wwn4bvx48dz7tw5NBoNjY2NTJ48Wey7IUOGkJycTHV1Nc3NzTQ3N/eK4xnEIAYxiP8UrhmDBrpjCebOnYtWq+W2224jIyODm266idbWVgDGjx/P9ddfj1wuZ8GCBcycOZMDBw4QFBSEwWBg9uzZNDQ0iPkplUree+89ysrKyM/P54UXXkClUjFjxoxeE7dOp8PT0xN3d3cWLlzIxIkTiYuLE7d1PD09xfLLy8uZP38+ISEhzJ49m/T0dIxGIxMmTOD6669HJpNRX1/P3LlzMRgMODo6snLlSlpaWvD19eX+++/nwIEDuLi4oFAomDBhAgaDgWnTpuHu7k5jYyMPPPCAqGwtkUi46667KCgoYMaMGUilUp599llsbGxIT08nIiKCcePG4e7uTlVVFSNHjhSPDs+ePZtHH30UhULB3LlzMRqNeHl5ERQUJMbMnD9/XtyGGzlyJPfdd5+oxDxz5ky8vLzw8fFBpVIxZ84cOjo6ADAajchkMiIiIlCpVCxZskR8od5xxx2iMrigU2NnZ8dTTz2Fg4MD586dIyIiglmzZqFQKGhoaGD48OEiQdmkSZM4evQoixYtIigoiMmTJ2MwGLCxsWHYsGEolUry8vIIDg5mzpw5vcbPrFmzqK+v54YbbmDq1KkcPXqUBQsWEB4ejouLCz4+PtTW1jJ37lzUajVvvvkmra2tpKWl8fzzz2NlZcWMGTN4+umnGTJkCK+99hq+vr6ipEBycjKtra1ERUWxfPlyTp8+TWdnJ4sXLxbHk1Qq5aGHHiIzM5Oqqioee+wxpFIpHh4eLF26lICAgF7jftasWaJ3KDQ0VNSGUiqVNDQ00NraysMPP0xHR4c4duvq6khLS2Pjxo2sWbOG9vZ25s6dS0NDAyNHjkQqlaLT6ZgyZQojR47ktttu4+jRo4wePZpJkyZhaWlJeHg4w4YN484772TXrl1MnDiRoKCgP+DJHsQgBjGIS+OaJdbreVt6vZ6dO3fS0NDAPffcI24p9bzmcjhhfgs3y0DomY9OpxuwbpeT9+XS21+sPhfjPem5PXSpPC5Vv0vhYvW/UN6X0xaXQ93/e/v+YvW6WH2vpM/OnDlDSkoKUVFRTJ48+bKkFC6Uf3l5OR9++KGo+dTY2Mi0adMuKm54sba43PYcxCAGMYg/EtfMKae+6DmpCnERM2fOvOy4kyu55vekEeo2a9asXnW70rx/D7nY5eR5ufd0tYjJLuea39IWl1vW77nucut1uX0mHBv39PRkwoQJv7sdvL29Wbp0KcnJyYSGhjJhwoRLKjX/XvK63wqTyURWVpaoOi6TyTCZTAQFBYlaSH80zGYzFRUVVFZW4unpSWBgoNheQsC6XC7H3t6ewMBAMjIygO6FlEKhwGg04uHhQV1dHSEhIbi4uIgGZ05ODjqdjmHDhvWTrLgSmEwmiouLMZlMhIWFXdCQb25uJjc3F7PZjLOzM05OTnh7e/e6T19fX9FDl5ubi42NDcHBwZjN3RxcarWaiIgIMW7t98JkMlFSUoLBYMDLy4ucnBxRgmIgaDQaTp8+zeTJk8Wg/v8UhGD4UaNG9bp/g8FAdnY2jo6O/TyofyQMBgMZGRn4+vr+x56HPzOuqS2nC0GhUHDrrbcyfPjwP93qUajbsGHD/nR1G8SfAxKJhJEjR7JkyRIUCsXvzk8ulzN79mzuu+8+pkyZctVeTH8EdDodv/zyCyaTibfffpusrCzKysoum8ixs7PzsrhxdDod1dXVF/398ccfp7Ozsxc78pdffklVVRVSqZSvvvqK2tpaDh48iE6n47nnnqOhoYHU1FRKS0t56623eO+990SPVn19Pffeey9xcXFX5cV87Ngxdu3aNeBvZnO3svnLL78MdBu1e/bsYf/+/eI1bW1tPPHEE1RWVgIgk8nIzMxk2bJlYsB/U1MTJ06cQC6/umvh48eP8+uvv6LX62loaOjn9ROMsfb2doxGI/X19f8nsVparZb3338flUrVr34///wzJ0+e/I/Wx2Qy8e2335Kfn/8fLffPimvOQ2MymVCpVFhZWaFUKnudnrkSCEGmlpaWV7x1otVqUSgUlz1JXU7+QtBzXxE6YcUqkUiuystOKEuj0WBhYXHJbYjOzk6srKx6XScQsmm12itqP4HcTjjhBPS6Z5lM1uuYdc90Wq0WmUx2VdtACEYWOIauJoQTb5e7KhfuVwgWVyqV/dqrb5vo9fqLjoue15tMJrq6usTTS38WyOVyHnnkEVxdXbGxsWHIkCFMmTKF2tpa8Vj8sGHDcHR07Fdvo9HIV199BcCDDz5IR0cHeXl5+Pr6EhwcTGlpKe3t7Xh4eHDw4EHy8/N57rnncHFx6ZWPRCLB2dkZa2trXF1dexk0BQUFKJVKbrzxRh566CEcHBx47LHHUCgUKJVKhg4dyowZM9BoNEyaNIm9e/dSVlZGYGAgycnJ+Pv74+rqOqCBYDZ3q3sXFRWh0+kYOXIkNTU1qFQq6urqCAgIICQkhLa2NgoLC0UixoFgMBj417/+xZ133iluWT7xxBOcP39eLOvcuXN4enqyY8cOnnnmGWQyGcHBwcyYMYMXX3yR9evX4+PjQ1hY2IBjymw2U1ZWhkqlorKykoCAAIKCgsjLy8NgMGBvb4+zszO5ubm4uLgwbNiwfnWXyWT4+/sDoFKpKCwsxGQyER4ezmuvvcZ1113H1KlTGTJkCIDYpxqNhqioKCQSCXl5eSgUClpbW5kyZcqAbavRaMjKyhLJSUePHk12djbe3t4MHTqUhoYGzp8/j1QqJTo6GrlcTkFBAW1tbWIeLS0tZGdnY2try4gRI0Q+K51OJ/ZZeHj4RRcMnZ2dJCcnEx4eTmVlJZaWloSEhJCXl0dUVBRlZWWUlZURFRWFm5tbvzGuVCpxcHCgrq6OPXv2MHbsWNzc3CgsLKS1tZXg4GC6urooKSlhwoQJVFVVoVAosLe3Jzs7GxsbG6Kjoy/pof1vwbVxF/8/zGYze/fu5eDBg2zcuPE35SGcAklPT2fLli1XnL6pqYmnn376N1H5XwrffPMN3377La+88gpvvPEG33zzDV9//TVvvPEGKSkpV62ciooKHn30UZqbmy96ndFo5JNPPhH5cHoiNTWVl19+eUDJhoEgkCEeOnSIL7/8UgxqLi0t5d577+W7777jp59+4t///jdFRUX9Ju5t27aJrMEXq6+Q76Wg1+v55JNP2Ldv34B1/b3yGidOnBgw70shLi6OI0eOYDZ3SyYI7SUYfULdNBrNFY2LlpYWPv7448vur74Qyr3aq2a5XN7PlS6Tydi2bRs6nQ6pVMrf//53NBpNv7TCSTE7OztaW1vZvn07crmc559/npKSEr7//nvs7e1FYkMrK6srOnoukUh4/PHH2b59Ow8//DBWVlY4ODjg5OTU6zpra2scHBzw9vZmxowZbNq0STwheKntia1bt9LW1saBAwdISUnhyJEjfPHFFxgMBt5++206Ojr4/PPPcXV17cXk3BcqlYq8vDwiIyORSCR0dXVx/PhxCgoK6OzsRKfTUVFRwfPPP8/+/fvFgxQSiYS7776bYcOG8dZbb11QsgUQCT7XrFmDyWTi+eefJysri+eee46ioiLOnz/Pxo0bUSqVvPLKKxQXF/P555/j4uIi1r2oqIj33nuPzs5O1qxZg5ubG7GxsRQXF9PR0YGrqyv19fX885//pKuri88++wy5XI5Wq+Uf//gH9fX1vPzyy7S1tbF27VpycnIGrKtGo+Gf//wnxcXFpKSk8MUXX2BhYcG7775LfX09n376Kb6+vpw6dYq1a9dy8uRJkpKS8PT0RKPR0NnZyfr167GwsOCtt96ioKBAzDs5OZmUlBS0Wi0tLS0X7V+5XM6PP/5IRUUFWVlZ7N+/H4PBQE5ODqmpqSQnJ9PY2Mirr756QUV1s9lMTU0NjY2NfPbZZ5SUlPDrr79iNBr58ssvkUqlrFixgs7OTvLz81Gr1WLd//Wvf11T3p1rxqARVqvbtm1jxowZLFu2TNx37+zsxGQyodfrRT4Q4Xi20WgUj/12dXXxzTffoFKpiI6OZtmyZUD3g9re3o5GoxHTaLVa1Gp1v0ncxcVFtNIFD4aw+hCObQurYeEFq1KpxOv1ej1arVYsqyeio6N5+OGHcXJywsPDg0cffZRRo0Yhk8no6OjodV/C8fWex8uFdtLr9ahUKrHuBoOhV7t4enrS1dWFwWCgq6tLbD/hOwEymYynnnoKJycnkV+nvb0ds9mMl5cXjY2NYvqBXMidnZ2o1WpRD2vTpk0MHz5cPNUjkUjw9/ensbGRKVOmcNdddxETE8Pjjz9OeXl5r/yWLVtGdHS0WA+hzgIMBgM//PADlZWVYv9cqH2EY/NWVlbi/Wi1Wtrb2zGZTOTn54vEeyaTCbVaLfaX0H9dXV3i0eb29naxLbu6ujCbzUyZMoV58+aJeWu12l5SBz3HTc+6zZkzh2nTpqHX6/npp5/E9hIYspubm/n2229Ffp6+40Lo+570BABOTk4iu3TP50R40Wi1Wjo7O8V2FWgONBoNGo0GtVrN119/PWBfX210dXVx9OhRQkJCGDNmDDU1NdTU1PS7TiqVYmNjg6OjI9nZ2TQ1NaFQKHjiiSdwdHSktbWVzz//HF9fX2xsbLCzs7vs7TdhnERERLBz5058fX3561//2msFP1B9/vKXv4iLLuHZvRiuu+46APFIvLu7O/7+/sTExKDRaCgpKRG9IUOGDLmgd00qlSKTyUSDxNLSkqqqKhISErC0tCQnJ4eSkhLxRXzw4EGxH5VKJa+++iq1tbWsWbPmgv0rk8nw9PRk6NChLFq0iNDQUFpaWnBxcWH+/Pk4OztTUVEheoeam5spLy8XPTlSqRQfHx9kMhmVlZVUV1fj5+fHyy+/TFRUFFZWVjg7O+Pl5YWlpSUqlYqkpCSCg4OZMGECxcXFGAwGnJycGDt2LP7+/gMutqCbo8nBwYERI0YQFRWFvb09o0ePRqPRkJ6eTnNzM97e3syZM4dDhw6xc+dOhg0bhre3N05OTpSXl4veo0cffbSXEevr68vOnTs5d+7cJQ95KBQKli1bRlxcHHZ2dpw9e5aMjAzGjh1LfHw8JpMJPz8/7rrrrgvmIWxJT5w4kfr6enx9fZk0aRKVlZXU1tbi7e3N+PHjOXbsGHq9HqlUSn5+PiaTiUceeUScv68FXFNbTsXFxZSXl5ORkYGFhQX79+/nueee47XXXuOll15i3bp1uLi4UFlZKR6T/vXXX3F2dubYsWPccsstbN26lbFjx1JbW0teXh7PPfccmzZtEq31hQsXkp2dTVZWFs7Ozmi1Wl599dVeE5MwqSQmJlJVVUVlZSULFizg6NGjdHZ28tRTT7FhwwYWL17M6dOn0el0JCUl8cQTT/D+++8THByMlZUVjz32WK/7mzx5cq/PcrmciRMnsnfvXhITE0USvFtvvZUDBw7Q2dlJUVERr7zyirht0tzczI4dOwgKCiIhIYFHHnmE9evXI5FIUKlUBAUFcc8994hBixs3bqS+vp4XX3yRn376iRtvvBE3NzcAGhsbeeONN/j73/9OdXU1FRUV6HQ6tFot8+fPp6GhgS+//JLU1FTefPPNXnpPCQkJ1NXVice4J0+eTHFxMRkZGXh7ew8Y1C2TyRg3bhxDhgxh3759PProo0D3Nsxnn31GdHQ0ERERpKSkIJFImDhxohjw2Nrayo8//ii6hQ8ePIi/vz+nT5/mmWeeEVmfjUYjW7duxdPTk4yMDKZNm0Z5eTmpqank5uYSFhZGZWUlubm5TJ06lby8PDo7Ozl8+DCPPPIIq1atIjAwUDSC3d3dRU9XcHAwCQkJvPPOO+zZs4euri7mz5/Pm2++yaxZszh69Cj/+te/qKuro6amhuPHj3PnnXcSHR2NRCLBYDCwfv16rK2tmTp1KiUlJaSnp4vtZTabSU1NZffu3cyaNQuz2dxrXCxZsoQjR47Q1NRER0cHTz75pGgI5ebm8sUXX/DOO+/w0Ucf4enpSVlZGZMmTcLT05M1a9YwYsQIkpOTefvtt9m6dSvTpk3DbDZz6NAhFi9ezM6dO5k6dapIDnm1IRh7gixIdXU1YWFhuLi4DMjaLYwdo9GIhYUFNTU1jBgxQozFeOaZZzhy5Aiff/45Y8aMwWQyidu3A53kEiZ9YfF05swZmpqaWLZsGS+++CIJCQk0Nzfj6OjY63ohjV6vJyQkhMjISA4cOMANN9wgXmc2mwcs88svv+TGG28Utc76/i6TyaiqqqKzs/OiHkh7e3vmzJnD7t27CQ8PF0k0LSwsMJvNnDlzhieeeAJbW1tcXFxYtWoVN954o7jla29vz4cffsitt9560f4VxqHZbMbS0hIvLy/xN0tLS6qrq4mIiKCrq4uioiKqq6sH5AATqBXq6uqQy+XiQkq4RtiuNZvN1NXV4eHhIW5L9m2jS6HviUBHR0dqamrQaDSiQaFQKCguLmbkyJEYjUbkcjk1NTWEhoZiMpl6LT6USiWffPIJK1asID4+nokTJ6JUKnuRYfYse8qUKXz33XcEBgYSHh7OoUOHePXVV1EoFLS1tTFmzBhKS0tF2ZWB7q9n/U+dOkVSUhKLFi3i6NGjQPeC7+9//ztvvfUWlpaW1NXVERoaitlsvioiu38WXDMeGolEQmhoKG5ubkyYMAE3Nzdqa2txcHAQA/mEa+644w6OHz9OQkICer2e2bNns3jxYry8vHB3dyc6OlrkGxEEFGfOnMm0adP4/PPPsbOzw9LSkgcffJD09PQLumHt7OwYOnQoarWarKwsrrvuOs6ePUtLSwuOjo60tLRw4MABLC0tRUNArVZz3XXX8be//a3f/fWdaCUSiejJmDBhAvfffz8nTpzg1KlTnDt3Djs7O5qamnqtGn/++Wesra2ZPXs2lpaW7NmzBwsLC3x9fbnvvvtITEwUJ0WpVMq8efNIS0ujublZnOwEODs709XVRUdHB19//TVTpkxh6dKl/Pzzz1RVVeHg4MDTTz/Nddddx48//ig+dFqtlm+++Ybp06ezdOlStm7dioWFBR4eHowfP/6CLyfhnl1cXHoF5VlYWGBjY4NKpaK+vp7Y2FjCwsJ6nRpzdHTE09OTUaNGsXnzZnx9fZk7dy4qlUrUfgIoLCwkIyODGTNmiAKVGzZsoK2tDU9PT/Ly8vD39yc4OBgLCwu2bNkiMhU3NTXR2dnJrFmzeOqpp1AoFAQGBnLXXXeRn5/PrbfeikKhoLq6GldXVxobG3F1daWtrY3Fixfj6+tLVlYWa9euxWg04u7u3stlLpfLcXFxoampCX9/fzw8PJgwYYLYXhKJBD8/P7y9vcWTLj3HxYEDBygqKsLZ2ZmysrJe2zQ+Pj7U1dWhVCqRyWSEhYVx2223kZCQgKenJ1qtlscff5zJkyeze/dupFIparUaDw8Pamtr8fLyEuUl/ogYnIqKCpRKJcXFxZjNZl566SUOHz5MfHw8d9xxB66urgOmi4qKIjk5GU9PTywtLXnuuec4ceIEFhYWHD58mNDQUGbMmEF4eDh5eXm9tg4ECFshjo6O7Nq1i02bNvHuu+8ilUrJzMxkx44d/PrrryxduhQ/Pz/MZjPFxcU4ODiQl5eH0WikrKyM7OxscTv37rvvRqVSodFoqKurE7mDekIikeDo6CgGRWdkZNDU1ITJZKK2tlb0Kk2fPp3333+f3NxcjEZjL4JGAVKplH/84x8olUq++uordu/eTUVFBXPnziUnJ4fS0lIsLS2xtLTEz88Po9HI0aNHSU9PJz09HYPBQEBAAJ9++uklT9Okp6cTGxsrPstWVlZUVVURHh5OcHAwzz77LHv37mXEiBHMnDmT999/n5ycHEwmE+fPn8fS0hJra2vmzZvHs88+y86dO7G3t2fkyJHs3buXgoICrKysUKvV4hbZsWPHeOCBBzAYDNja2lJbWyt6vwcyalQqFdbW1tTX19PW1oZWq6WmpgZbW1ucnJy48cYb2b17NxkZGTz++OM88MADHDlyhHXr1mFnZ4eFhQXjx4/nueeeY/v27dja2ore0OLiYjIzM5k+fTphYWH88MMPrF69+oLGlb29PTfccAMTJ05k6dKljBs3TvTcJCUl8dprr1FVVTVgvJ1Wq8VsNtPY2EhdXR0KhQKJRCJuWXV2dlJTU0N4eDhz5swhPDwcPz8/YmJiePbZZ9m6dStOTk5/qri534NrykPTEwN1kFQqRaFQIJfLMRqN1NTU0NbWJgZ+9XxJCkFSwl632WzG09MTnU4nBloKXpkLDdTa2lq6urpwdHTEZDIRGBhIQEAA69at44YbbqC1tRULCwvmzp3L7Nmz0Wq1oiCl4FHpu3IbaJUmBIkKQpX19fW4urqycOFC5syZ0yu9YJgA+Pn50dDQ0Ot++q7wfH19GTp0KGvWrOH666/vFzwmlUrFUwc6nQ5LS0scHR3RarVinUJCQigsLBTTGI1GGhoa0Gq1ODo64uDgMOAk3BdCsHJeXh7PPPOM+H1PQy8yMpJ58+bxyiuv8PbbbzN69Oh++TQ1NaFWq5FIJPj6+opeIoC6ujrxN6F/q6urWbRoEZGRkXR0dIgq44IMxKxZs5g7dy4dHR1i/wkB1YI3QS6Xi/IBgttXgBDsLJfL0Wg0NDU1MXnyZBYtWtRr9SQYsJeLvuOipqaGgIAAFi9ezPz588XVXk/DWPjc8zkR8pHJZAQGBvZ66Q809gfyNvxe+Pj4iJ5EgVBy5MiRoifgQuVNnz6dCRMmYGVlxcqVK8WtRIlEwl/+8hcMBoO4zfTJJ59cMK/Ro0fz888/i59vvvlm5HI5kydPFo0IOzs7pFIpZrOZ4cOHs2PHDrFtAwIC+PDDD3sFtptMJj7++GMkEskFTw0988wz4vFv4di62WxGKpWyevVqMSaos7NT7KOBxohEIsHBwYFnn32Wzs5O9Ho9CxYsEPv4zTffFOsQFhbGzz//3Ksd5HK56PW8WCyH2WwmKiqKGTNmYGVlhdlsZs2aNeJ9v/XWW2g0GlGG47nnnutVd7PZzLhx45DL5TzxxBP89a9/Fa998MEHMRgMyOVyvvvuO+RyOQEBAUycOBHoXtiYTCZWr16NTCbjzTffvOC4cHJyEuNLIiMjxTb95ptvxGDorq4uFAqFGAC9atUqzGYzf/3rX5HL5YwaNYquri7xUMSzzz4LdD/PQkC+Uqlk6dKlFz1pJ4xFYewIHll/f39xcdM36F+AEAcjzIExMTHI5XJRSPm2224T5Wwef/xxZDIZEomEF198ka6uLlH8+FrBNWPQCK6z9vZ22tvbsbW1pampifj4eJqamsTtDUEssLOzk3HjxvHkk0+iVCpxdXVl+vTpGI1GMjIyaGtrQ61WixH92dnZlJSUcPPNN4sxGmq1ms7OTrRarejqFKjy1Wo1sbGxhIaGiiswg8HAPffcw/vvv88zzzyDWq0mOzub1atX4+joyOTJk+no6KCtrY2Kigr27t3Lww8/3Gsg6/V6mpub0Wg06HQ6UYpAuO+Ojg4iIyNZs2YN9vb2WFpasnDhQnHSXrBgAZ9//jlVVVXU1dWxcOFCduzY0et+hPZRq9W4u7tz9913889//lN8YPvC0tKSBQsWEBcXx8yZMxk6dChDhgwRX6KFhYXcfPPN4vVWVlbMmzePuLg4Jk+eTGRkJM7OzqjVatra2sQXouDa7+jooLi4GL1ez6FDh5g5c6Y4iUG3gaRSqbCwsCAjI4PQ0FDuuuuuXpOIRCLBwsKC1NRU5s2bx549e8T2FmKlAIYOHUp5ebnozXB1dWXcuHG899573HTTTdjZ2WFvb8/58+fFWKdPPvmEgIAAoqOjUavVqFQq0Q0txMII/xfaVRhfQlyVEH8kGL7Lly8nJiaGwMBAJkyYAHS/ANva2sQ2aW9v79VeQtu2tLSQk5Mjjgnh+jFjxrB8+XL0ej1KpZIbbrih17Mj1FP4k8vl4lZGU1MTlZWVFBQUsHDhQk6ePMnJkydpbm6mpaVFjLHKyspixIgRV32SFIzCvuPuUuipMdU3D6VS2esE24UCgi92UkxI3zP2ZqDrB6p/T+PmQuX2TNc3fc/PA21nXCjPvtsyA+V7IQPrYsaX8Bz21LsDet2jTCbrVb5EIulX957597z2Qm3Rcxxcqk0FSKXSS55e7NtOfceboGIvoGd+Pevg4eFxSeHWgU5vCnPWxSAYTX0hjOWe47JnmwnxZdcarhmmYGEvtbi4mCFDhuDp6UlaWhpmsxlra2vc3NwoLy/H1tZWpMsfMWIEpaWllJWVMX78eBwdHUlLS8PCwgK5XE5LSwsjRowQY1GcnJwICgqiuLiYzs5OvLy8KCkpYdiwYTg4OCCRSMTtJQ8PDxQKBQUFBQwZMoS6ujoxNictLY1FixYB3ayt2dnZREVF4ejoKKZ1cXGhuLiYUaNG9Rrk7e3tZGdnI5VKRRp/QWzS0tJSjBMQjvuNGTOm31HT0tJSGhsb8fb2xsPDg+zsbBQKBba2tlRXV+Pv709paSkBAQF4e3tTW1tLcnIyixcv7pWPEFT26quv4uPjQ3Z2NhKJhJCQEKytrcnLyxOVznuqQ0O3qzQ7OxuZTEZISIhI4tWTuEzo08LCQvHl4+npiYeHR69VqE6nIzs7WxR0FNSng4ODe3m6ioqKxIDvkpISVCoVfn5+uLu792ufwsJCPD09cXJywt3dnTNnzmA2mxk7dixms5nk5GRGjhyJXq/n7NmzhISE4O3tTWZmJm5ubvj5+YnHIq2tramsrCQ0NJSSkhJcXFzEYFpfX1+Ki4sJDg6msrISe3t7PD09OXXqFC4uLowYMUKciAwGA3l5eej1enx9fcU69iR6MxqNnDp1Ci8vL5qbm3uNi6ioKAoLC2loaGDcuHHimO357AQHB1NVVYW9vT0ymYzGxkbs7e35xz/+wYsvvoinpyf+/v50dHSQnJxMYGAgOp1O1LRycXEhKCjomnFhD+LyIRwwgP/n0RnEIP6TuGYMmj87TCYTn332Gc3NzTz99NMX3PP/M8FkMvHll19SXV3N008/jYeHh/hbUVERZ8+epbi4mGeeeeaatPYHgUhh8K9//Yvvv/8eGxubwRfVZUCI37C1tb3qJHR/BpjNZtra2rC2thbjNgYxiP9rDBo0/yGYzWZKSkqwsrLCy8vrN00Aer2empqaS7ovrxYEb4VCocDHx6dXmZ2dnRQUFODv739NBZUNojfMZjOVlZXiiY6BCOz+CKhUKlJSUoiJibkqTMZ6vZ4TJ07Q2trK4pOw6+kAAQAASURBVMWLxXiNuro6Dh8+zJgxYwgPD78q92Yymfjpp5/w8/MTT7iYzWbq6+uxsLDAwcEBtVpNTU0N9vb24gmmyspKtFotfn5+WFhY9CM+zMzMxMbGRtwu8fb2pr6+HpVKhZeXF7a2tmi1WiorK1EqlaKEQU8ItBOVlZX4+PigUCjQ6XSi9ICwxd7e3i4eqnB3dxfrp9PpxFM/RUVF7Nmzh4cffviiCxq9Xk92djbV1dXMnj37N0k8NDU1ceTIEQIDAxk7diytra0cOXIELy8vYmJirjoxnNlspqGhgaamJlxdXXF1dcVgMFBeXo5UKsXf3x+pVEprayv19fW4uLiIkha/ByqVirS0NCZMmHDBdjIYDCQmJtLQ0MCSJUuQy+VifePi4hg1ahQRERH/k3PyNXPK6UogvKgTExNJTEwkKyvrDz+HL5FICA4O7nck+XJhMpk4ffo0586do7Cw8D/CGyCRSBgyZEi/7SLo3qMdOXIkzs7O/5MPzv8KhJNT48eP/0MM1/b29gEDJvV6PStXrqS9vf2qlCMEa//4449ioLXBYGDfvn3k5uZeVCVcOHJdUVFBfn4+5eXlF33+jEYj586dY+zYseKWp06n46OPPqK0tBStVssPP/yAra0tBw8epKioiJMnT4oaSz1PBAoQiOY+//xzPvzwQ9avX09hYSGHDx/GxsaG9evX09XVxU8//YTZ3K0RlZCQ0C8f4WX4+uuvi0eNDx06hFQqJT09nc8++4zOzk5+/PFH7OzsiI2N5fz585w4cYL8/HwMBoNIWhoaGkpnZycNDQ0XbbujR49y/vx5JBLJbyakdHBwoLW1lY8//hiNRoO9vT3QHfP2R8w/5eXlbNmyBSsrK/7+97+LdBcdHR1UVVURGxuLSqVi06ZNODg4sGPHDmpra393uVqtlo8++kiMQxoIMpkMa2trNmzYIB5kMRqNHDhwgPT0dIKDgy+YVhjLlZWV5OfnU1ZWdtlko/8NuKYMGoH742KTjU6no76+HolEwtChQ4mMjMTV1ZW6uroB2UaFfeHLNSCEFVBHRwcmk6kfgdlvuaeGhgbq6+sJDw/nxhtvxN3dnbq6ugGPJArl/xkcbwJBXF+Oif8mCLFCfYn6fks+Go3mdzMMC3kJJ1X+SAj3fjFK/d8Dg8HAypUr2bJlCyqVivz8fJKSksRYNxsbG4xGI8nJyRQVFaHRaDh37hxxcXF0dHRQVlZGSkoKBw8evKxFiV6vJyAggJaWFsxmM+Xl5dTU1DB58uSLbpu0traydetW8vPzaW1tvWR7CL/1jMvKy8sTV9KFhYU0Nzfj5eWFv78/Bw4cYNeuXQwdOlSMRerLMCscu/7444+54447ePjhh4mLixOP6Tc3N3P27FnOnTvHkCFDiIiIYM+ePf3qKZfLmTRpEtbW1uJckZiYiJWVFVOnTiUhIYGsrCyRpsDX11esX0REBCEhIZw5c0akgujL4dIX7e3t7Nu3j9LSUsaMGYNeryc5OZkTJ06gVqupq6vj7NmzHDhw4KKsuoKMgIeHB2fPnhVpDS5lZPfk+Rno70IQ5AxcXFzEwP6jR48SFhbGsGHDiI2N5ezZs5jNZjw8PHB2dubEiRMD5iWQcSYlJVFRUUFHRwcnT54kLS2N2NhYtFotRqORzMzMAWkDBkJXVxfBwcHiCVxhLE+aNKmX5E9fqFQqtm3bRm5urjiWryVcUwZNY2Mj//73vy94rBC6LeB169axcuVKLC0tsbe3Jz09nZ07d/azioVtory8vCuqx9mzZ8XjiImJiRe1ti8FIcjzr3/9KxqNBolEQnNzM6tXr+63MjKbzbS0tJCcnHzRNrhaEJhyL4T6+noee+wxqqurSUxM/EMfnkvV5Ury6dt2FRUVPPLIIxdkHb0caLVaXnvtNVJTU39vFQHYvHkzmZmZVyWviyE1NZWnn36637F6gXn790A48u/i4oJarWbjxo04OjqyYsUKkaG4tLRUJE7bvn07ra2tJCQk8N1335GcnMyqVauwsLBg+fLlF6XlF4JVfX19qa2tpa2tDaPRSHV1tSgHMBDMZjNHjhxh1qxZzJo1i3Hjxl3UK2A2m0VJBuHkSn19vcghBNDQ0CBukVhaWlJRUUF9fb34IhJYrHvCysoKd3d3SkpKkMlkuLq6Ul1dLW5LSKVSiouL0Wq14pZUXV1dv5d233rLZDL++c9/4uPjQ2ZmJjExMbS2top1F+rX0NCAQqFAKpWKrNmAyP90IePA0tISOzs73N3dUSqVrFixAjc3N5qbm3nnnXdISUlh+fLltLe3X7T/Ghsb8fDw4JZbbmHbtm00NDRclsewoqKCHTt2sH379l5/Z8+evWg6Z2dnoqOj2blzp3hKVK1Wi9QfAsux0GcC8/JAaGhoYNOmTdjb27Ny5UrUajVvvPEGzc3N7N69m+TkZOLi4sjJycHb2/ui7QD/T8LF39+fmpoakem+qqqKESNGXHRsxsfHM23aNGbPnn3JsfzfiGvKoHF1deWJJ55AJpNhMBhob2/vNxHb2dlx8803U1paCnQPNhsbGx566CGcnZ17XdvY2MihQ4cIDw/HZDKh0WjEo7UajYa2tjZx4hWO0QIEBgbS2tqKUqkkKCiInTt39nvZCqvsnnkIlPk9PUUKhYI5c+agUChQq9VotVrKysp49tln+50mMRgMbN68maFDh4pcCK2treh0OrEM4Qiv4HrUarW0tbWJXiiDwSCmEbwKwtF0vV5PW1ubSI2/evVqampqMJu75Qb63otCoRC5dYKDg9mxY8cFjQ7hRSCUbTQaUavVordLIJDSaDR0dHT0epnq9XrWrl1LeXm56E0Rjk4LfSXkJfSbcL9Go5HW1lbx/tatW9dvS0GYZHr2tZC/cK9qtbqXFIIgcSDAwsIChUJBZ2cnra2tYvnCPQseLOH+dDqd2BcDaSTdfvvtDBs2TBxHgkewb5sK41JYkXZ1daFSqdBqtSLFgNDWQp2E9hPGsnA0vGe+hw4dIiEhAaPRKNZzILmOi0EikYj6R+7u7lx33XWUlJTQ0NAgHgP/7LPPmDBhApaWlsTHxyOTyZg3bx4zZszAxcUFX19fxo4dK8p3XAgCmaWvry9lZWVUVlbi4OCAVqu9KFGcsLqOj49ny5YtbNmyRdTSGgg6nY7NmzcTExODUqlEr9eTmpqKs7OzSMnQ06thMpnEo8bCdwInSt+2Ati9e7cocdA3Tc/juyaT6bLiSoQj5lVVVdTU1PDEE0+IniQhH7lcLpYlfC/kPW/ePH755ZcLbjspFAqsra2xt7dHpVKRm5uLr68vU6dOFck/vb29WbZsGZ6engPmIcQ6eXh4MGbMGJqbm0WyxEvBy8uL+fPnc9111/X6i4qKumg6QZh42bJlHD58mIKCgl4syMLx9b79OBCcnZ2ZO3cu58+fF983jo6OjBw5kpCQEOrr69m9ezfDhw/Hy8tL3E67ENra2nBwcMDPz4+ysjLKy8tF2guBGX0gCJ7CY8eOiWP58OHD19SW0zUTfm82m8nIyODrr7/m008/Zfv27YSHh1NVVdXruDF0T9JOTk4cP34cKysrJk6cOCCldGxsLN7e3kilUjZs2EBmZiYSiYRbb72VqqoqcnNzCQ0NJSAggJycHFQqFRMnThTVYqH7gTp79iyzZ88WB5vZ3E01XlxcjEwmo76+nhtvvJE33niDYcOG4ePj00u7w9LSkrlz5/LLL78wefJkoqOjsbW17WdZFxQUiCsXnU7HunXr8Pb2ZseOHTz66KPodDrUajUHDhzgkUceYffu3aJR4+npyQMPPEBsbCwGg4GUlBTuvvtuPvjgAyIiIggLC6OsrIzRo0dz9uxZ7rrrLrZs2UJYWBidnZ1kZ2eTk5PD0KFDmTx5Mrt27SI4OJjq6moA8Rj9nDlz8PHx6dd/jY2N7Nixg7CwMOLj43nooYdYsWIFjz76KJmZmVRVVTFr1ixWrFhBSEgIM2bMYNq0aUD3A75lyxZ8fX1pa2vj9OnTQPcev9FoZMuWLbi6unLdddeRk5ODXC4nLS2NZ599luzsbEwmE8ePH+fpp59my5YteHl54evrO+A2REJCArW1taIxcNttt7Fz506GDBmClZUVHf8fe+cdHlWV///XTPokk957T0ggEDoRkSq9qWB31bWsbRV3V1ddd0VZXdcurquiIiqLggrSa0JPgISQTjrpbdLbTDIz9/dHfvd8M8kkgmVVlvfzzAOZufe0e+6953zK+93ZiVKppKqqihtuuGFQYN/x48fZs2cPUVFRLFu2jKSkJEHFP3v2bF599VUiIiIYP3483d3d9PT0kJ2dzbPPPivK0mq1/POf/+Sqq64iKCiI7OxsDAYD06ZNE5lokiSRlpZGXV0dR44cYeXKlahUKlJTU/Hx8eHbb79l1apVvPTSS7zyyiu88cYbzJo1CycnJ86ePYtOp2PMmDFERkYOula9vb3iWsXExPDtt98SFRVFYmIit956KxERERe961MoFEKM7+uvv+aRRx4RKft2dnbMmDGDV155hdWrVwvepbFjxwrGYBnDuRAkSaK6uhpvb286OjrYunUrM2fOpLi4mKCgoGH5SJRKJZGRkUydOlVIfsjtNgdra2tWrlzJCy+8wOLFiwWzdWVlJVVVVdTW1hIXFycsRs3NzURHR2Nrayu04WxtbXFxcaG6uhpXV1fBf6LVaklNTeXuu+8G+ua3bB3R6XSMHDmS06dPi81cUFCQcEfIAcLmXNQajYa0tDQWL15Mfn4+Pj4+YnHY3NwsgqU7OztxdnbGzs5OaBQdOnSIxYsXm4zNULC3txcLeoVCgZ+f36Dg54GQr2tbW5sgLlywYAF79uwRPErDoaamhuTk5EH9DgkJYeLEiUPWuW3bNpqbm/ntb38L9BGSuri40NvbS2dnJz4+PkRFRXHixAkkSaKlpYWIiAiz5eXm5rJt2zYefPDBQS46+f8WFhaCckSWm5AXTgPbVl1djZeXl4hnmjNnDhUVFfj6+g4bcK1QKIiMjGTSpEkmi8HLRWkbLjMLjZ+fn6AGP3HiBEVFRUyYMGHQcRYWFsyZM4cdO3YwYcKEIR9omZmZeHp6olQqcXV1xcHBgTVr1rB3715aW1vx8fGhqKgIBwcHRo4cSW9v7yDFZ6VSia2tLWVlZeI7g8HAe++9R0JCAkuWLOHQoUO0tbXR3t7OzTffzMqVK03KUCgUzJ49m6SkJEJCQoY0tRYUFAiip/r6eo4fP86sWbNQKpX4+vry2WefYTQa8fDwoKysDHt7e/z8/Ljrrrs4deoUZ8+eJTk5GTs7O3p6erC2tqatrY377ruP+fPnExcXh6urK8nJyahUKry8vBg9ejT/+c9/aG9vF+Px5ZdfEhgYyNVXXy20XORxKC0tNTvWW7duxcHBgRkzZqBWq9m9e7ewBLm6utLY2CheSKtWrSIhIUGc6+TkhJeXF2PGjBGCcwCnT58W7M5//etfaW9vR6PRMHfuXCGseODAAbFw0Wq1ohxzRGqyVWrq1KksW7aM7du3U1xczN69e1EqlYSHh1NQUMDhw4eZNGnSoB2bzLR65513kpycLAIuXV1dRUZJe3s7jz76KLa2tpw5cwZ7e3uxs5dha2uLvb09HR0dNDQ0sGvXLqKiokx2dnq9nvXr19Pb24uPjw95eXmi7RMmTBA0783NzVhZWWFtbU1HRwcODg6MGjUKg8EwpFneysqKgIAAYmNjyc3NpaamhmnTphEfH88HH3xg9pyhEB8fz4kTJ0RWzdGjR7G0tOTMmTM4OjoyceJEFAoF27Zt48477xRBsdbW1rS3t6NUKqmpqRGK2uZe2BcuXGDnzp1UVFQQFBQkxD2PHj2KWq0e1hWqUCiYMWOGsEhlZWUNOYfl421tbcVLydramgkTJuDl5SUYw4OCgoiOjiY9PR2NRsOCBQu4+eabyczM5PTp08yfPx8HBwe++OILsTiXr6nMowWwcOFC6urqSE9PJzY2ltjYWObPn8/p06fJysrilltuobOzk9dff13EpxgMBk6fPo1SqSQlJUUEEm/evJknnniCI0eOEBERQUREBOnp6bS0tDB//nxuueUWMjIyOH36NIsXLxZZTW1tbcNmvnV1daHVaqmvr8fBwYGHH36YHTt2cPz4cR5++GGam5tRKpVDxs9IksSxY8dITk6msbERhULB3LlzWbRo0UWlwwcEBLBy5UpuvPFGk89QixkZ48ePx8fHhxMnThAaGso111zDihUrOHXqFGfOnOHWW29l9OjReHh4kJGRgdFoZOrUqWbLkvm9jh07hoWFBefOncPJyYm6ujohinzPPfewa9cuPvvsM9zc3NBoNGbHoqKigu3bt1NWVoa/vz/XXHONCLx2dHQcNohenstJSUkcO3aMrKysQRuDXzsuGwtNf/p7mYZ67dq1ZGZm8sILLwya/C0tLcycOROVSjXkzdif+l2pVAq3QV1dHYsWLSIuLo729nbS0tJoaGjA1dXVbGCx3D4ZRqNR6P5YW1vj5uYmhMesra3Nvkzb29uFxtRQ7e0vXeDp6UlISAiHDh1i/vz52Nra0t7ezjXXXMO8efOE5kh/yYOmpiYcHR2ZN28es2bNoq2tDSsrK0Epnp6ezuLFi7GysjK5CWpra1m+fDmxsbG0t7fz/PPPi92KOdP5wKBJ6LOyyLtCf39/s5kvsnlcJj40N7b19fXk5eUJzSOZbdPa2ppx48aRmJhIVlYWN954I+3t7djY2DB37lxmz55toqdjbndkMBjQaDTodDpsbW1xcnJCpVLx6KOP8uabb7JkyRKuu+46Pv/8c55++mn+/e9/m2hfASZSBHV1dQQGBrJs2TLmzp1rMt4tLS24uroK+Yr+491/rsfFxTF37lyefvpp/v73vwvhQL1eT2NjIwkJCSxevJjGxka2bdsmrkf/MuT+yteyrKwMV1fXYbNXZLS3twvxTR8fn+/0//eHQqEQ7iMbGxvGjRuHlZUVK1euxNLSkqlTp6JUKhk1ahSSJAmpATmOITQ0VMhxrFu3bkiTf2BgIE8++aSQnvDz80OSJMF8/V0vRldXV1auXEllZaWQxfiufg08xs/PjxdffFHUt2LFCtrb24mNjRXPAD8/P7FoUSgU3HrrrSbSFw4ODrz++uuCBVZuV3d3N2PGjBHuuI6ODsaOHSvS3e+++26xc7ewsGDSpEnihW5paclDDz0khHDlMVq5cqVJ+1xcXAgICDBpnxw0Ptx4qFQqnnnmGVHXwoULRZyPlZUVBoOBmTNnDnntFAoFV111FQkJCeIYV1fXi7LOyOdfKhQKBcHBwfj4+NDd3S0s+FOnThVxNPI1uOmmm+jo6CAmJmZINunQ0FA++ugjMbetrKxEqvkf/vAH4T5cu3YtAPfcc4+QKBjYLj8/P5544glxzrJly5AkiUceeQT47rns7OwsPAzyhuBywmWzoOkfH9LW1sbp06d54okneO+990x8hJIkUVtbS0pKCr/73e+GLXPSpElUVFQgSX0kUnIgYUJCAn//+99ZunQpbm5upKSkYGtrK9pQX18vdiYWFhYYDAaTtFArKyuWLVvGvn37hIChr6+vaPtA5tr29nYOHjyIv78/er1+yBtn5MiRnDp1SsRidHR0UF9fj6+vL1ZWVri4uLBmzRpGjhxJXFwcHR0dWFpaCrryqKgo3n77bVxdXXF0dGT8+PFCVqGrq4vDhw8zcuRIWlpaqK6uxtramtTUVEGpv2TJEtzc3Jg5cyabNm3Czs6OhoYGKisrhYk0NDSUt956i/j4eKZNmyb6uWDBAt5++23Ky8upra1l0aJFtLW1ceDAAZRKJY2NjTQ1NQlZif5WKnlXnJqaSk5ODu3t7Xh5eYmy5JgROai3vr6esLAwgoODycvLY+3atbi5uTF79mxRjqurq8gEkSUL9Ho9ixYtYt++fUyfPp2RI0eKAMEnnniCkydPcvToUVasWCHOkRc0cmxNS0uLcE1NmDCBNWvW0NnZib29PaNHjxb9GzNmDO+++y5qtRq1Ws2iRYvEddbr9bS1tdHa2kp6ejoRERHceOONJrtcGxsbIiMjee6550hISCAsLIz4+Hjefvttrr76ajo7O8Xiae/evYKX5Ny5c+j1elQqFTU1NTQ0NIi53J/63dnZmYyMDBYvXszWrVvJz8/n/Pnzg6yL3wXZcgf/Ry0/HHV9/zb0f3gP90I0R4cvb1IuBjK9/HCp3f1hYWFBbGwsaWlpTJo0SegUDZQAkN02MvpLL8jxbAEBASbt6E+1Ly/W+7sZlEqliaXOYDDg7OwsLCrmpAvMvQQvpn3FxcXY2dkN624yJwMx1DUc6vyBx5hbMP7YkJ8pA+UOBgrnWlpaDhqnocqC/5unA/8FLoqf58eayyEhIRd1/K8Nlw2xnhw4Vl5eTnh4uAgslEmR+i8QysvL0Wg0ODk5ERoaOuRkaG9v56uvvuKmm26ivLyc7u5uEXB79uxZJEkiPj5eBLuFhoZSVVUl5AIiIyOFf/qGG24wqUev15Ofny8WO3q9noKCAvz8/Ey4aiRJorGxkYqKCqysrIiIiBhy4svxIjNnzqSrq4vExERcXV1pbm7G19eXhIQE0tLS8PX1Fe4RpVKJWq2mpqaGmJgYQYEfFxeHUqmkpKSEoKAg3NzcSEtLw9XVldbWVgIDA2ltbUWr1RIZGcm5c+cAxE4xMzOT3t5eHBwcCAoKor6+ntOnT7Ny5UpOnjxJW1sb8+fPN+lnRUUFGo0GHx8fvL29aWlpISsri6CgIIxGIzY2NlRWVhIWFmbCfyNJEqWlpbS3twvF6rCwMCoqKvDw8KClpYXIyEjS09OpqqrC0tKSuro6Zs+ejVqt5vz588TExODj40NpaSltbW2MGjVKBEK2tLRQWFhIYGAgbm5uIustNDRU+L6VSiUBAQE0NzfT3t6Ovb09AQEB4prL5GJqtRpra2vq6+sZMWIEJSUlaDQaxo4dS3d3N2VlZYSGhuLm5kZJSQkXLlxg9OjRJqRdPT09nD9/Xuzs6+vrhcWi/8O/s7OT1NRUXFxciI2Npbu7mzNnzuDs7MyaNWv47LPPqK+vp6amBm9vb0HalpOTI6QY/P39qa6uJjIy0uTB3dTUxPnz5xk7dixtbW1UVFTg7u5u0uf/VfQPNnd2dr5smYKbm5uxt7cfNk34Cq7gv4nLZkHzU0C25mg0mmFTO4eCTqfj3LlzjBkz5nuxY14qZGtCdnY2VVVVnDhxQnBsjB49+r/GMDwQOp2OjIwM4uLisLGxEcGOPwYD7MVCkiReeuklrKysBBvq1KlTvzOj4HKDvPh76KGH2Lhx46DMvp+rTTqdDp1Oh6Oj4//Ey1G2vMoigf8Lfb6CK/ipcWVB8x3on6Z3qQ8dc7Ei/w3I6colJSVotVqCg4NRq9U/20Pz5xqHgW3QarUUFRUJN993ZVhcjpAkiZqaGmprawkPD/9FLOgaGxvZsGEDN9xwAwEBAXR3d1NeXo6tra3IGKyqqqKzsxN/f3/s7e3R6XSUl5djaWlJUFDQIDO8TEtQXl5OSEiI+F3epNja2uLs7ExzczN1dXW4uroKiv/y8nJ0Oh2BgYHY2toOijVqa2ujuLgYBwcHwsPDhUWqu7ub2tpagoODBb2CtbW1oMmXqfQ9PT1xdXWlrq6OzZs3s3TpUoKCggaNS0dHB+fOnUOr1TJz5syLtnz19PSQlJREW1sb1113HUajkUOHDtHT08OsWbMuWXdNHrPm5mYCAgJwcHCgvr6epqYmvLy8cHFxwWAwCNbZoTLHtFotp06dwsLCAq1Wi42NDQaDgXHjxglXTldXFykpKYwfPx5HR0dhec/JyeGaa665aGtXU1MT586dw9XVlZEjRyJJfZpkKpWK6OjoS7YiyhbkiooKEhISzD432tvbOXv2LCqVSojKylmULi4uFBUV4e7uTmxsrNl+GAwGMjIycHR0JDw8/JLaNxSMRiO5ubkoFApiYmIu++fd/7Zt+CIg+ye/b3DZzzGBlEollpaWREZGEhcX97Pven+ucRjYBjs7O0aNGiXSZH/uNv0cUCgU+Pr6Mnbs2F/EYgagsrISCwsLAgICMBgMHD58GFdXV8rLyzl79qzI/nFwcGD9+vX09PSwadMmAEpLSzl06NCgTA2dTsexY8d4/vnnTQKVe3p6ePXVV4VrcdOmTbi5ufHtt99SXV3NyZMnOX/+PFZWVmzYsGEQR0d3dzcHDhzAw8ODTz/9VHDSSJLEjh072LFjB5IksXHjRiwsLCgqKuLw4cPU19ezdetW3N3d2bRpE83NzXh5eeHr60tubu6gMZEkic8//1y89GXOo/68M0OlqltZWWFra8unn35KcXExlpaWqNVqoqKiLtkqKkkSqampHD9+nI6ODr788kuqq6vZsWMHbm5ubNy4kba2Nnbv3i3irTZv3my2XQ0NDZw9exYrKyuefvppjEYjBQUFgvdJ/vzrX/+isbFRlNHV1cXbb78tuKcG9ttcXba2tmzcuJGdO3cKksPk5GRBTjqwj+YoAAaWW11dzUcffTRkVpAcj7ZhwwYsLCxQKpVkZmbS2NiIvb09zz77LC0tLWb1teRN8759+0hJSRmyTRcLSZIoKyujt7eX5ORk9u/fP2isLrXMXwMuywWN7Hr5oUymsi+8o6PjJ7vw8o5P5qUYWPdAKQc5WPDHaJM8TnK80cWe093dTVdXF93d3d8pCyFbRuTrIZ8rk9B1dXUJEj9z58rBtD8F87Ek9RHPDRz7HwNGo3EQIZ1cp9ynH2NOyfOhoaFhELGdPIfa2tpoa2sTZH3myuju7qa+vl4QGF4s5L78UHIuedGr1+vZv3+/iHHT6XTs2rWLsLAw/Pz8KC8vJy8vj9OnTxMSEkJMTAy7du0a1GYbGxsSEhJMXL2S1EcsJgfwnzt3DkmS8PDwwM3NjaNHj7J9+3aioqIIDAwkOztbZHDJ6OjoIC0tDS8vL+Lj4zly5AjQ97KTs+RaWlo4e/YswcHBjBgxgl27dnH8+HEcHR1xd3cXwfRyn80RItbW1rJ//34uXLjAlClTqK2t5cSJE5w8eRKdTkdRURFpaWns3LnTbGaZ0Wjklltu4euvvxZ/h4SEmLxM+7+4B35k6PV6PvnkE0aNGoWvry/Lly/nyJEjuLm5iWDg1NRU9u7dS1RUFJGRkRw9etQkY1CGu7s799xzD1FRUVhbWxMeHs5tt92GJEmcOnWKrKwsbGxssLe3Jzs7my+++AKNRoOLi4sILG5qauLo0aMcP34cvV5PZWUl586dEzpYMuzs7Pjd737H4cOHBSGphYXFICZdrVZLeno6R48epbW1ldraWk6ePMnhw4c5fvy4kP44deoUlZWVg/rUH9bW1vz2t78lIyMDjUYj7r2EhATUajV2dna4u7ubXIPe3l6ysrI4c+aMSHZobGwUgfayxfDEiRMizT4xMZHz589z7tw5Dh8+LCxf/ekHqqurWbVqFefOncPZ2Zm2tjZ27NhBRkYGktQnZJmUlERKSsp/hVX+v4XLakEj++I7Ozt5/PHHKSws/MFlpqSk8NRTT/1kbIqSJLFu3TqztNlnz55l1apVgxZmJ06c4Mknn/zBE7GxsZGvvvqKTz75RDAnfxf0ej2bNm3it7/9Lbt37+bf//43GzZsGFLfRpIkdu/ezW9+8xvq6ur4z3/+w+9+9ztaWlro7u7mlVdeGVZaorq6mt/97ndmeRm+LyRJEvTyjz76qAlH0I+Fzs5O3nrrLbMvm+LiYh566CGT34xG47BMt0NBkiQ2b95MYmIi27dvN/m+oaGBl156ia+++oqDBw/yxz/+kZSUlEFlVFZWsnPnTpKTk/nqq68uqf6qqioeeOABWlpaLrntcjubmppwcnIC+hYi06dP55577hEv0traWhP3YEVFhdhpW1lZ0djYOGihZm4XLksQyHFDtbW1wjVia2tLZWWlUMSWJQj6p00DeHh48Pzzz4vA95kzZ9LT00NFRYVwE8gvUDmjRM70k8u1sbGhpqYG6EvFHigdoFAoBCVAYGAg7e3tvPPOO0RHR3PmzBnWr1/Pzp07+fzzz4X1pj9kdu65c+dy9uxZIYEwMNuooKCALVu2DPpkZ2eLY7q6usjNzaWpqYnMzEySk5OF5ILcv4qKCiGXYGFhQWdnp9kFjZ2d3aAsIQsLCzZt2oSfnx9vvPGGIBfs6OigtraWZ599VjzndDodH330Efb29qxdu5YzZ87w3nvv4erqKgg8+4+hHEyflJRETk4OkZGRg1yTp0+fJjc3l9raWr766ivKysp46aWXUKvVvPnmmzQ0NPD+++9jZ2cnXGvDISQkhNGjR7Nt2zZKS0vx8vISGU7mLMK5ubkcPnwYOzs7sXhuaGjA0dGRF198kcrKSt577z1GjBjB8ePH+fLLL0lLSyM5ORmdTsfnn3+OQqEQVkUZKpUKvV4vaD5kWpEXX3yRpqYm1q9fj6OjI//85z/Jy8sbtk+/Jlw2Cxp5Jfvhhx9iaWmJUqmkq6uLpqYmwbrY09Nj9uEHff7L1tZW8WKWJImOjg6cnZ0FR4pM+y9bHIxGIy0tLWK33dPTI6jl5dW5TIGv0+lobm4WNPbd3d2C0v/+++/H19dX1Cnv7IODg83qpISEhAiiNbkN/entOzo6xK5ZlisYuPOWyQdra2t54IEHCAoKEqnAcjnd3d2D5COsrKwIDAzEYDCwfPly7r33Xi5cuMCaNWvMLvqUSiUzZ86ksbGRzs5O5s6dS1FRET09PVhaWhIQEEBMTIxgEO1fhkKhwMvLS5CU9ZdtkFOg5bZ2dnbS3NwsBEFlKYaBCy1Jkjh79ixff/01dnZ2op/950n/v83NEVkCQ7YcyWOt1WrFefb29jz66KNYW1uLAFBZLsHDw4POzk4T6YujR4+yb98+Qfhnzmont6G5uVn0q6Ghgd27dzN79myWLVsmjtPr9TzzzDOEhoZy5513snz5cu68885BD3SDwcA333yDnZ0d06ZNY+7cuRiNRtra2kQKuWz56ezsFNdIHicbG5tBMg+XgoqKChITE5kxYwaAqPPTTz/FwsKC//znPya0+0aj0YTBWZKki3IJ6/V6zp49K8gLm5ubTawjBoMBS0tLEzp7MM+jZGlpSXJyMhMmTBAU/k5OTrS2tg7SJJLbJ3MPyX2QYygmTJhAbm4u58+fN6lXThl2dnamoKCA7u5u3NzcmDFjBkeOHMHDw4Pw8HCWLl06yI3U2NiIu7s7Li4uTJgwgc2bN5slvwsJCWH+/PmDPgPZoW1tbYmJiSEuLo7//Oc/ol/9r8fAcbrYGBVra2tuuOEGiouLaW5upqenR/C03HHHHVy4cEFo4dXX15Ofn49CoeCRRx4hKCiIkJAQ/vSnP5mNC7KysuLOO+/kk08+4eTJk4wbN27QGIwePRofHx/q6+uFNcjT05MxY8Zgb29PXV0dycnJhIeHC/bw4aBUKvnNb37Dli1b2L9/P5MnTx52bvr6+nLq1Cm2bt0qsgijoqIYO3asEGPt7e3F1dWVGTNmkJiYyJIlSzh+/Dg9PT1UV1eTmppKWFiYSdusra2xtrYWbmWZtkGS+tLtS0tLUSqVPP7443h5eV02rqfLKp8wNTWVI0eOsHDhQoxGI4cPH2bjxo0kJCQwe/ZsDh06RHt7OzU1NTz55JMmgYIbN27Ex8eHY8eO8fvf/57CwkIqKyvRarVotVpqa2t54okneOuttygrKxMEW729vezatYuHHnqIzz//HDs7O3p7exk1ahR+fn6UlpYyb9481q1bR3x8PPn5+SxevJhnnnmGUaNG4enpSXp6Ovfccw89PT3k5OTQ0dHB+PHjzcrA9785DAYDmzZtwtfXl/T0dOLj4/Hz8yMnJweAyZMnc+jQIWJiYqioqDB54cmpvxUVFdTW1gpzr6wLEh4ezmuvvUZUVBQzZ840YcHs3wZbW1vuuOMOVqxYweOPP26Wk8LJyYmpU6eyfft2brjhBoxGI/v27WPSpEmEhYWRn59PSUkJqampTJ48mXnz5g16CHR1dbF9+3ZmzJiBVquloqKClJQUpk+fjr+/PwUFBWRkZAiW27179wq207feekvo9UhSn9hgdXU1M2fORK/Xc/DgQYqKipg7dy6jR48mLS2NCxcu4OTkxF133SXa0t3dzaOPPspjjz1GUlISKpWKm2++mU2bNrFw4UJOnTpFbW0t0Kdx89JLL/Hmm2+Sn59PYWEh+fn5WFhYcPfdd9PR0cG3335LUlISTz31FLt378bBwYHRo0dz8OBBQkNDUalUJoymPT09bNiwQZj1582bR29vLzU1NeTn5wtWXeijfE9NTeX5558XgoJy0KvBYBBzv7u7W5ih29raOH78OK+99hrvv/++4DiZO3cuHR0dlJaWimsUExPDgQMHCA4O/kHWs4CAAGbPnk1iYiJ33nmnsJCEhYXxl7/8hf/85z9ERESIhR9AZGQkarVa8PEEBgYiSX2MwEFBQSbkjf0f1N7e3tTV1Ql+nZCQEOFGkCUI5A2Jh4cHNjY2ODs7U1ZWho+Pj1icZmVlodVqmThxIpmZmUJHp6KigqamJqytrbG3t0ev19Pa2ipcT7IVtLW1lcmTJwN9z6yYmBiio6MHjY3cdg8PDyoqKtDpdOj1esLCwoZcMEhSHzW+zJlz3XXX8cADD5jlByotLRV0C/0RExMjtI5UKhXh4eFiU2Fra0tUVJTYULW3tzNhwgQyMjLEwlaWa5DpAIaKWZGtc2+//Tb//Oc/BYGnDIPBQEREhLBwWFtbU1tbK6wejY2NTJ06lZiYGN566y0++ugjE64chULBtGnTePXVV9HpdGb5YrZs2YKdnR2BgYHk5uaauNzkxWhnZye1tbXo9XrhKh9q/GXLkLu7O6WlpYIpHRAL2v7Q6/W89NJLrF+/nm+//XbQGHl6elJWVkZPTw+9vb2EhYURFhaGtbU1J0+e5KabbuKjjz7itddeMxln+R4YGBIgSZIQ0pQlaDo6OnB3dzfbn18bLhsLjUKhwN/fH39/f4KCglAqlcyYMYObbrqJ5ORkDhw4QFFRkdihDTSJRkRECOG6srIy1q9fz/Tp07nqqquwsbEhICCAyMhIsrKyaG5uJjIyki+++AKj0Yi3tzf19fXY2toycuRInnnmGYqLizly5AhTpkzB2tqaiRMnYm9vT3JyMm5ubnR2dnLXXXdx++23Y2FhISj+ZQmFi1Flrqqq4tChQ1xzzTUsWrSId955h9raWnbs2EFYWBiOjo6cPn2a7OzsQVTftra2RERECM6TL7/8kunTp7Ns2TI++eQT1Go13d3d/OEPf2DKlCnDtkNOOx0uFmXp0qXs3r2bo0eP8thjj/HNN99w5swZYmNj2bBhA11dXfj7+5t1fckv8mXLlhEeHs6GDRvQarX4+/tTWlrKp59+SkdHh4ix8PLywsHBgT/84Q+EhIRw+PBhUZbMFxMaGoqnp6eQwbj++utJSUlhy5YtQtm3pqbGxEpjb2/PpEmTOHfuHBEREZw5c4a6ujpGjx7Nzp07qampwcPDg/r6ejw9PYVFZs+ePbi6unL11Vfj4OCAg4MDKpWKFStWMHLkSPLz8wkICCAqKgpbW1v27t2L0WgclOmQm5tLdnY206ZNY+bMmfzrX/8iPDwcd3d3xo0bZ5I5IVsQ5RgSo9HIG2+8weeff27ygJPlL0aPHk1CQgKdnZ1YW1tzxx13sH//foKDgwkMDOTTTz81uUYbNmxg1KhRTJ069Qc9DBUKhfDxQx9LtEKh4MSJE6SlpTF37lyuv/56Lly4wJkzZ5g6dSrBwcEsW7aMlJQUMjIyuO2222htbWXt2rXCRdTb20tKSgo2NjZCyyc+Pl5kGBmNRmJiYvD29iY9PR2FQsHVV1/NzTffTHZ2NqdOnWLhwoXY29vz6aefCrN8R0cH69at47PPPuNPf/oTTU1NREZGCr0jR0dHbGxsWLJkCSkpKWRmZnLrrbcKscr09HTc3NwYM2YM0LcgkIP2B3JP2draUl5eTmRkJEuXLmXHjh0UFhZy77330tHRIQgP+6OoqIi9e/dSUVEB9Flh7rvvvkGM1dC3MFy5cuWgz8iRI8UxlpaW3HbbbRw+fFiQkc6YMUMwh/v6+hIXF8dtt93GmTNnOHXqFDfddBOdnZ28+eabg1yukiRRUFCAt7c358+fF+zo33zzDR4eHuTk5DB16lSOHz/OsWPHeOSRR0yUtZcuXcrTTz/Nl19+iYeHB8eOHUOhULBs2TKz1BgqlYrHHnuMJUuWmLWU+Pr6cuLECTQajWDJdnR0pLq6GrVaTU9PD7/73e94++23SUpKwt3dfZAbciCsrKx46KGHWLlypVhYlJSU4OnpSV5enkkIQVNTEykpKcTHxxMfHy/cnDU1NTg7OxMQEMC8efPYtWsXFy5c4L777sPCwoIbb7yRq6++mnnz5jFz5sxBizVra2sRvyV7Fqqrq3F2dkalUjF79myefPJJvv32Wzw8PC6bBInLykIzELL5WKas9/X1ZdmyZVx77bUmD39J6hO2jI+PF4GI1dXVYiUuP2xWrFjBW2+9xQ033ICtrS2dnZ3MmDGDuXPn0t7ezrFjx4Q8wu23386nn37Ks88+y+rVqzl9+jQLFy4UO2NZP8fS0lJMJpnZ1sPDw0S7ZyjodDoRfObi4oKFhQUjR45kyZIl/OUvf+H5559n1apVvPvuu2RlZfHyyy+bTReUafJ7enpwcHDAzs5OmJJlaYShIEl96ZDh4eG4ubmJl+XA3UJMTIyIO/jb3/7Ge++9J6wg9fX1jBs3jtDQUOFG6n++LOCZmJjIihUrqK+vZ/z48YLI7sknn+S2224jOjqa1tZWcnJyhNvRz89v2PbLLgRLS0v0ej11dXVMnDiRq666ymxsw4IFC3jmmWeYPXs2SqWS/fv3c9ttt7F//34mTJjA3LlzmTt3rkn7Fy5cyPbt2xk9erSgbJczL2T6dxnOzs48/vjjvPnmmyxYsIDbb79dlKXVatFoNMJtNVyWgo+PD2q1msLCQiZOnIidnR2urq64uLhcVOqrra0tsbGxHDx4kGuuucbkGjU3N/PEE08IWZCBNPiX+nCUz5fjPJYvX05HR4cJC+51111HT08PEydORKlUMn36dDo7O7GwsBA7+Ntuu02Y3a2srJg2bRpXX3018H+MrH5+frzwwgviuxtvvJHOzk5GjRqFpaUl9vb2+Pj4CFJIgDvuuEOMmYODA2+88YZou/x8sLe3F6KRSqUSLy+vQe276aab6OrqEim9cp/NUdy7ubnx1ltviSzLO+64Q8TGyFa+/v2SERYWxlNPPSXKVCqV3HDDDUOO+8Vcm4kTJxIbG4tSqRTurRtvvJHu7m5Gjx6NhYUFo0aNIjw8HEmShJXkhRdeMOuiGT9+PF9++aXo2zvvvCPmgLwRlV3Ssrvx3//+N0qlknvuuYfbb79duLnuuusu9Hq9mBfm2r9w4cIh+yrLvFhZWaHX67GwsBAp8m+//TZKpZL4+HjxzrgYVl6FQiEs2nK/IiMj+fzzzwHTayYL/8pzPyYmRpwn9/muu+5Cp9OJPsuWJ/m4m2++eVAbLCwseO6550zaoFQqee+991AqlTz88MNCr+9yIsK8rBY0arWa6upq4bZpaWkRGUETJ07kqaeeQqvV4ujoyNKlS8V5vb297N+/H0dHR5qamqioqGDMmDF8+OGHhIeH09jYSEtLCzExMSI638XFBXt7e55//nmio6MZP3680LUxGAwkJSVx2223sW7dOoqLi0lJSRHMwcXFxSJWRqVS0d7eTktLCykpKRiNRqytrWlsbKSurk5kE8kPdknqk2Ho6urCy8uL8PBwkpOT6enp4frrryc3N5fg4GBuuukmNBoN2dnZPPbYY3zwwQeDzLlNTU00NzejVqu56qqrSExMxM/Pj9mzZ6NQKEQsTn+ZAYPBQG1tLa2trWRkZFBeXs7hw4dZvXo1BoOBV199lYcffnhQSrCtrS3Lly8nJCQEtVrNddddJ/gYJk2axN/+9jfmzJlDYGCgiKeQ4zR6enr47W9/yz//+U98fHwYN24cf/3rX5k5cybBwcEkJCSwevVq5s+fj5+fHyqViurqakpKSmhubmbFihUmbXFxceH48eMUFRUJ4ceWlhY6Ozu59tpreeWVVygsLMTd3Z0FCxaYnOvv74+7u7uwgJWXl+Po6MjVV1/NK6+8QnV1NS4uLkyePFnEg8hihnLcjUqlorOzk/b2dlpbW4XAX2ZmJiEhIdTX1/P4448PEoeMjY3F3t6ec+fOUVpayk033UR7e7v49GdsdXR05LnnnuOjjz7CYDDg5eVFQ0MDI0aMMClTjgGS2yG3a9u2bTzwwAOsXr2a/fv3M3HiRJNrNHfuXNavX09XVxeNjY1UV1fz4YcfMnXqVK666qpLWtTIkh7l5eUEBgZiYWEhgoRlDKT4VyhMZQD0ej2enp7iGIVCYXYhK2sVybCwsDA7V+VFiCRJWFhYCHVieQE8EOa+798+6NtgyXVJUh+/Sm1trVldooHtlNvVvyxzMHfeD8XAsQbEZqf/Mf3dPYBZLhpz18Xcoqd/XwcuIvr/Zk4KYCCGGw9Zo89ce4eq82LQ/9yh5qJ8XP9jh7quA+sfOIcHYqh5+kP69GvAZUWsp9frhRm0sbERFxcXkT4bFRXFhQsXaGhoYPTo0UJgDRAmwba2NhwdHTEajfj6+nLu3DlcXFywtrYmODhYWArc3NywsLCgtbWVzMxM/P398fPzo7CwUGi+yGmcarUad3d30tPT8fHxoaGhAXd3dxobG/Hz88PR0ZGCggKcnZ1xcHAgPz+f4OBg4X+ur68nNDRUPOBlH3lDQ4NwSciK38HBwbS3t1NXVycUkcvLyzEYDLi5uZnQ5+v1eoqLi+nt7RXlFBQUYGNjQ0hICM3NzVRVVREcHGyyoNHr9RQVFdHR0YGtrS0qlUrI1huNRrKzs4cUamttbcXa2ho7OzuT//f09HDu3Dmsra2JjY0V58pxHSUlJfj4+Ihg68DAQLKyslCpVMTExGA0GsnIyBA7xTNnzrBhwwYeeOAB/P39TfoN/xc3EhYWRlVVFe7u7vT29tLZ2Ul4eDj5+flotVpGjx496KaXJEmkFOv1enp6enB2dsZgMJCXlydERLVaLRcuXMDf35+DBw/i6OiITqejqqqKxYsX09zcTGBgoIgZ8fDwIDc3l+joaGpra1EoFAQEBGBnZ2cyT1tbWyktLRVigQ0NDVRXVwtZhoEkcA0NDRQVFQF9MSSBgYEmD7re3l7y8/Oxt7fH1tZWCGbK8668vBy9Xk9QUBBZWVniGikUCjIzM4UWU2BgIGfOnKG3t5dZs2Zd0oJGkvpS+3t6en52zqT/FuTAdgsLC5NrfAVXcAXfH5fVguYKrkBOE//22295++23f/ZdiFarZdWqVYwZMwZPT09cXFyEHMXlBHmh7ebm9rOP+Y8Jcy7UX3L5Ax/n/62F0k89Tt+3DZfSLnNjN9Tr8aecDwOFYAfWOdANbq6M/r/3/1un013WDOlXFjRXcFlBTt/v6OggODj4kmnef4r2yOKNTk5O+Pr6fm/m6Sv470Kv15OWloZarWbEiBE/6jWTU/nT09PR6XQiJkun03H27FkhcOri4kJERARKpVIkJFx11VXCPdLa2kpKSgqhoaEEBweTmppKZ2cnarWakSNHUllZSUlJiYiNkzPECgoKxPdytqacuPB9+pKfn09ra6tJpl1nZydpaWn09PRgb2+Ps7MzISEh2NjYUFZWRl5eHvb29tjb2xMZGSncWiUlJfj7+1+y/l19fT25ublcc8014kWen59Pe3s748ePH/b6NTY2kpqayvTp0zly5AgTJ07EycmJ2tpazp07h4eHB46OjmRlZTFp0iT8/f0veZy+C5IkifT9ESNGoNfrSUxMpLGxkZUrV4q4q8rKSvbv309CQgLR0dGDFnDnz5/nyJEjLF68GF9fX/Lz8zl9+jRTpkwRWVqXqwzC5RMNdAVXQN+OJCgoSMSb/NyQSdJGjhxJQEAAFhYWl+WD5IdguOBmc7/3T/sd6rdLKX+o+hQKBSdPnuTYsWOX1LaLPXbDhg04OjqKwFfoiyfJyspiw4YNeHh48OGHH/Lee++hUCjo6urijjvu4MyZM6Ksbdu28eCDD6JUKjl48CCHDx8mMjKSHTt2CAqBZ555BkdHR86fP88DDzxAV1cXSqVSfB8QECBEeC+mf+bGs6CggG+++cbkOxsbG/bv38/+/ftxd3cnNTWVRx99lOrqaqytrXn22WextLSkqKiI+++/n46ODnp6enj66acF/f+lQo4RKSsro7u7m/Pnz7N161bRdnOQpD4OsX/961/odDqx4dDr9bz77ruEh4dTWVnJl19+SWRkpOBl6p/ePdScNPd9dXW1yOrrj66uLtLS0oiMjBQxMGq1mm+++UawAPf29rJ7926qqqqIiIgw2x8vLy8KCwv57LPPgD5Xs6urK8HBwURGRpKWlvadmVq/VvxPLmgkqU+4bqCswKXCaDTS0NBgluL+vw1JkgQ5nVarFQHA8veNjY0izdNgMIiAVZn639xLQA6q/qGQyeB+yjGSKcqH69OPXV97e/uPMj5yJp45EsX/JuR29Od8+anr02g0nDp1iqNHj9Le3s7u3bvJycmhs7NTUOjn5OSwfft2KioqqKqq4vjx43z77bfU1taSnp7O4cOHqampQZL6SMMSExM5cuSIeHEcO3aMffv2DSK8KysrY+vWrbS3t5Ofn8/58+cpKysjJSWF06dPo1AoBKtwcXExx44do7m5mV27dolA/qNHj3LgwAFBcnb27FkhqdC/rpaWFk6ePMnBgwfRaDRUV1dz8OBBiouLmThxogjsVCqVODs7o1ariYiIYPz48YLdWSbWkwO9Ozo6qKurw9PTUyQ0pKenY2try+9//3scHR1FGnlQUBDXXHONsFg4OTlha2uLn58f/v7+jB8/XvCSyOju7iY1NZWkpCSam5spKSnhxIkT7Nq1i9OnTyNJfTwyMkHnQMhB0LKV6dZbbyU8PJxXXnlFxGwFBARw7bXXUlhYSH19PXl5eYwYMUKwIA+ERqNh69at1NXVsWvXLvLy8mhsbOT06dP09vZiZWWFRqNh1apVJCcn4+joSGdnJzt27ODEiROD7q/e3l5SU1PJy8sTRKS2trbo9XpycnIEM3JHRwc5OTlYWFjg5+fHqVOn2LlzJ62trZSUlJCcnMw333xDc3MzZ86cYceOHbS0tAwas46ODv70pz8JsdD+cyQrKwsvLy+TIN+enh78/f1F9mdZWRl1dXUkJCSYZMjKUCgUNDY2cuutt5KWlkZTU5OIIZWzpPz9/cnIyBj+5vyV4n9yQQOwb98+Xnrppe99viRJ7N27l5SUFNavX/9feQEM15bCwkJeffVVkpKS+POf/0xdXZ1ggX3jjTc4ceIEn376KU8//TSdnZ28/vrr/OlPf2Lfvn289dZbbN68edADf/369VRVVSFJklkK/+9Cb28ver2elpYW/v3vf/8kmkkyenp6+Pe//83vf/979u/fz9q1a9m4ceMgviGZHOuHoqmpiQcffFBQ2H9f6PV6EVD82GOPXZT8gexn/6GLH/m69i8nNzeXhx56aBC/yU+FzZs3Y2Njw759+ygoKCA/P5/Dhw/T09NDbW0tqampwi3zl7/8haysLF599VWsra0pKCjg+PHjWFlZ8cEHH9Dc3MzatWuJjIzkH//4Bw0NDXz88cdYW1uzdetW9uzZY1K3Wq3mk08+ob29nfLycpRKJevXr8ff358PPviAxsZGAOGS+fjjj7Gzs2PTpk1UVVXx4Ycfir8TExNZt24darV6kP6TwWDgtddew83NDQcHB55++mns7e1Rq9WEhYWZFYysrKxk7dq1nDx5klWrVokX1+LFiyksLOT8+fOcPXuWsWPHCqvEokWLcHd35/rrryc1NVVwz2i1WrZs2cJrr73Gww8/LMgvOzs72bp1Kx988AHbtm0zKwtQVFSERqPhq6++Ijc3l/feew8vLy9eeeUVWlpaWLt2Lf7+/kIZezgolUquuuoqMjIyBFHc6dOn2bBhA7NmzcLX15esrCx+97vfUVlZSWFh4aAyZdHJ7u5uIVvQ1tZGS0sLpaWlvPfeeyL+JDg4GKVSSVNTEwEBAbz++usmdBiyhauuro6QkBD0ej29vb28++67VFZW4u3tjb29PREREQQEBODk5ERAQABffvklnZ2dnD17lg8//JBjx46xfv16bGxs+PLLL2lpaSEnJ4d///vf5Obm8v777+Pp6ckrr7wiqA6Cg4MHZSEVFhYKqQLoe37K1BN1dXU0NzcDfaSZcnr3QMiJAFFRUYwfP54DBw5QX1+Pl5cX0DeX5QSWn3sT/lPgslrQyJkDjY2Ngtmyo6NDiPNJ/58ro7W1VRDs9T9XFk6sr68XO3ytVktDQwO9vb3CCtDa2kpzczNbtmwhIiKCBx54AKPRiEajEZT0PT09dHV1CTpv+f+yzkljY+OgRVBvby8ajUZIKciZN01NTYKFUxYiHEg3n5iYiLW1NUuXLuXGG29Ep9Nx8uRJPv30Ux555BEWL17M3Xffzbhx47Czs8PDwwNnZ2eWL1/O/fffT2JioomSrEKh4J577sHHx4fMzEw+//xzdDqd4PSR+ylLOLS1tdHb2yssQXq9nvXr15Obm4uTkxMPPvigMK23tbWJaySLG/aXMuh/TXQ6nUl/tVot3d3daDQak4WJjY2NINRbtmwZ999/P6mpqbzzzjuiTL1ez4YNG8jOzhZSBQNFHWUrhSyloNfr6ezsFHw/cp8BsaNraWkRFr/+11+j0YgFg9yP/kKcBoOBzz//nHPnzuHp6SnSr2Vrlryz7z8ussn6nXfeobu7W6Tfy9ZGnU4n5prRaKS5uZnW1laz87mlpYU333zTxMIYFBQkxlo+Xy5Lvr9aW1sxGAzo9Xqam5vp6uoS1/VSF/YLFy6kqalJyHfccMMNnDx5koyMDCZMmMDRo0exsbEhIiKCRx55BG9vb3x9fZk3bx4TJ04UcSLy/VRXV4fRaGTkyJG4urqSlpYmRANl6nd5fru4uDB79mz27t1LZ2cnoaGhXH/99RQXFwuJCxlyzIrMHdTS0kJWVhaWlpb87ne/Y+TIkYwaNYonnnhikHWwra2Ns2fP4ufnx5gxY6isrBQEhmq12mx6rbu7O93d3aIv8kvOzc2NpUuX8sEHH1BcXGyShm9tbc2bb77Jk08+yVNPPUVaWpr43sfHh6ysLK655hqxALK3t2fJkiXceeedzJkzZ9C9N2bMGEEU2dTUhIuLCz4+PowaNQqFQkFlZSUFBQWC0PS70sXlOeLq6ip4ojw9Pbn++ut54YUXaGhoICcnh4MHD+Lq6srGjRsHlWFvb8/s2bNJSkrCycmJpKQkcnNzGTduHD4+PlhYWGBpaYmVlRVqtRqFQkFgYCAjR44UMh0yZNLL6OhoPDw8BOGli4sL0EfKZ21tjZOTEw4ODtjY2GBra8vRo0extbVl4cKFLFy4EC8vL4KCgpg3bx5paWnY2dkxZ84cli5diqurKz4+PuIayrQcarV60Hh1dHSYLG7lMQ8ICKC0tJSqqiocHBzQ6/VmGdnlPul0Ouzs7LjuuuvYsWMHTU1NJu53lUp12bqcLisemoKCAg4cOICTk5MIwHzttdeYO3cuSUlJvPnmmyQnJ2M0Grlw4YIJY2NbWxurVq0iNjaWwsJCRowYwc0338yRI0doa2ujvr6eefPm8Y9//INRo0YRGBhIVVUVhYWFuLq6snnzZuLj49mzZw8rVqxg586d4oEUEhJCbm4usbGxnDhxgsWLF7N//37uu+8+xo4dC/T5T9evX8+YMWPYu3cvK1asICsri1OnThEWFkZFRQXPP/88e/bswcrKigMHDvDiiy8KsbcpU6bw29/+VixEnJ2defvtt5k8ebI4Rg7ak29q+SFpb2/P7bffzrPPPsvdd9+Nra0tbW1tPP3009xzzz0kJyeTlZXF3LlzOXPmDDY2NuzYsYM//elPvPjii0RFRWFhYYHRaGTKlCns2bOHxx57jO3bt+Pq6oper+edd97hnXfeITk5WSwA6uvrmTNnDn/7299YvHgx+/bt4x//+IcIuGtqamLTpk2MGTOGnTt3cs899/DRRx/h6upKU1MTERERQppgIKGbnZ0dd9xxB4888ggPPvgg9vb26HQ6tm/fjlqtxtXVVZDd7dq1i0cffVTQlO/Zs4etW7fi5eXFjBkz6OzsFOym1157Ldu2bWPkyJFUV1djMBh48cUXWbFihXCfPProo2zfvh0/Pz8OHjzIH/7wB5KSkujt7SUvL4+//e1vWFlZodPp2LFjh3jhtLS0sGPHDg4dOsSaNWuoqKigs7OTffv2ceuttzJu3DgAcnJyOHbsGMuXLycpKYmwsDAOHjzIQw89xL/+9S+sra0xGo2MGzcOo9Eojrnhhhs4ceIEHR0dVFVVsWDBAvbv38/ChQtxcnIS4ycHVG7duhWVSkVFRQUqlYoZM2Zw6NAhPD09RRp3SEgIRUVFLFu2jNWrV/OnP/1JcLZ8F4xGI6+++ir33nuvoFjw8/PDz8+PEydOMHXqVBwcHCgrK2PhwoUUFxebLD4TExPJzc1l+vTpnDhxArVazbXXXitiRVxdXWltbcXKyoqQkBCzAoZLlizhgQce4LnnnkOr1fLqq6/y8ssvY2trK1wektRHStbb24tOp6O7uxtLS0saGxuxt7fH3d2duro6oqOjeeGFF3jhhReYMGGCeDHa2dlhaWlJdXU1vr6+eHt74+DgMGy8j42NDQ8//DAPP/ywsDzKi8gbb7yRxYsXs3TpUqysrMRic8uWLSxevJgFCxawc+dOKisrxY4/ISGBhoYGnnjiCdatWyfqsbS0xNbWFh8fH06ePGnCIfTll1/i4OCAv7//IDVrSZKwtrZGo9Gg0WiExUWOPerfF/nT1NTEV199xd133y1iyYKCgggICECSJI4ePcqDDz6Ij48PU6ZM4Y477qCmpgYfHx+TuTl//nwefPBBHn30Uc6fP091dTWurq5iI9DfwtFfbmDgWCsUCmHpk6kb5E1L//b377O8KJEJEouLi8UiXq63o6OD+Ph4ioqKhqxfblv/RY2jo6PQrZIkiaqqKgICAmhtbWXbtm3MmTOH3NxcQkNDRYDwhQsX8Pb2Frp07e3tggYgNDQUR0dHGhsbTerp6uoSFrXKykp8fX2/k8vn14LLykLj5OQkHvqpqalCdOuWW27BysqK4uJivvrqK2bMmMGUKVNMUmfl4Dz5Bbt37142b95MeXk5zs7OaLVa8XJetWoVN954I+7u7owdO5aMjAwaGxtJSEhg4sSJfPrppzg6OhIeHs7q1asJCgrCwcFBULRPnTpVUOjLOHPmDDU1NSQkJJCQkMC6deuEb/zOO+/kwoULZGRkcOzYMUEG1t+0HRcXx2effUZ2djZ33HEHjY2NtLW1mazMKysrueuuuwaxEMvEWf13lo6OjlhbW9Pb20tAQABBQUHodDr2798P9OnLyAF0c+bM4aGHHmLSpEk4OTmRk5ODXq/Hy8uLuLg4wsLCaGtrE2q5U6dOZcmSJSL2QK/Xs2zZMvz8/EykD3bv3o2VlRVXXXUVISEhbN26FXt7e6Kjo7npppsGEc8N7JO9vb2JGrFKpcLLy4tRo0axd+9eHB0dmTp1Kt7e3nz99dfiXF9fXywtLXnuuefEAszX15eGhga2bNlCYGAgU6ZMwc/PT/ildTodHh4etLS0sHfvXhHvcPvtt3P8+HHy8/PFS0x+aNnZ2eHt7S12j46Ojtx0002EhoaSk5PD559/Tk9PD8HBwUJzSKFQ4Ovri4+PDyEhIYwaNQovLy/KyspoaWnBwcGBUaNG8cwzz7B3716Cg4MZO3Ys3t7eHD58mNLSUsGG7ezsjKenpwhClCFJfXw3mzdvZtq0aSxfvpzPP/+c2tpadu7ciYODA35+fuTn55OYmMiUKVOwsbFh9erVQjPrYqBQKMR1leeN0WjklltuYerUqVhYWLBy5UoKCwt57rnnhMCnSqWiqakJHx8f8vPzKSgooLe3l7KyMo4dO8aWLVt4/fXXycvL49577+Xll1/m448/Ntu2gIAAVqxYwYgRI7C0tMTd3Z0tW7bg6+tLWloanZ2d9Pb2CuvEhg0bCA8Px2g0cuedd7JmzRo+++wzPD09SU1NpaOjg+uvv96EiM7W1pann36aQ4cOcfjwYR544AG0Wi0ODg5UVFSYxIrodDrBTdTe3s6aNWvIy8vj2LFjZGRkkJeXh6urK3/961+ZNGkSBQUFBAQEUFRUhFKpZPPmzXzzzTeEhoYyY8YMioqKCAoKorCwkFtuuYW4uDg2bdpEVlYWAQEBbN++nS+++IK33noLlUplMg8CAwM5ceIEzc3NNDQ0UFVVJUgr3d3dUSqV3H777bz++utkZmbi4uJiYtWSJWaam5v54osv2LJlC3feeSfz5s2jpKSEgIAACgoKMBgMghBVtq6oVCoiIyNJTEwcZPXz9/dn2bJlTJ48mRtvvJGEhAQAqqurcXR0pKOjgwkTJrBr1y7q6+uxsrIShJeNjY0mVrqHHnqIXbt28cUXXxAYGMiFCxdQqVTU1dVRWVmJi4sLFy5coKamBhsbG1paWnjwwQfZsGEDb775plBLNxqNdHd38+CDD7Jx40Zee+017O3taW9vR6VSUVNTg7u7O83NzSQkJLBnzx4R6Cu3JTIykoqKCiRJorS0lH379lFWVkZISAhz586lu7ublJQU1Gq1sAR//fXXIh6mu7ubbdu2kZeXR09PD0qlknvvvddElFNexERFRdHd3c0777xzWVlrLisLTVVVlWBbLSgoABA7AVlkTDZJm8s0keno5R18c3MzUVFRLF++nDlz5tDW1iZo8vuvaGVKegBPT09RviyDINcvR67L7JD941LkMiRJEmXI7ZdX121tbVhZWQm67v7xFqmpqYwdO5b333+f2267jRMnTjBp0iSOHz+OXq/HyspKvKjlQEcZRqOR5ORkEhISBrGx9od8A86ZM4fZs2fT2dkpdjmSJHHixAmWL1+OSqUy2Y3I5cjK4J2dnbi5uYlduTwespVHhuyag75I/aamJsG62T8zxBwkSeLUqVNMmDBhEIOpJPUpR8v98fLyoq6uTrRV7pOlpSUajYb58+czbtw4WltbTSxIA8dHbo/sIlMqlQQHB5OcnIy7uzuLFy9m1qxZQ5rm5WttYWEh1NmvuuoqFi5cSEtLy6C6jEYjp0+f5qqrrsLJyUnsIK2trbGysuK6667j4MGDBAUFMW3aNDZv3oyfnx/Lli1jzpw5Jgvi/jtb6HOHNTY2otPpsLe3R6VS4e3tzYMPPsjrr7/OPffcw913381nn33G3/72N959991BL8TvgkKh4NFHHxVzQJL6WHlHjx4t2uPt7c27774rpDgkSWLGjBkolUrc3Nx49913sbS0ZMWKFRw9epT58+cza9YsNBoN9fX1LFq0iLlz54qxNRdEeccdd4hr8vLLL5v8JkOpVIqYO/l+njBhAkuWLBFl33rrrRgMBiZNmmRyriwfIFtjZRfTm2++KcqWYWNjw4MPPijOkynw+0OpVLJ8+XIAEhISmDJlCgqFgkmTJgl3p42NDUql0oS5WaFQ8Mwzz4h5KjNyA6xcuXIQu+3cuXOZOXMmlpaWQtBUvi7/+te/UCgUhIeHs2LFChGg2r/farWaZ5991qTdMvordyuVSnx8fFizZo04xsvLiw8//HDQefLfsjV6xowZok65r0qlklWrVgnLmtxmWWKhfx9Hjhxpch2USiVxcXHiODnDLDo6mvnz54t5t379evEsCgkJEefHx8eb/BYWFiakUOQxCwkJEeru/REbG8s333yDXq8nODiYJ598Uoypr68vkiSxatUqkzG5+eabhevdzs6O22+/XfyuUCiIj483qUNmel+2bBkqlYo1a9ZclBTKrwWXT0+AjIwMER9RUVEh2Ho7Ojpob2/H0tKSsLAwocra0NBAZ2en2E3p9XrOnz9PTU0Nc+bMITY2lmeeeYbGxkbc3NwYNWqUSVlyLMjkyZPZvn072dnZ5OXlsWLFCiGIKCvudnV1iQyc1tZW2tvb0el04mE+YcIEvvrqK7KyssjJyeHmm28W0gdtbW10dXURGBhIXl4er776Kp6eniaU6QcPHqSqqooRI0bg5uZGdHQ0fn5+FBQUsG7dOmbOnEl5eTk+Pj4i1qC2tpa0tDTy8vIoLy/nD3/4g7iRe3p6RLCds7MzhYWFGI1G6uvr+cc//oGfnx/Tp0+no6MDjUaDp6cnhw8fFkq8xcXFODs7c/LkSSZOnCgyj5YvX86ePXu45ppriIuLw9XVFa1WK2Kd5FgQhULBtddey4svvkhhYSHl5eUsWbKEzz77TGSYyLtn2ZJUW1tLQ0MDaWlpFBYWkp2dzVNPPWXyQHRyciI5OZlp06aJOITa2lqWLFkC9L3Ym5ubhUVp+vTprFmzhkWLFuHn58fcuXNZt24d9vb21NfXU1lZib+/P/v27cPX15empiZuv/12nnzySaG0PHnyZFatWoVer8fV1VW8jKBPuyk5OZnJkyeLuSXHI0VFRfH0008zceJExowZIxaiDg4OaDQa0tPTSUxMxM3NjaamJnJycmhpaRHxJIWFhWJsNRoNU6dOFe1yc3Pjmmuuobu7m7NnzzJhwgSUSqWYa1ZWVsyYMYODBw8SGBjItddeK+KW7r//flpaWjh06BC33norH374IVqtljfffJP77rtvSP/+QPRfnJr7Tf63/wN34IJEjm2RGboPHz7MhQsXGD16NFdffbVY4H1XG2QM93A391v/soej4VcoFIPIFIfq98Dvv0uLbLh2mlvAXSz6j91w8TFD8cWY68vFtmu4+vovnAYuUGRcrAvlYq/LQPQ/Zzgph6HG21z/7OzshNUtJiZm0DHmNlBykK/8+8C2DLS8FhYWik2eub7/2nFZEet1dnaSk5NDcHAwVVVVeHt7C4kB2UyqUqnIycnB3d0dGxsb4T+UJIn777+fZcuWERYWRmhoKBYWFhQXF9Pc3MzIkSOFjLwcAFdeXo6bm5uIf6ioqMDd3R0PDw9KS0uFj7impgatVounp6eInpdf3LKkgmzm719mVVUVXV1duLu7U11dTVBQEF1dXZSUlBAZGSmo7iVJora2VmhJyVIM8H8SB42NjTg6OhIUFISNjQ0lJSViMaBWq0W6oHwDdHd3U1pailqtxtPTk+zsbCIjI9HpdBQUFAgto5KSEhwdHfH19aWwsFD0RQ56a2howNPTk7q6OgICAnBwcKCkpASj0UhwcDA6nY6Kigp8fX2pr68X7gy5Xw0NDdTV1eHl5YWrqyslJSXC8tXS0kJYWBg2Njb09PRQUlIiBNccHBzw8vIySW2UJIn6+nrq6+uJiYmhoaGBhoYGfHx8TMayqqpKLEZsbGzIy8vDYDAQExODpaUl+fn56HQ6HBwcxM4pNzdXaBL5+vpSXl6ORqNh5MiR2NraUlZWRm1tLbGxsYLMTJL60pblOAF5btXV1aFSqXB3dyc7OxsXFxfCwsLEA85gMJCdnY2vr69YJDs4OIgYACsrK3x8fPjiiy8ICAigq6uL6upq7rzzTqqqqmhqaiI2NhaVSkVeXp6IN4O+uKXq6moCAgJQqVQUFRVhbW1NUFCQSE2WFcvr6urQarU4OTnh7u7O+fPnxfX4OdA/VmGgteAKruDXADkZRLbs/1rK/qXgslrQ/BB0d3fz29/+lvvuu49p06ZdVgqkV/C/h+bmZv7whz8wZcoUXFxc8Pf3H1KR+JeI/nEOv8a6zLlcf6y2XGx7zQXB/pj4scft+47Zd5X3U8+hH7vdQ5Xff2P2U9X1a8eVBc3/R3d3N2VlZWLX/Wt58F/BFZiDnPJdV1eHq6urCOL8JaG3t5cjR44AfZkXdnZ2Qiw1Pz+fuXPnDusu+qGQM2+Sk5O59tprf7S6jEYjOTk5tLa2XpLyuE6n48SJEwQFBREWFjbkcbm5uXR3d4sECHNoaGjg1KlTzJkzh4MHD5KQkCCyrn4o5Ll1/Phxrr322ou2yOn1elJTU9FoNIwdO1ZkL0lSn1xJVlYWTk5OeHp64uzsTFpaGlZWVmi1WpHO7OHhQWVlJdOnTxcJD11dXRw+fBhfX18R/yJLFlx77bU/aQaPnIp+9uxZurq6RJzNj4XW1laOHj3KzJkzsbe3R6/Xk5KSIoL5r8AUv6wn3M8IOzs7oqOjL4pP4Qqu4JcOhaKPZyU6OhpPT89f5Jzu6OggKysLHx8fXnnlFZEpkpeXxwcffDCI9M9cmnP/778rSLz/MfK/er2e999/f9i6BqYff1c7ACoqKkwo94cqr//fsnxBdnb2sOMm86EM1XdJ6uNS+uCDD4SOkvxSN3ds/7EoKCgYMm0Z+q6ZrMAulz9UeVqtluLiYvGbUqmkvr6etWvX4uTkJL7v6upi9erVREVFodfr2b17N+fPn0en06HT6Xj99dfx8/MjOzsbvV7Po48+yo4dO0Rdx48f5/7776enp0csHru6ugSj8nBzyFwfzV3nof4G+OSTT3B2djZJxf+uuTPUdRv4N8BHH30kMpEUCgWJiYmCjXqoeoqKigTFQElJidk6Lua++bXhsgoKvoIruIJfDxwcHLjnnnuwsrLCxsYGPz8/IiMjaW5uZsOGDZw6dYq6ujqWLl2KhYUF6enp1NXVMXPmTJHVVVtbS1ZWFkqlkp6eHmbOnMmRI0eIjY2lqamJ5uZm4uLiOHXqFK6urhQVFTF58mTOnDlDbGwsQUFBWFpacvjwYVpaWliyZAnW1tacO3eOmpoaZsyYQWNjI9XV1dTU1DBt2jQqKyvp6elh7NixwnLQ2dkpxCMnT56Mi4uLeLmWlJSQm5sriPVkun5nZ2fCw8NJSUlBqVRiaWmJs7MzAJmZmVy4cIE5c+ZQWFiIo6MjwcHBgsyxp6eHlpYWUlJScHBwoKamhiVLlqBQKIS4pdFoFJmQPT099PT0kJWVRUNDAzExMbi5uZGVlSXEG48dO8Y333zD6tWrUalUglZixowZItNp7dq19PT0cPfdd2NlZWVyjRQKhUl5W7ZsIS8vjz//+c94eXmhVCpxcnISsgfy+BgMBkpKSoRVzs/PT3CrZGdnY2NjQ3h4OH5+frS3t7NgwQI+/fRTlixZgqWlJYWFhYLJVy5TrVYjSRJJSUl0dHSwaNEiurq6yMvLo7Ozk6uuukowREuSxPjx4wXXUHJyMhYWFpw/f55p06YRGhpKbm4uhYWFxMfH4+DgQEZGBl1dXbi6upKYmIiPjw9z586loKCAmpoarKysmDRpEoWFhSIuLTY2VhCPWlhY4O3tTWZmJgsXLkSlUpGRkUFlZSXTpk3DxcWF9PR02tvbTfjSLCwscHJyorKyko8//phx48YRFxdHaWmpmGOenp489thjPPnkk2RlZVFSUsITTzyBTqcjOzsbR0dH4uPjhbyHwWAwyRb7NeOXt227giu4gv8JyGyu/SEHdGu1WtRqNRkZGRw+fJhNmzbR1dVFdna2IIaDPpXkN954A09PTz755BMaGho4ceKE4LSRGa5fe+01LC0tycjIYOfOnQQGBvLWW28JkjS1Wk1+fj5r167liy++oL29nfPnz/Pee+9x5swZ1q9fj7OzM4mJiZSWlpqwzhqNRj766CPc3NzQarUcP35c9Ke8vJxt27bh7OzMX/7yF8rLy/nyyy8JDQ3ljTfeoL29nVdeeYXm5mYT141KpWL9+vXo9Xry8/NNqAdSUlLYuXOnIAK0sbHh4MGDZGRk8PXXX9Pc3ExERIQg4vvggw+orq5m27ZtdHZ24u7uzo4dO0hJSaGyspKGhgbBBeTo6IiTkxPr1q3DxcWFdevWCb4npVKJvb09vr6+ODo6Ci6dzMxMDh8+TEpKClVVVdTV1fHtt9/i4OCAp6fnd7q61Go1Tz75JGvWrOGJJ57AxcUFtVo9KANH5gqbMmUKFhYWHD9+nLy8PCIiIsy6Czs6OvD29mbXrl1kZ2ezZ88eDAYD6enpnDx5kg0bNmBlZSUIUOVr+fHHH1NQUICrqyt//vOfSUpKIi0tDUmS+Mtf/kJFRQX/+Mc/BOO6o6MjERERVFRUsGXLFsaPH8/BgwfZtm0bBw4cENe/qKiIDz/8kODgYN566y2Rfbhjxw62bdsmsiZff/11EhMTOX/+PLGxsYN4eOQEksDAQJ544gnS0tJEHc8++yxarRYLCwvCw8NFUockSXz++ee4ubnx4osvUlRUxBtvvEFtba1Z6Y1fK64saK7gCq7gFweVSkVMTAzh4eHU19dz/Phx7O3tWbRokUixVyj6lMxlmgInJyeRcg6IbA57e3vx0gkKCsLd3Z3w8HAhLqpWqxk3bhy33XYbZ8+e5ciRIzg4ODB//nyuu+46PD09Rer6lClT+OKLLzhy5Ihw+8i6PgEBASxdupQFCxaIfmRmZgoSw9WrV+Pu7s68efPIz8+npaUFW1tbXF1dmTBhAhMmTBDnhYaGEhUVxf79+1EoFLi7uwN9iwovLy/RLycnJ6Kjo/H19aWxsZH9+/cTGRmJu7u74A9ydnbGaDSSmJhIWFgY48aN4/7772fs2LE4OjpSX18vFlTW1tZCVNHGxoYnnniC0NBQsYC0trYWzMcqlYrY2FjCw8NpbGxk7NixqNVqGhoaaGlpwdraGltb22FjkySpT45j9uzZfPPNN7S0tPDss88OK6Fha2vLb37zG7HYGiqWyNnZmREjRuDn50dLSwvTp09Hq9UKOZDx48fz7LPP0tDQIKwT1tbWuLq6Ehsby/z58zEYDOzatQtbW1siIiL44x//iIeHBx4eHkyfPl0sphwcHEhPT8fGxga1Ws3VV1/N8ePH8fDwIDQ0lGuvvRYfHx88PDwIDw/H2dmZwMBAQkJCqK2tFXNu2rRp3Hbbbezbt4+wsDCcnZ1N3HPyvI+NjWXWrFn4+PiwY8cOenp6xBxzcnLCysoKe3t74Z4sKyujsbEROzs7nnvuOQIDA3F1dWXs2LGCx+hywJUFzRVcwRX87BhIEW8uFqGlpYXQ0FDBkipj4HkWFhZ0dXUJniJzcQrm4mm0Wi3x8fFYWlrS1NREWFgYnZ2dJi9Xo9HI2rVrKSsr48yZM0DfwqmtrY0zZ87Q2NgoYhYkScLJyYn8/Hwh4JiXl8fHH39MTEzMICLJ/m1SKBTcdtttvP/++4SHhw9J4jgQMs2A7GLqz/5tY2PDoUOHaG1tJT8/n88++0ywYMuiiTIJZ3V1taBQ6E/qKLv25HL7fz799FNaW1vx8fHBYDAIF465YyWpjxgyKyuLAwcOcPDgQby9vXnkkUeoqakRsS9Go9FkbshutDlz5lBbW4utrS329vbimKHmhXzd7OzscHNzw2Aw4OnpyZtvvsnXX38tiDVlGAwGdDodPj4++Pv7U1RUREhIyCBC1P79CQwMJD8/H4PBgFarJTo6elA7zEGS+mQkamtrCQsLQ6vVolKpyMrKQq/Xm8hK9D/HYDCIxdf58+fx9/c3YRHu7e1FoVDQ09ODnZ0dxcXFeHp64uHhIfTbLqf4GbiyoLmCK7iCnxGSJFFcXIy/vz+FhYX09vbS3NyMl5cX1dXV4mX44IMP8sUXX7B27VoRnyJJEo2NjXh5eVFTU4OzszPNzc1ce+21wmQfGhoqeI5qa2uBvozGuro6vL296e7uZurUqezfv5+ioiIeeOABHn74YbZs2cKbb76Jo6Mjra2tKJVK2tvbqa6uJjMzk2nTpokXlo2NDatWreL999/ns88+w9XVlYaGBhwcHBgxYgQhISE888wznD9/XsTs7N+/Hx8fH5KTk3Fzc6O8vJzu7m4UCgUdHR0YDAaioqK45pprTKQpZPV6CwsLE/4iWRvs4YcfZt++fWzZsoWwsDDKyspQq9VoNBoeeughkpKSeOONN4SK9PHjx+nq6kKj0eDm5oZer6e6upqbb76Z559/nu3bt4tsJIVCwdixY8nNzSUrKwsvLy+qqqrEoiU0NJTjx4+j1Wqpr68XfGBFRUWi7bKUzBdffMH69ev5z3/+Q3h4OKdOneLbb7/lyJEjPPHEE1hZWaHX6ykpKcHX15fi4mLBvyQrRT/77LNce+21FBYWEhQUZKKdJPNfVVdXY21tTWdnJ5GRkezevRuVSkVxcTHp6enU1NSwbNkyE/Z0o9HI8ePH2bt3L/fddx+33347NTU1rF69WhCjurm5UVtbS2NjI05OTlRUVDBp0iSuuuoqdu3aRVdXFzfeeCPd3d1CnLilpQUnJydqa2vx8PCgpqaGrq4urK2tufvuu9m3bx+vvPIKKpWKe++9l5ycHNavX09wcLBQ2gaYOHEiVVVV7N27l5UrV7J48WKTOebj40N8fDx79+4lKiqKiooKlEol06ZN4y9/+QsnTpzA2toaFxcXqqqqLllQ9peMK2nbV3AFV/CzYeAusT/Xhrxokb+XrQgyrXv/na+5Y+XMLvm3gcfI30FfCnl/Esb+dfU/Z+Bv/dur1+sFOeXAevpT3cttGyhO2P+8mpoaSktLaW9vZ968eSb1mOvDwL4Dg8Zp4BhC3yKjv7SB/K/8m9yfgf2UxREHlj+wPPlv+ThzrxuFok/wUW637Co0NzcGXgu5TQPnj7l5AYgxl/+V2yn3UZIkHn/8cVauXMmECRPEWMgLpf6SK0Pxwgw3Dwb+v3+75f4PrFO+jv3rG3jswDkmL1KUSqW4Bv2v98B2Xy4upytZTv/jGGo9+1NOcEmSqKmpEezEV/C/i6Eepuao7QdKDww85lKOHfj/gXEew8kHmJNAGEgjP7Ce/gsX+XxzqfTyyy49PZ20tDQef/zxQWWZ68PF9N3cb3Kb5e/7/26OFr9/P83VPbC8gWMy1HPFnGyEueOHmytDfd//d3nMB/4ro6OjA0dHRxNtN7l9Q9U38O/h5sFwbR54bYaTjTB3jfv3pf+5/dtzMbIMv2ZcsdD8j0OSJM6fP09iYiKjR4+mtbWVyZMn4+rqetETvr6+ngsXLiBJEnZ2dsTGxg55M0qSRF5enpCoGD9+/GV5Y13BFfwQDGV1uIKfFrJVyNxi6Ap++bhioRkGcjCZ/HDpr3U03DldXV309vbi7Ow8yDwol2nOZN3fxCibumUTcUdHh8gs+L43mRxgB5jU3d3dja2tLQkJCRQXF/PZZ5/x6KOPDttHjUaDk5OTSOUMDQ0VferfP3Nj6e/vT3R0NF1dXaJ/vb29dHV1mfB3XMH/Fsy5E35IOf3xc86pofp1Ke6TXwp+ivaZc7/8XPguK8qlYrhr/GPN81/iPPm5cCUo+DuQnZ3No48+yq5du2hra/vO4/V6PUlJScJC0dzczNtvv81f//pX2traSE1NZc2aNeTm5pqcZzAY2L17Nw888AAFBQUA1NXVsWbNGs6dO4dC0ccQKUvFfx80NDTw/PPP8+GHH1JWVia+z8rKIi4uTiykWltbhyxDkvoEHjMyMtDpdHzyySc8++yzQkm6qqqKl19+mdLS0kHnpqen88ADD3D69Gmgz8T6+uuvs3fvXgCSk5OHrfsKLk9IkiS4S+TF//ctp7m5mX379rFjxw6SkpI4deqUye/FxcXiPpYkifz8fHbs2CEyQ2pqati5cydlZWU/OANEkiQqKyvZuXPnoMDLjo4Otm/fTn19/aBzdu7caRIE+ktCQ0MD+/fvZ//+/WJ8enp6OHr0KNu2bSMxMZG0tDR0Oh09PT0kJyezb98+k+dWXV0dX3/9NUVFRfT29nL06FF2797NkSNHaGtr49y5c3z99dccOnSIkydP0tbWJizJX3/9NQcPHiQxMZHdu3df1DP5UpGZmcnZs2cv+frr9Xqys7PFeb29vRw7dozMzExxTE9PD/v376eiouIHtdFoNJKWlkZaWtoPKudyw5UFzTBQKBR4enrS2NjI3LlzB/EBmENmZqagwQZwcXFh3rx5glK8vb2dBx98kJEjRw7ybc+cOZPa2losLS0xGo2Ul5dz2223CX4HR0dHMjMzv/eD1s3NjYqKCmbOnElwcLAIxistLSU8PBy9Xs/Ro0dZtGjRsOUcOnSIMWPG4ODgwLXXXsv58+dRqVTo9Xo0Gg333nsvoaGhJufIGRIqlYrOzk4UCgWVlZXMmDGDefPmYWNjw8iRI0lMTLzsUgmvwDwkqU9LSc6U+eSTTy6aqt7cHKmrq+Oxxx7DysqKqVOnYmVlxRdffCGONRgMvPzyy+zcuVN85+zszAsvvMBrr72GwWDAycmJ1NRUwdg7sL3mPsO1S6fTsWHDBtEv+WNlZcU333wjMq9kKBQKXF1dRdDtcP3VarVi8zNce4Zrt3zscNdI/t1oNPLBBx8QEhKCo6Oj+N7S0pKysjL+85//EBYWxtdffy2IDNvb27nvvvvIysoSZW3cuJFnn30WtVpNUlISx44dY/z48Zw4cYKamhoAXnnlFZEG/fDDD9PT04NKpeLFF1/Ey8uLESNG0NvbS2Nj4/e6TvJ8yMnJGXScvb29eH4PN/4D6ywqKuLRRx+lqakJ6LOCnzt3jlOnTom6lEolu3btorS0dMi5/V31yC6x7Oxsjhw5clHnDVfW5YQrLqdhIEkSOTk5hIeHX5QAmyRJnD59muuuu87EBxsZGYmTkxOffPIJt956K+7u7mbNhCqViunTp7Nnzx5mzJiBk5MTYWFh4tioqCi2bt3K2LFjxTl6vd6EF0GGpaUl1tbWJvW0tLTQ2dmJv7+/+F7WWsnNzaWrq4vx48cTGxs7pBmzq6uLmpoaEWPj5+eHv78/KSkpODo6EhkZiaen55DBe9dddx2ffvopY8eORaPRmChA+/r68uWXXwoujCu4vKHX6/nHP/6Bp6cnN9xwAwqFgsOHD6PRaLjuuusEVXtHRwcJCQnU1NRQUVFBW1sbbm5uTJs2zcRl+/HHH+Pn58fMmTNRKBRMnToVLy8v8XtpaSlRUVFs2bKF5cuXY2dnh1qtZuXKlezbt49vvvmG66+/nuDgYBPXriT1pQGfPn0aJycnysvLmTBhAikpKUyePJnw8HCys7Opr6/H2tqahIQEenp6SE1NpaWlRbhaCwsLyc3NJSAggLFjxwpxxf6Q9Xe0Wi1VVVVUV1fT3NyMt7e3idClJPUxv2ZkZPD000+j1+vJzMzEzs6O8ePHk5KSIoJqjUYjLi4uZGdnM3HiREpKSrCysuLaa6/lwoULaDQarKysGDNmjEn59fX1Qipg8uTJNDc3c/ToUeLi4pg9e7ZJYK2TkxMqlYrAwEDi4+P56quvkCQJV1dXrrnmGtavX88bb7xBU1OTcMerVCq6urpISUnhN7/5DQ888AAGg4HW1lZsbW3x9fVl8uTJvP/++3R3d+Pg4ICNjQ0eHh64u7sTFRVFQECAGDtJkqioqCA3NxdLS0taW1sZMWIEp0+fZvbs2Xh7e5ORkUFxcTGTJk2ipKSEd999l+eff57GxkYMBgMajYbw8HBhKSwsLKSiogJbW1umTJkyrCbauXPnCA0NZe/evdxyyy1CpkAWHX3rrbd46aWXcHBwoLS0lOLiYiGt0H/+TJw4kaNHjwpOmsrKSubMmSPO6e3tJTo6GkdHRwoLC9m0aROBgYGD5sfJkyfR6XQEBARw9uxZsbH28/MDICMjA0tLSyFtcTngioXmO3D69GkmTZoEmMaBnDhxgo0bN3Ls2DEOHTqEVqvFaDTS2NiIo6PjoHJkDZfhgm0VCgULFy5k+/btaLVaE/4JACcnJzQajYn5ur6+nv3797N3716TT0ZGhknZsnk9MDAQOzs7EZ8THh7Os88+y+TJk5k1a5awHA2MH5IhM4DK7VIqlSxdupSPPvoIDw8PwVkxVP/Gjh1LVVUVZ8+eZdy4cYOyP4xGI52dncNdkiu4TGBhYYG9vT3BwcHY29vT3d0tuFlSU1NJSkpCq9VSVFREYmIiJSUlbNiwgfDwcN5++20h2Af/Z4KX56/MJfLNN99QXV0NQFpaGrfeeis2NjacOnVKzG0vLy9eeeUV3nrrLc6ePWu2rV1dXbz99tu4u7tz/PhxTp06hb+/P++++y7nzp1j165dTJ48mT179rB7927Wr1+PtbU14eHhGI1GSkpK2LlzJy4uLjz77LNiFz8QRqORTz75hOrqagoKCvj8888JDw/nzTffpKury+RYWZna0tKSDRs24OHhwZtvvklOTg6vvfaakI/YtGkT5eXlqNVqXnvtNWJiYtiwYQP19fV88MEHeHh4DLrndDod//znP4mIiEClUvHXv/5V0PzHxsYKluT+qKqq4tVXX+Xo0aP8+c9/FnF6N9xwA2lpaZSWlnLq1CkmT54snhFz5sxh5MiRrFixgv379wuZhLa2Nj766CPWrl3L008/LZ6pXV1dbN26lQ8//JBvvvlmUPKBRqPh3XffJTg4mM2bN1NWVoadnR2fffYZhw4dIj8/n97eXp5//nnUajVqtRo/Pz82btxITk4Obm5unD59WvASHTx4kPj4ePbt2zfsc0mj0aBUKrn//vvZvHmzkMWQ4eLigoODAwEBARiNRvR6PR4eHrzzzjukp6eze/duJk+ezO7duzlw4ADnzp0jPT0dKysrPv/8czo7O3n99dcZNWoUH330ERqNBoD29nZiYmJ48803B9XZ09PDt99+i5OTE5s2bRJWf71ez8cff4y7uzvvvvsuOTk5Q/br14YrC5phoNfryc3NJS4uDujTjblw4QLQt5Dw8vIiISEBo9Eo4kDkgFgZBoOB1NRUZs6cSX19vUnsijmTn52dHf7+/owePXrQwkAOuO1/jpOTE3FxcYwZM8bkExQUNKg/p0+fFvTq8o5OoVCIlMn+gcI5OTl88803g8roz+8BfYsUe3t7oqKiRGDwcP2zsrLCzc2NKVOmmLXCyNwVV3D5Q04BtrOzw8LCArVaLXbdLS0tJCQkYDAYaGpqoq2tDScnJ7y8vIiMjBSMtf3Lkon1ZJO8h4cHhw4dws3NDY1Gw7lz5zhw4AAeHh7CvSUjLi6OJ554gj//+c/C7dG/bLVajZOTE6Ghofj4+ODp6UlISIiJQKS9vT1TpkwhKSmJI0eOEBoairu7OzY2NmRkZGA0GvHw8ODFF1800WbqDzs7O+HucnZ2xtvbm6ioKAATkUKFQoG1tTU2NjZUVVWh0WhwcHDg+eefJyIiAldXV8aNG8fEiROFsGVYWBgODg4EBQVhbW1Nd3c3Y8aM4bHHHkOr1Zq0Q6PRUFhYiLu7O/Hx8ZSWlqLT6bCyskKlUpm1VHh6euLg4EBbWxsRERHiWeDj48Ps2bP54IMPaGxsHPRsWrNmDS+//DIvv/wyJ06cABCJBkVFRcTHx4uyVCoVS5cu5e6772bJkiUmz0OFok8Kw9XVlZCQENzd3fHz8yMkJISGhgaOHTuGWq1mzJgxPP7441hbWws9MVdXV0aMGMH06dPx8fEB4NSpUzg7O+Pi4sJf//pX4YYaCEmSOHbsGBUVFWRlZVFXV0dycrLJs8/KygorKytsbW1RKpVEREQQHR1NW1sbp06dwt7eHnt7exISEjh+/LhJ+rtsaWtra6OpqYmRI0cSEhICQFBQEJGRkQCD7oeJEyei0WgoLS3F2tqaxMREbGxs6OjooKamBnt7e/72t78RGBh42bierixohoAkSVRVVQkdlqamJr755huxg8jPzycmJkao5DY3N6NQKFCpVOh0OiSpL9tp7969uLi4EBMTw7Rp09i8ebOgsU5NTTUJDjYYDJw9e5aIiAizac+yGFz/37RaLXV1ddTW1pp8+tOVS1KfXkp6ejoxMTF0dnayZ88eMYk1Gg2NjY3o9XqMRiMajYaqqirUajVtbW20trbS0tJCT08PDg4OwsUlSZJQp42Pjx/U3srKSpOYGKPRSFlZGVZWVmZjFGRyraEe9ldw+UGmkpfnnrwINhqNvPvuuygUfTpG/Sn8wXxmzK233sq+fftEoK3M16FQKDh69Ch33nknt9xyC8899xzFxcWcP38eo9GIVqtFoVCwaNEi5s6dS0pKyqB2movFkP8NDQ0VYpjd3d2MGjUKCwsLCgoKRHCso6Mj2dnZeHl5YWVlRUdHx5AvEXN1mDtWlhawtbWluLhYCEu2tLQMGU/T37prMBgYOXIka9as4eOPPzaxADk6OmI0GqmtrUWn0xEYGIhKpTLLKiuXa2FhwV133YWjoyOvvfYavb294iPrE40aNUpYgCVJYsuWLTQ2NnL11Vczbdo06urqxIL0mmuuYfny5Tz55JO0t7eLeuRnrpubG0eOHBl2/Pr/39HRUchQaLVaId7Z09Njdqw8PT3Zv38/DQ0N5Ofn09HRIQKZ+9fV2dlJTU0NjzzyCHfccQe///3vhbCofIxMcCfX1f8TEhJCbm6umD+xsbFYWlrS2dlJR0cHPT09WFlZMW/ePMrKyrj99tuFtpd8PczND5VKxVVXXcX69etZtWoV69evJzo6GrVazYULF7C3t8fFxWVIa+GvEZeH4+wnQHd3Nzk5OcyZM4e0tDQaGxvx9fXF2dmZ7u5u2tvbcXd3p6enh7Nnz4oYgKioKCorK4mOjubs2bPU1NQwceJEDAYDERERIg4gJCQESZIoKCggNjYW6Atq1Gq1REVF0dnZaaJELEl9GRPybk2Gm5sbCQkJw/ZFkiQyMzOZNGkSNTU1lJeX09jYyOzZszlz5ozQdlmyZAkVFRX09vaSlpbGypUrSU1NpbKyEj8/P4KDgwkJCcHa2hqtVoutrS2lpaV4e3tjb29Pb2+vCUGZQqGgqKiI6dOnY2FhQWdnJ4WFhUyfPp2Ghgbhy5XR0dGBq6urWXP2Ffz3IT9wZQoB+WXS3yUp79T7/3YpSEhIYP/+/Xh4eODv709VVZWg8R85ciSHDh0S95S7uzvOzs5UV1fj7+9Pc3OzcOEqFAqmT5+OUqlkw4YNBAQE0NHRwb333kt7eztZWVmMGTNGUC9MmjSJo0eP0t7eTnl5ObW1tfj4+PDwww8TEBAwyNKo0WgIDAwU6sTt7e00NDQImvnq6mp2796NpaUlN954I3FxcWzcuJERI0YQHR1NTEwMoaGhPPfcc8yfPx9vb29cXFxobGw0GcfW1lacnZ1paGgQCyG5vzJ1vty22NhYkfm4aNEinn/+ea655hqmTp2Kj48PlZWVwmIhZwO5uLhQXV2Nt7c3TU1NQq36pptuMrnv1Go1zzzzDAcPHsTd3Z3HHntMxPKUl5fj5eUl2tzT00N9fT1ubm40NTXx17/+lb///e8kJiZSXV2NJEnEx8ezZs0aYmJiOHr0KJGRkRQXF+Pk5MS2bdtwcXEhODiYOXPmcPr0acLCwigtLeWuu+6ioaGB7du3ExQURFRUFLt27UKlUnH+/HkhVCpfJ/k5XVNTg5ubGw0NDRiNRlxdXbn++ut54403+Mc//sHKlSsJCQnBycmJo0ePolKp0Gg06HQ62trasLKyYsKECZw6dYrVq1ezcuVKwsLC2LBhA4899hhubm6i3jNnzoigb6VSSUBAAJaWlkKYVI4lkhdgshXLysoKT09PJkyYQE1NjZg/ixYtoqysjPfeew9LS0tGjhxJc3MzycnJ6HQ6LCwsWL58OTqdDktLS6qrq/Hz86O5udlkfigUChYvXoy/vz/jxo3j5ptvJigoCKVSyYoVK1izZg1Tp05l5cqVl03q9xVivUuEJEmkpKRw6NAhZs2aRWtrK6GhocLE2tzcTFJSEkuXLv1OFtzGxkYaGxuFyXA4GAwGtm/fzvTp04WV6MfoyyuvvMKdd97J+vXrue2229iwYQP3338/69at48477xQWJHd3d0JDQ7G3tyc9PR1Jkhg7duywN4JOpyM3N9ck2HC4thw/fhxPT89BsUNX8N+HJEnk5ubi6OiIl5cXu3fvprGxkbvvvhuA6upq3nnnHR577DE8PT1JT0/Hz89vyIDw4eqR6eflnXl/yL/JL335mP7HDlx8yDEKlpaW4qUrH9/fOtA/gNLcb/1/73+cuXOBQTT/MsdS/7JlS8bAc/uXZ678oY7tL1XQ//9DtXFgef3bZI4npz9v1Xe1eeC16D925r6X64c+d5oseTCwHnPlybjU6zSwP/1DBAbOI3PXbODCfWCdA8/vD3N1mZs/MmT3vkKh4PTp06Snp7NixQpaW1spKytj+vTpZusb2I6Bx/SfnxfDrfZrwhULzffAhAkTGD9+PPB/ZHvypHB2dmb06NHU19cLX+xQsLW1HZTePBTq6+uJi4sz66r5Ibj66quFCmtXVxfTpk2joKAALy8vurq6CAoKoqSkBDs7O+zt7VEoFIwePZqMjAx6enqGzf5SKBRER0df1A3T1dWFWq028b1fwc+H/Px88vLyuO6668SDUFaRhr4dfH5+Pj09PSgUCkaMGMHnn3/OrbfeekkuQ9ktJP9/IOTf5BeKuYf2wPKGotHv//+Bfw/1m7k6hzp3IK38d9HTm+vDUG0b6tj+dQwl1zBceQPbNLD84cZxqDZ/1/Hm2jHwOXKp5Zkr01y/B/ZnOHkB+d+hJAWGqmuodpqra6j5A6bXMyQkhKSkJD766CNGjRplkuE3VH0Dvx/4/8sls6k/Lr8e/cT4romgUCguepFiLm1zKHh7e4vyfywoFAomT56MJElMnjwZgPDwcCRJYtKkSaKukJAQk4eWhYWF2ZiZgRiojzMcVCqV2UDoK/jvQ5IkvvjiC26++WbxMO8/VxUKBXZ2dibX19bWFknqSxWdNWvWlet4BVfwI8LDw4Mnn3xS/H3l/jKPK0HBPwGG27H8ksqUy+2v5ir/3f87c7ERP3Z7fqr+XcGlo7e3l/z8/EuyBioUCnx9fYdMezaHgYG2PzcGBs3+2OX27+/AIOeLOffHaMfF1vtd5fxSrtl/A7J75vtch4EBwN/3Wg50cV0qfsx59EvGlQXNFVzBFfwo6B8TcTHo7e1l3bp1ZrOK/tswGAxs3bqVbdu2/ehl5+TksHHjRqAv6Hffvn18/vnnHDhwYNjzJEmipKSE1atXD+IYuVQYjUaOHTvGhg0bOHDgwA96MR86dIiDBw/+oPb80jDUIk2SJLKysvjqq6/45JNPLrlcg8EgrnVhYeH3KkNuW2dnJ6+++qpZItWLKeO99977wZILv3RccTldwX8NktSXCt/Q0HBRgcKXI4Z6kfySxsLKyoro6Gg0Gg1eXl4YjUZaW1vp6uoSgYtarZbOzk6hgQR9cV6jR4++pHqampqoqakRAZPyomiooFYZQwWnDnXMUEGb/YNHJamPvn4gBrZnqPL7t6X/76GhoXh4eADw1VdfERUVxYoVK4RG0VD9gT5XQ1ZWlrCsDFfPcOPS1tbG5s2b+fvf/y7Sjs31a6gA6Z6eHg4ePMi8efMYP368SbuHGxtz3w9XT/9Fsbng3u9q5/c9prm5mXPnzjFjxgyTcTMYDHz44Yc89dRT4vuBXGND9VeS+rKetFotZWVlXH311SxYsMBkHIZrq3wN09PTcXFxITAwkJtvvhlra+vvDMQ2F4u1bNky3NzczF6DywVXFjRXcEnofyNezI3Q/6bT6XQUFxdjbW1NXV0d3t7el1zewDJ/LlxKGwY+vHJzc6mqqgL6yMhGjx59yX36PuN2sVAoFNxyyy1kZWUxYsQIDAYDoaGheHl5CS6k7u5uHnvsMaE5JFP2T5069aLbI8ejZWdni7TqBQsWsGPHDuzs7Ojq6iIuLo4PP/yQP/7xj5SVldHV1YWbmxv5+fkUFxdz3333mWisXbhwgS+//JLf/e53fPjhhyxatAilUklFRQWtra3Mnz+fzMxMIVJ59913k5mZSU1NDWfPnjWhSoC+cT569ChKpZLu7m5CQ0PZtGkTnp6enD9/nscff1yQVObl5fGb3/wG6GMlrq6uZvz48XR3d1NQUMDSpUvZv3+/4DNpaWnh97//PWfPnqWhoYHCwkLB4yKPj0x4Cf8XqN3S0kJMTAyjRo0iMTERCwsLIiIiaG9vF+RrS5YsMcl4SklJITc3l/T0dLq7u2ltbWXKlCl8+umn3H///Xz66acEBASQnp7Ob37zG0JDQ0lKSqKzs5Pe3l68vLx444038PHxoaKiAgsLC+bNmydEKisrK7n++uvZt28fPT09VFZWMm7cOObPn28yH9rb2zl69Citra24u7tz1VVXsX//fiwsLGhra2PJkiV89tlneHl5kZubK3iz6urqeOCBB/jiiy9QKBQUFhYyfvx4MaYKhYLq6mpWrlzJwYMHaW5uRqPREBUVxZIlSzh58iQlJSVYWloyZ84c1q1bR2RkJGlpaTz66KNs376dkydPEhYWRmBgoGhzbm4uGRkZJCcnM2vWLHbt2oVCoaCqqorrrruOnTt3otfrqa+v509/+pMI7JXjyVpaWsjMzCQuLo6srCzOnTvHddddx0cffYSrqyv29vZER0dTVFRETU0N9957LxcuXODChQsUFhayePFi/v73vzN79mxmzpzJt99+y4MPPkhNTQ3Z2dm0trYSERGBt7c3mzZtIjg4mMLCQv70pz+ZEAF2dnby1VdfMW/ePGpra0lLS8PS0hJLS0vuu+++78zI/bXgisvpfwjd3d1CH0YmqroUSFKfuKa8izUYDIJXp6qqitraWhOSKlmQrbu7m+rqahobG5k0aRKTJ0/G1taWqqoquru7yc/Pp6ura8j2SJJEd3e34HooLy9Ho9F8L3+wvFuVx2EgQ+rFltHU1CQUmfV6PVVVVWIc6urqTGIVent7TcbBxcWF6dOnM3v2bAIDA6mpqREp7sOZkwfGQBQXF9Pa2vqT+MXDw8OJiIigsrISKysr4uPjmTp1Kmq1GoWiT7R11qxZhIeHo1AoOH/+PMuXL7+kQHe5T35+ftx8881s2bKFrq4umpqaGDFiBFu2bMHb2xuNRiM0jfz8/Fi3bh0uLi4UFhYK5XboWwA4OjqSmpqKnZ0dDQ0NNDQ0kJSURGVlJSNHjqSiooKvv/4ab29vDh06RHp6Op999hlz5swhLi5u0GLMYDCwefNmVCoVQUFB2NnZkZKSwvLlywkPD+f999/nvffew87OjpaWFvbs2cO6deuIj49n2rRptLS0YGlpyZkzZ3BycsLf358pU6YwceJE8vLyqK2tZc+ePcyZM4egoKBB8gYydDodr7/+OldddRVz5szhxRdfJD8/n927dxMVFYW9vT07duygvb2dESNGmJyrUCgYNWoU3t7eJCQkYGtrS1ZWFh4eHpw7dw4LCwtycnLw9/cnPj6eXbt2sX//fnp7e5k7dy5GoxE/Pz98fX2Ji4ujq6uLvLw8UlNTSUtLY/78+RiNRjZs2IBGo6Grq4vrrruOLVu2mMQkSZLEpk2b8PHxYc6cOXR0dLBlyxZ6e3tZuHAhqampYuGhUqlYsGAB3377LTfccAN5eXk0NDRQUlKCWq3mwQcf5KOPPmLbtm3k5+ezYMEC2tra+OKLL2hvb0ej0bBixQq++uorzp49S1JSEj4+PmzduhWtVktqairx8fG4urqSnJxMcHAw4eHhJosZgMjISLy8vLj66qvJyMggOzub+fPn093dzaZNm7hw4QJKpZLf/OY3JouC2tpatm7dypw5c4iJiQEQkhvyHBo3bhzx8fF89tlneHp6cubMGU6dOsXGjRuZMWMGo0aNore3Fz8/PyZMmIC3tzfJycl0d3fz+uuvM3HiRObNm8fLL7+MTqcjJSWFefPmUV9fT2FhockcsLW1JT8/X6i45+Xlccstt3Dw4EHa29sv5Zb9RePKguZ/CDqdjr///e/s3bt3yAfncDAajSQmJuLh4SFMo4cOHeK1115Do9GQlZXF66+/LkzpsrS9JEl8+eWXPP/884JuPi0tjW+//RatVoubmxsHDx40+2I2Go3s27ePzMxMtm3bRkdHB56eniQlJZlQfQ/s53ALg5aWFh555BGys7OHLGM46PV6Dh8+LDhXDAYDX3/9tdBYSU1N5a233kKr1SJJfbpfKpWK3t5e3n//fdauXSvM1keOHGH//v309PSgVqs5fPiw2XHQ6/Wkp6fz/PPP09raCvRZdw4ePGhC4f9jQaFQMHLkyEHkh0MhLi4Ob2/vS7YWKRQKXFxcUKvVGAwGLC0tsbe3p7q6WvBkXH/99Xz55Zd0d3djZWVFW1sbcXFxvPzyyyJLr395A10uixYt4uTJk6xdu1YIM8bHx/P+++9jaWkp9Mns7OwGtc/CwoJbb72VNWvWkJSUhJWVFfb29jg5OTFhwgRqa2upqKhg1KhRPPnkkyxcuJC8vDxUKhVRUVFMmzYNOzu7Qe2SrS6VlZV0dHRgaWnJ0qVLhWtqILq7u7lw4QKWlpZ4eHhgYWGBo6MjI0eOZNWqVVRXV3PTTTfx5ZdfsnHjRhP5kIF1D3RFKJVKrK2tcXJywtHREa1WS0ZGBra2ttjb23PjjTcKegqlUinGqbS0lN7eXjFX5LF1cnJCrVYLRnQZkiSRkZGBvb09Hh4eLFu2jPPnz2NpaYmFhQXR0dGUlZVhbW2NWq3GwcEBlUqFnZ0dlpaWghXZ1dUVf39/4Y6T+X7kNsh9UavVGI1GiouLcXFxYeLEiaxbtw43NzdsbW1xdHRErVaj0+nMuvsGjllJSYmoKzY2ltLSUmxtbfHw8MDPz88kvbusrAytVoulpaW4/ra2tsIqqVKp8PLyEpuU0aNHs3btWry9vamursbKyooZM2YIcWKFok/qQnb3ymPt5uaGlZUVXV1dQmxVpVINeq5ZWFhgY2ODQqEQcg8ODg4olcqf5Pnxc+HKguZ/CPIucu7cuXh5eV3yy6esrAyj0ShM/BYWFjQ1NXH11VcLBV5nZ2cOHDhAb28vqampYvd4/fXXU1lZKRRtbWxsuO+++3B1dcXd3Z329vZBGjoAFRUVpKamMnHiRGxtbTlw4AC2tra4u7uTmZlp9uVfVFREfn6+2T7IDwZra2umTZtmVkj0u5CTk4Ojo6N4sFtbW6PRaJg5cyZxcXEsWLCArq4uTp8+LdiRg4ODcXJyYsWKFRQUFAB9Yn6+vr7ccccdODo64u/vT3l5uYlshQylUkloaCi1tbXioapWq7G0tBSK6T82ZJfHxR53qfNJkiRBjS/TwhcXF7N161bh1mpvb2fq1Knk5OQQFBSEq6srDQ0NHD16lLy8PKGtJkN+8RUXF1NfX09bWxtFRUW8+OKLODg4YDQaOXPmDJmZmZw7dw6FQkF9fT1ZWVnU1tbS3NxsshjQ6/V0dXXx7rvvkp6eLhbLXV1dFBcXM3/+fCwtLdm9ezdFRUWC0+n999/n9OnTgq+pt7cXo9Eo+qrX6+nt7cXf35/09HT27t1LUlKSEB2Ux0c+x87OjvDwcM6dO0dzczOhoaEYDAauvvpqHnroIY4ePUp5eTlvvfUWzc3NNDY2Dhrnnp4eIS1SX19PaWkpra2tdHZ2otfrMRgMon2yCOKZM2c4cuQICoWC7u5u6uvrRX/GjBlDYWEhbW1tVFdXM2vWLNE3+TNwsTlixAjeeecdYY2ZOnUqycnJaLVaWltbmTx5smiLPCfk/8sW2ra2NjQaDb6+vixcuJCcnBwhPzBjxoxBYxwcHMy+ffsoKCggNTWV1tbWQf21trampaVFaIHJ42YwGIQ0x9ixY8nLyxNaSHJd/a3SMvz8/CguLqagoIC6ujoaGxvR6XT09vaa1Ovp6UlBQQGnT58mMzMTrVZLY2MjX331FcePH6eiogIrKytqa2uFDIKNjQ1RUVGcPXuWlpYWgoODhTyIwWCgp6fHrFSIfN3ksZHbIY/r5YArC5pfOdra2sjJyRn0qaysHDRJ6+vrUSgUQ+4Ch4Mk9cknBAUFmQTHZWdnDwrwra+vp66uDisrK8HZ4+fnh5+fH4mJieTk5DBhwgQTQrXAwECzqq9lZWVi5+Ps7Ex2djYAYWFhgxTFZRgMhmF3Hbm5uYSFhQ1LCjjcOGRkZJgIcfb09FBSUmJCImgwGISYqZOTk+hDVFQUlpaWnDhxgtLSUhPFcYVCgZeXl9nAVKVSiVqtHkTMNtw4/NKh1+uJjo7Gzc2Njo4OFixYgJubGzNnzqSuro7rrruOtrY21Go1zz77LNHR0djb2/Piiy9SUFBAW1vbIOJGBwcH7r//fiHdERQUhL29Pbm5uSxatIhZs2bx9NNPc+rUKcF9tHr1ao4dO4anpydz5swxaaMsCHvhwgUeeughbG1t6enp4eTJk/j6+rJkyRKef/55mpqaKC8vJz4+nqeffhobGxsuXLhAdHS0WHg0NjYyZcoUoeszffp01Go1L7zwApmZmTg4OAy6N5ubm1m0aBEdHR2sXr2atrY2cnNz+eMf/4irqytVVVXY2dlx8803Y2try/nz57ntttvw8vIyKaetrY158+ah0WgYNWoUkydPpqysjAcffFC0RZIkIXY7e/Zsli5dyqlTp4iOjsbf35/58+dTVlaGk5MT0dHRBAcHs2rVKk6cOEFgYCDz588nLCwMX19furu7mTt3roklWKFQcNdddxEXF0dGRgYjR45k3rx5zJ8/n2PHjjFv3jwiIiKYMGECdnZ2GAwGZsyYQX19PTNnzhTuq+LiYs6ePcujjz7KpEmTeOCBBzh27BgxMTHMmjULPz8/wsLCxJwKDw8Xx3h7e2Nra8uCBQvo6OgQYqNxcXFEREQM2kw0NTWxYMECGhsbiY2N5eGHH+b48eNEREQwY8YMRowYgYODwyBBXT8/P5566ikOHTpEZGQkU6ZMQavVMmPGDKqqqrj22mtpaWnB1dWVF154gXPnziFJEnFxcfzjH/+grKyM3t5eAgMDufXWWwWr/KJFi+js7ORvf/sbWq2WrKws/vCHP6BQ9GmRaTQapkyZMujZ1tnZKVTOra2tmTJlCg0NDSxatOh7ud1/sZCu4FeN5uZmKS0tTUpNTTX5lJSUSEajURxnNBqlXbt2SatXr5aMRqNkNBqljo4Ok2OGg9FolN577z2puLhYfKfRaKRbbrlF6urqkiRJkvR6vXT//fdLx44dk9LT06WNGzeK8o1Go7Rlyxbp3nvvldrb2weVn5WVJX3yySeDvj9w4ID04osvSkajUTpx4oT09NNPS0ajUdJoNNJLL70kyu/p6ZFSU1OlEydOSJ9++qm0fv166cSJE9Lp06clnU5n0o+XXnpJ2rZtmyRJkmQwGKTOzs5LGodXXnlFqq+vF99duHBBuuuuu6Te3l5JkiRJp9NJN910k5STkyMdOnRI2rNnj8n5//rXv6Q///nPUnd396Dyjx49Km3fvt1s3fL4ajQak7rXrl17UW3/tUCen/Kn/3cDf7+Y84cqy9zfA8sc+FtNTY20dOlSk3vnu8q/lP5eynHf1c+LKWe4Yy+m7Evta/9zLrUMo9EoPfXUU9KOHTsuebwvZc5cTNsvpr/fZ5yH+/uHtOV/BVeynH7l0Ol0NDQ0DLLGmGMzPn36NFOmTAH6VuyZmZlMnjyZ8vJyDAYDQUFBtLW1UVdXh5eXlwjcValU+Pj4CL0dGYWFhQQEBAiW2JycHFQqFePHj6egoGBQmxobG5k5c6bZwFFJkkT5xn56Ka6urmL3093djbu7u9iR93eHKJVK3N3d6e3tpbm5GYPBgKenp4meD/RZBXJycrjxxhuRJInq6mpaW1uJjo6mvLxcWIuamprQaDT4+fmhUCioqanByclJxC/071tWVhYxMTHi+1OnThEREUFkZCT19fWDxqG5uZnZs2ebtRANNQ7D4XLJUJBhrr/mYhsu5fzhfh/u+IFpuB0dHSxbtkyIVV5qed+nvUMd932z275vfT+krOHOudgytFotY8eOFfFoMlP1j9HOH3NMfmiZFzs+l1O69Y+FKy6nXzmcnZ0ZO3bsoE9ISIg4RpIkysrKSE9Px8HBgZycHN566y3c3NxobGzkP//5j0hd3bx5s4jL+Pe//01tbS3nz58H+syo9fX1QF9g7b59+/Dz86OkpIRjx45x8uRJnnrqKWxsbPD09BQKv5IkUVtby8mTJ0Ug3EDU19cTEBBAT08PL7zwgqgnMjISa2trmpubuXDhArNnzxb19zerW1hYEBQURHh4OAEBAQQEBBAeHk5wcLBY3On1ejIyMmhubhaZDh988AF+fn4i+yUrK4sLFy6wc+dOJKmP2Oydd95Bo9GIzAFvb28aGxuRpD4V5oMHD+Lj40NxcTGHDh0iPz+fP/7xj1hYWODr6yuOlf5/hpbMKzEQkiTR0NCAv78/HR0dPPfccyIDwWg0UllZSWtrKxcuXBBcIvKi6wp+eigUCsLDw7nzzjtNXK9X8N+BnZ0dN9xwAzNn/j/2zjs+qir9/++ZySST3ntCGgkhCQQISaihdywgooioqOvadd3Vta3urr3rqqjYQUURpErvgYRASAIhnfTeJ5OZSTLt/v7ge+8vDcSyq+J8Xq+8kszce+45zzn33Oc+5zmfz/QfJatixR8H1gjN7xx2dnY/mBMjOhBPP/00Dg4OmM1mrrrqKiIiIsjNzSUxMZGTJ0/i6+uLj48PQ4YMwcHBga6uLoYNGyadP2rUKA4fPsy4ceOkMpRKJSaTibCwMCZMmCAlh3p5eaFQKOjp6cHOzg6LxcJDDz00gOdDrF9dXR3z589HqVQyd+5caZeSo6MjK1eupLKykilTphAeHo7wfwRoiYmJg7ZXpVINmkMjCIKUsyDuOlixYgWurq4UFRURFxdHQUEBSqWS0NBQwsLCqK+vl3g+RDuMHTuWwsJCKT/i5ptvxs7ODrPZzPDhw5k2bZoUWRkyZAiHDh3qo6T7j3/8Y1CHxmKxoFariYiIQKVSMWPGjAG7FR577DHs7e0lB6m8vJzk5OSL9v9vHaJdhQuQy4nfizvrfgiDORqivX4pIrHedb6QuOOllPFL1umXxI+JEA52vPB/u3f6i/f+GnX7PeDHtOlC98CF7g+xvN730eUKmXApM4QVly0KCgro6urC2dmZwMBATp8+jZOTEwEBAVRVVaFQKHB1dWXIkCEIgsC+ffskTouL3RiCIFBYWIhGoyEpKemCxwqCQHV1NaWlpUydOhU4v4soMjLygkm7arWa1NRUaYdJf4jOzKUuxQiCIG3/dHd3x93dnYKCAhwdHQkODqakpASVSoWnpyf+/v5YLBZ27tzJhAkTcHd3/0E7ZGVlYWtrS1xc3EXtUFxcTHt7O8nJyQiCwNmzZ4mJiRm0jWI0Jysri9mzZ//kh+qvDYvFwvHjxzl16hTu7u7odDrc3d259tprpQl6+/btTJw4EQ8PD4nCf//+/TzwwAN4eHhQWFjIli1buOGGGwgODh7gDIlLRW+99RY33XQTISEhP7veRqORTz75hNjYWCZNmvSTymhoaGDVqlU89NBDP0o3638Bk8nEm2++yS233IKXl9cPHt/W1sZHH33EQw89hMVi4eDBg/T09NDQ0MDtt9/+k8bnhR7AJpOJN954g1tvvRVPT88fXe6vhd6P2v5tMhgMvP7669x9990/uPPSbDazZ88eDhw4wNixYzEYDISGhpKcnExnZycffPABDg4OuLu74+HhwZw5c1AqlRQWFtLZ2UliYuLl69T8Eok4Vvx+8WOTL7u7u4Xq6upLToqrrKyUkmUvdExVVZVgMBik/41G40XLb2ho+FEJzZeCH2sHnU4n1NXVXVIdzGazUFlZKZjN5otev6KiQjCZTNL/JpPpokmMNTU1QldX1+82KdBisQj79+8XbrjhBqGpqUkwm81CUVGR8MYbb0htam9vF+bOnSt88cUXUh80NDQI48ePF+69916hu7tb6OrqEv7zn/8MsJfFYhHa29uFnTt3CiaTSVi5cqWQl5fX57jedhZ/zGaz9PeF+sBisQgvv/yy8O23315yef3R1dUlLFmyRGhoaBhwTu//zWbzgP/FnwvVrf8x/cu5UH/0Pqeurk66dwdrV+/PjEajUFtbK1gs55P333vvPcFgMAhVVVUD6t373N7/9y9vy5YtQnd396CJyGLdBit7MPv3t4P4/8X+Hqwv+n/2Y9qk0+mEbdu2DdoHFotFqK2t7XOti4297OxsYfr06UJnZ6fQ1NQkPPLII8Lzzz8vmEwm4Z577hE+/fRTQa1WC0uWLBE2b94smM1m4dlnnxVWrFghzbWXI6xLTn9wXCwBbTAv3s7OjqCgoEsuW0ycvNgxwcHBff4fLCLRG/23pP4S+LF2cHBwwMHB4ZLKlsvll2SH3pGDH+KAkclkv/vcGYvFwurVq7nyyiulKEBoaCg33HCDFH7PysrilltuYd26dSxevBh7e3sUCgW33XYbO3bs4KOPPuL222+XGIz799XWrVv5/vvviYyMRCaTcejQIT777DOuvfZa4uLiSE1NpbS0lICAAGJiYli3bh329vYkJCRgNpspKyvDy8uLRYsWDdgyr1AoyM3N5ezZs8TFxXHVVVexefNmnJyc0Gg0LF68mD179kjkcOPGjetTN5HU7KuvvqKkpIQ77riDiIgIjh49SnFxMXFxcSQkJHDs2DGKi4tJTEzEycmJ7du34+fnR0NDAw8//HCfSKbwf3lfa9asYejQoRQWFvLoo49SX19Pbm4unZ2dhIaGMmXKlAGRrDNnzkiSAVOnTuXzzz9nxYoVuLu78/3336PVamlpaWHevHls2bKFiIgITp06xW233YZCoWDDhg3ce++97N69m7a2NrZs2cLx48d55plnqKmpkeQqFi5cSGtrK83NzVRWVnL99dezb98+WltbaWtrIzIykpiYGF555RU8PT2ZMGFCH7u1t7fz2WefcdNNN5Gbm0txcbFEuBgbG8vRo0e58847KSsrIyMjA5lMhl6v589//jNffPEFbm5uyOVyxowZQ01NDY2NjSQmJtLa2kpGRgZ/+9vf2LFjByNHjqSlpYWKigrq6+u5+eab+fTTTwkMDCQ/P5/hw4fT1dVFfX09f/3rXykoKODcuXOUlZVx6623sm7dOtzc3CgoKGDevHnodDpWrVpFWFiYxB4sorm5mU8//ZQ//elPHDp0qI8txGhl/7Enl8tRKpU4OTlx3333cfXVV3PjjTeiUChQKBTSuNPr9TQ2NhIWFkZGRgYFBQWMGDHisozS/D7j1FZYYcXvHkajUSIXlMnOMy5v2bKFxx57DLVajcFgoKioiJSUFACOHz8uhe1dXV155ZVXWL9+Pfv377/gNcRcKDH3avTo0cyYMYMtW7Zw4MAB8vPz8fPz49tvv0Umk5GRkcH1119Pe3s7p0+fxs/Pjw0bNqDT6QYtPzg4mBUrVrBp0yb0ej1arZbo6Gi+++472tra2LBhA35+fvj5+V2wjvPnz+eKK67glVdeYcOGDdTU1ODl5cU333zDV199RUtLC+7u7nz99dfY2dlx+vRpSQ6gsbFxQHnOzs6cOnWK+fPn09HRQX5+Pq+//jrJycksWLCA1157rQ/xnohdu3bR3NzMyJEjcXJyIj8/H61WS3Z2Nnl5ecycOZOcnBxCQ0MpKioiIiKC+Ph49u7di4uLCydOnMDGxoahQ4cSExPDrFmzKC8vR6/X88knnzBlyhQmTJhAZ2cnNTU1REVFUVxcTHFxMTqdDo1Gww033MDmzZvx9vbGz8+PhISEAfUUNzZ0dXUhk8morKxk2bJlHDt2jJEjRxIeHs6RI0eQyWRUVVVx991309jYyMmTJ8nJyZF4dt59912mT5/O1KlTefbZZ4mKiiI7OxutVovFYsHGxoZvvvkGb29vTp48SVVVFdXV1bi5ubFw4UJ27tzJ0qVLKS4uprCwkE8//RQPDw+KiorIy8ujvr4eR0dH5s2bx+bNmwkJCSEwMJDo6OhB+ywnJweDwTDAFv05bgaDh4eHtHlCEATS09N5//33GT16NAsXLiQ9PZ2RI0cyffp01q5de9kQ6fWH1aGxwgorfhUolUqCg4MpKytDEARsbGwIDg5Gq9Xi5uZGfn4+3d3dHD9+nLi4OD7//PM+yd5hYWE888wzPPvss1RXVw8ov3+CsZgPJtLdFxUVERISwvTp03n77bdxcnLC0dERHx8f6uvrCQ4OZtq0aaxatQqVSjVo+W5ubjg6OkoyDaLwqsViwdHRkUWLFvHII49w+vTpQR8iSqUSd3d3Ro0ahcFgkPLHFi5cyEsvvURxcTHDhg1j0aJFPPPMMyiVSik6aGtrO+BhJ76929vb4+joiEqlorOzk/LycmxsbHB3d0elUknyGb2xdOlSdu3axerVqzGbzdJOouDgYBoaGigvL2fGjBnY29v3kSfo7u7G1tZ2QAKw+LdOp6OiogKlUsnYsWOJjY3F29ubsrIyZDIZJpNJouMXbSn0yp/pH0lQKBSSCKRoD3t7e+nH0dERnU6Hra2tJG8QHx9PZ2cn9vb2eHl5YWdnR11dHXK5HH9/f4xGI46OjowZM0bSEGtra0Mmk5GQkMC7775LZGSkFBERr6lSqbCxsaGhoQG9Xs/o0aN5/fXXSUhIQKlU4uLigpOTk5TcL7anf5tsbGxQKpV9pAn62+JCEP4vn87GxgZ/f39kMhlJSUnceeedPPjgg8hk57XWiouLcXNz4+jRo5I4riAIl5VzY3VorLDCil8FcrmcBx98kB07dnD27FlJmFKlUmE0Gtm7dy8333wzV199NQ888AD5+fmcOXMGnU4n8ftMmjSJO+64g7q6ukGvYWdnR3t7O83NzRL1u8FgwGAwEB0dzbp16ygsLCQ9PR2dTifRxkdHR7N+/XoKCgokav7eEP6PSl48XmSL3rRpE3Z2duj1eqqrq3F3d+fll1/m6NGjF9yBotfrqaysZMaMGYwYMYK1a9dSXFzM8ePHCQ8PZ82aNRQXF5ORkSFd72IU970p7Q0GA0qlkhEjRnDixAna2toICQnB399/QF1KSkp45ZVXsFgsNDU1SVT9crmcsLAwenp6JMbe3uX3lq4Qf/f+3t7eHrlczqeffsqRI0coKiri3XffRaVSSVT/4rli/8hkMimC13/HosVi6dOXvWn8e//AeYZknU5Ha2sriYmJUp+JCf55eXk0NjYSHx+Pq6srixcvZufOnQwfPhx/f39yc3NJT08nMzOThoYGqV29ZRmMRiOurq7U1tZy6NAhsrOzqaqq6mMXo9GIjY0NnZ2dg3JTWSwWyd7ieWL7LL0EPsU+7urqoqenh/b2ds6dO8c777zDXXfdhYeHh/SduGR98OBBJkyYwKJFi1ixYgWjRo1i3bp1dHd3s3nz5suKKdiaQ2OFFVb8KpDJZCQmJvLKK69w/PhxCgsLsbOz49///jdqtZrg4GBMJpO02+muu+6ira0No9GIp6cnGo0GNzc3li9fTnx8/KA5AbGxsSQmJlJfX8+cOXMQBAF7e3smT57M5MmT0ev10vKMyWTi6quvpq2tjSlTpqDRaMjMzGTevHk4OTn1KddkMhEbG4uDgwM6nY6rr74ab29vFixYQHt7O8uWLcNkMtHV1YVareaee+4ZUD+lUsmDDz7ImTNn8PT05JZbbpGIFXNycli4cCEpKSl8++235ObmsnDhQtrb21mwYIFEyT+Yk9TT08PixYslnTVnZ2eeeOIJjh07Rn5+Pn/9618HFeIU6Quuv/56nJ2dmTt3LhaLhfb2dnQ6HceOHWPfvn0sWbKEuXPnIggCvr6+ODo60t7ezuLFi2lvbyckJASDwUBHRwfXXHMNRqOR559/nk2bNtHT08OwYcO48847qa6uZsmSJej1eoYMGYLZbEav17No0SJsbGy46aabqKurIzQ0tE89Ozs7mTdvHgaDARcXF5KTk2lra+Pqq69Gq9VK48ZkMtHZ2cmxY8e46qqr8PHxkSghlEolL7zwAidPnqSrq4uHHnoIhULB0KFDeeqppyRahWeffZajR48yfvx4fH19mThxomS7mTNn0tzczNy5c3FxceHFF19k3759uLu7ExoaSlJSEq6uriiVSqZPn463t7ckP9E/D7Cjo4OrrroKnU5HSEgIRqNRskVXV1efCKHo/Nx4442cOnUKGxsb7rrrLkJDQ1Gr1aSkpODs7ExXVxc2NjaYzWZpF53JZGLBggWo1Wq0Wq0koHu5wLpt2worrPhV0X8K+ikilxc692Lf/dxzf6gu/XGp1+//2U+pxw/V7cfU5cUXXyQ0NJS4uDhaWloYNmwYAQEBP+val9LnvZedfiwEQeDQoUNs3ryZ119/fVB+l1/a9j9nrP0a+K3V55eA1aGxwgorrLBiUIj5GUeOHEGlUpGQkPCDHFS/BVj+T1W9pqaG6dOnD0pkacXlB6tDY4UVVlhhxUXxc6IlVljxv8Jlk0Mj3nCDhTLFz34pNlUxM/y/RVku9KINH6zOws+kJf+jQ7SHuDXzUm3YP0T7W7blL9XnFypHTFT8OffUhXZYyGSyX2y8Xk5j32KxSFIL/dvU+ztxbhLt+1PES8Vzxf79KcuA/805svd1BuvfS7n+72Vs9L5PxP4Qk7/FnWU/p9zf+lz2Y3BZ7XL68ssv+eCDD3j22Wd56aWXWLVqFV999RWvvPIKZ8+e/cWuU11dzWOPPTbo1sefC0EQSEtL45tvvuHrr78esINBEASys7N5+eWXB2S/X6i8mpoa/vGPf9DV1fWL1/d/DXES+jno7Oxk27ZtvPzyyzQ0NFzSOSaTic2bN/Pwww/z7bffsmbNGlJTU6XdFL8ViGOkuLiYjz/++GeXtX//ft5///0B3x0+fJgdO3b8rITCrq4uXn75ZT777DMefvhhVq9ezUsvvcS+fft+sfH6U+zwW93KmpaWxpYtWygqKuLf//53n7F34sQJNm7cSGtrK0899RSVlZXk5uby5ZdfSu25lDZZLBbMZjPr16/nu+++u+ixg5Up3p/t7e08+eSTNDc3/7TGXgIEQeDo0aP85z//GTBP1tbW8vjjjw/KtyMeU1hYyLPPPntJPC+/FET7/Jjx1dXVxYsvvkhRUZH0mUaj4ZVXXvlZ809dXR1PPvmkJCJ8OeCycmhcXFy46aabMBqN2Nvbc9NNN+Hu7k5raytqtbrPFkcxCnKhG1Lc/y/+tlgs0vE+Pj6UlZVJW/h6v/GLN7l4bP/y+kM8Vvy+u7ubNWvWMH/+fObOndvnuPT0dEpLS3F3dycvLw+z2SydJ9Z1sOu4ublRWFjY5zq9j+3fhv51/6G29P5cdDbEdvW2Se9zL/Zd77/717WgoIDs7OwL2rn3z2C2EASB1NRUWltbueuuu/D09Bxgx8HOtbGxwdfXl8LCQq666irmz5/Pxo0befvttwfdVnmxMnvbdLDxcKm2GGysbtu2Da1WS2hoKIsXLx50jPXvr9627N8nzs7OfSZSEYmJiUyePBngova+GAwGA2PGjOHaa68lJyeH8ePHS2zA/cfrT7VpbztcbEyItrD8n06XSFB2oWv0Hm/963Shvu1/f/Xum95OR+9+6V3+6NGjmT59Op6enuTn5/exz8iRI5k1axYeHh7U19ej1+uJiopi4cKFCILA7t27aWlpGdCe3mNXHD9dXV0oFAoqKiouek8fPnyY6urqPm1pampi3759ODk5UVFRQU9PT5/2Xsg+g9nzYueJx7u6ulJQUNCnfgA+Pj6Ul5djMBgG2FSEh4fHD86jvc/7obF2ofuqd3k6nY4tW7YMcAYHu//E8+zt7WlqaqKjo0O6rpOTE7feeitKpfKS69x77FssFtzc3CguLv6fOnT/bVw2S05wnnFTXKaRy+U4Ozsze/Zsjhw5wvHjx9m5cydTpkwhJSWFw4cPU1FRgY+PD4sXL+6T5b5hwwbq6uq4++67+eabb0hKSuLEiRPY2tri7+9PcnIycrkco9HI22+/TUpKCp2dnZw+fZq77rqLo0ePUl1djU6nY/ny5Rw4cAA7OztCQ0MZPny4VF/L/4m4dXV1UV5ezty5c2lra6OgoICMjAymT58u1aurq4tXX32V2bNnM3XqVDo6Ovj0008pLy/niSeeoKmpibNnz3LmzBluvPFGQkNDpVCiWNeNGzdy+vRpHnzwQVJTU+ns7OT2229n48aNjBo1iq+//prAwECys7NZtmwZMTExpKenU1BQwKRJk7BYLBw+fJju7m5uvfVWwsLCsFgs7Nmzh8zMTIlT49FHH2XPnj04OztTVVXFNddcw/vvv4+bmxtms5mhQ4cik8k4d+4cK1asYP369chkMpqamnB1dcXDw4OcnBweffRRGhoaKCws5MyZM9xyyy288847+Pn5ERwczJkzZ6iursZisTB79mw+/PBDPD09CQgIwM7ODmdnZ1QqFePGjZPsqNPpSE1Npbu7m8mTJ1NUVERzczO1tbWMHz8euVzOxo0bsbGx4YorrmDUqFEAkh1FWnFvb2/uu+8+rrvuOm688UZ8fHykfq2oqODEiRN0dnYyefJkGhoaOHbsGHq9nsWLF1NbW0thYSETJkxgwoQJUt2MRiOfffYZcrn8B22xcuVKDh48KNHRDx8+nLFjx/LKK69IhF/Z2dnce++9HDx4kO7ubsrKypg7dy7l5eVkZ2dLZGj33HMP2dnZVFRUUFJSwj333ENFRQWlpaWUl5cPcNjMZjN79+7FaDQybtw43n//fWJjY8nKyuLxxx+/ZMFAFxcXZsyYgdFolAjhIiMjJaIzcbw+9NBDkgxCbm4uV111FZWVlRQUFNDW1sYtt9xCUVERRUVFjBs3jokTJ0pLLqdOnSIzM5M777yTXbt24eLigq2trUSnLwjnxUM7OztpaWkhKSmJl156iYcffhiFQsGZM2fQaDRcf/315OXloVQq6erq4oorrmDVqlUEBQVx+vRpli1bRlxcHFu2bJE4aa6//nqioqKA84Kqb731FgEBAWRnZ7No0SImTpzI9u3bcXBwoKqqihUrVnDixAns7Ozo6OggOTmZtLQ0zGYziYmJZGdno9FomDt3LlqtllWrVlFaWsrDDz9MUVERTU1N3HTTTdKyU3p6OiUlJSxevJiXXnqJ++67D5VKxcmTJ/n73//O/v37GTp0KMOGDQOgsrKSV155BScnJ2QyGWVlZbz66qs4Ojpy5513SnUtLy9nyZIlvPzyy1x33XVcd911qFQqBEFg/fr1ZGVlSer0u3btIj8/n5UrVxIWFkZqairnzp1j6NChzJs3T+qnzMxMduzYQVhYGPn5+UydOpUjR45w1VVXMXz4cHbs2IGDgwMNDQ2sWLGCiooK8vPzaWhokLbHHzlyRJrTr7zySml55vjx4xiNRlpaWrjqqqukZRq5XI5Wq+WLL76goKCAJ554gsLCQtra2qQ2fvTRR4wbN46kpCS+++47rrzySo4fP865c+cYNmwYc+bMkdpw9OhRDh8+jLu7O1qtloceeoiMjAza29upqqoiJSWFyspKPvzwQ6KjoyXGYIvFwpYtW3B0dKSiooKVK1dy9uxZ6uvrycvL44YbbkAmk5GWlsamTZuYPXs2YWFhrFmzhr///e+UlZVRXFzM2bNnufHGG/noo49YuHAhQ4cOZdeuXUydOpXs7GzOnj3LggULGDJkCLt378bR0ZG2trZLfLr+PnDZRGhEDaD+mhc2NjYIgkBMTAxLly5l8+bN7N27l8LCQnx9fdmyZUuf0LbIjZGRkYFer8fR0RFbW1vc3Nzw9vZmy5Yt0rFKpVIiSvLy8iIvL4+8vDz27duHr68vhw8f5uzZs2zfvp2IiIgBmfZ1dXV8++23zJkzhzFjxvDCCy8wbNgwfH19mTx5ssTUCWBvb09QUBCJiYk4ODhgZ2fHDTfcgFqtpri4mA8++AClUonZbGbfvn0D7COXy5k7dy5DhgwhIyODpKQkjh49SldXF4IgSBPJhAkTuOWWW3jrrbf45JNP6OzsxN3dnU2bNmFjY0NRURH3338/rq6utLa2otfrUSgUNDU18cADD9DY2MiRI0fQ6XRER0ezZ88ezGYzhYWFxMXFSbTyw4cPJyMjg5aWFnQ6HRaLheXLl7N7924WLFiAxWLh9OnTknKsTqcjLS2NkJAQRo0aRUVFBUePHsXPz4+9e/diMpnIyspi/vz5JCcn89133+Hj4zNA78jR0ZHhw4cTExODt7c3q1evZubMmUydOpXnnnsOGxsbTp8+zZ/+9CfpYXQheHl5IZPJBoRsW1paCA4OlhxWpVJJcXExDzzwAAcOHKCjowN3d3e+++67Ps6CjY3NJdvi2LFjdHV1YTQaWb58OVu3bsXT0xNfX1+SkpJwcXHhzJkz1NbW8u233zJ79mxGjx7Niy++iFwup6amhptuuonDhw9TU1PDmjVrcHd3p7S0lJMnT7J69WqmT5/O+PHjB6yvizkHxcXFuLq6kpuby8yZM7GxsflRS7uigzjYZ73H67Fjx/jwww8RBAFbW1t27NiB2WymoaGBv/zlL+zYsQO1Wo27uzsbN27sY1MXFxdOnz6NXq/nu+++w8vLa8CY2Lt3Lw0NDYwaNQovLy/8/PxITk7GxsaG6upqHnjgATZu3EhAQABXXnmlNH9UVVUxZMgQpk6dyo4dOygsLCQjI4N58+aRl5fXZ2uzo6MjJSUlJCQkcMcdd/D222+zefNm2tvbmT9/Pm1tbXz77bds2rQJQRAYNmwYpaWlHD58mBEjRuDo6IhCoZCiZTY2Ntx8882MHj2aDz74ABsbGylSAefnMXt7e/Ly8nBzc8PPz096MItOmkaj6aMf5uPjI40fmUyGt7c3t99+OwcPHqSzs5POzk6GDx/O/v37MZlMBAQEkJycLGlJyWQywsPDiYyMlPTZJk6cyIQJE9i9ezdbt26lsrISHx8fvvvuO4k9F85T/xcWFrJkyRIaGhrw9PRk1qxZfP/996xfvx6FQsGCBQsoLCxkx44drFq1iqlTpzJx4kQUCgX79u0bdE4XBIEdO3bQ3t7OyJEjB4xBGxsblixZglwuJy8vj+bmZsn2FRUVxMXFkZaWRkdHB+Hh4ezYsYOamhqpDb2XfOzt7SktLWXFihVkZmZSWlrKp59+ysyZM5k8eTLPPfccAQEBBAcH95E/MJvNaLVahg8fzr59+6iurmb9+vXMnDmTcePG0dPTA8CIESNYvHgxmzdvxtXVldOnT6PValm9ejVOTk60tbVx5swZhg4dyokTJ2hvbycqKorPP/9cYkLesmULX3zxBb6+vsyYMQNXV9eL3qO/N1xWEZoLQSaT4erqiqOjI0ajkZKSEiIiIpg1axYzZsyQaLTFY4cMGUJAQACbNm0iISEBR0dHWltbUSqVGI3GQcPJ4qQvan1MnDhRehikpKTw0EMP8eijj+Lr6ysdKwqQyWQyQkNDaWtr68OKOViilvhZb6pvrVZLTU0NY8aMYfLkyYPWT6TsFqnKhw4diq+vL9988w3x8fEolUopqiFq6xQVFTFnzhwiIiIwmUyUlZXh4uKCp6cnW7ZsITs7m4SEBIlO3tHRkbi4ODQaDTY2NhJ7q0wmQ6VS4eHhgYeHB46OjlRWVqJQKDCbzdjY2Eg07SqVCjs7O2xtbWlra6O5uZkxY8Ywfvx4BEGQ8jkqKirw9vZm8uTJUpTD3t4eT09PXF1dufbaa3n00UdZuXLlAPFLERqNhsbGRmQyGf7+/vT09EjhXC8vr4uKZAqCQF1dHc7Oznh7e/f5TtRlgfNRF5GC3dPTU2KEjYyMlEjjetdNqVResi2+/fZbnJycsLe3H8Cmamtri0wmo6Wlpc8Ya29vRyaTSdeQy+W0tLRgMBhITEwkKSmJpqYmampqpGv3T/yVyWTY2dlJb7l2dnYSFfwvlVPUe7zq9XrKy8tZuXIlM2bMwGw2k56ejqurK56enpSXlzN58mSGDRvGNddc04d6X7SDo6MjS5cu5fHHH+emm25iyJAh0nHXX389zz//PBkZGTz99NN9bOjq6oqXlxeVlZXSC1JISAiNjY3Y2Njg7OwsEZ0FBASg0+nIy8tjxowZfcRL5XK5NA58fHywtbWltLQUT09PZDIZw4YNo6ioiJUrV/Lyyy+TlJTEHXfcQWBgIA8//DAvvviiZHM4f/87OzszduxYcnJypHb2HwO9+wzA09OTSZMm8fnnnzNixIg+wpa9j4PzzqC9vb0UgbC1taW2tlZKSu5/zmASBaL0QE9PDyUlJUyaNIlx48Yxd+5cyZkVo3PieBfnNZFor7S0lNGjRyOXy4mKiqKoqIjKykpUKpU0PvvP6WLbZTIZK1as4Pnnn+f48eP885//7ONEq1Qq7O3tcXBwoKenBxcXF8rKypDL5ZhMJqZNm8b69es5fvw4M2fOZP/+/UybNo3ExETmzZvXJ2m6tyCkQqGgra2NxsZG5HI5QUFBaLXaPvd8799KpZK6ujpkMhnNzc00NDSgUCiYNGkSZrMZmUyGi4uL9AxTKpXI5XKJBXns2LGS+Kler+fBBx8kICCAWbNm8cEHH3DdddcxZ84cTCYTf/vb3xg/fjwKheIHhYB/b7hsIjQixJtAfNMVabJ7enqkn+HDh/Pll19y+vRpUlNT+7wpwPnJZ8mSJezbt4/IyEgOHTpEQUEBRqMRjUZDZ2enRGft4uIiCax1dHQwZMgQ9u7dS0ZGBqmpqRQVFRETE8PDDz/MiRMn+lwnJCQEmUxGeXk5VVVVTJ8+HYDu7u5BHSc7OzuqqqpQq9XS2nR3dzcymQxnZ2epTb3fksW1WZFWu6enB4PBgFwuZ8WKFezYsUNaBjOZTLS3t0sTT1xcHB9//DG5ublkZGTQ09MjMUteffXV/Otf/+LKK68EzofUOzs7UavVhIWF8eWXX2JnZ4dOp6O5uVmyvV6v58MPP+zzXW/6dLFdBoNBmpQ2bNhAVlYWRUVFqFQqamtr8fPzY/v27WRmZpKamirZxGAwoNFocHR05KWXXuLEiRMD1qq7u7vp7u7G29sbb29v8vPzqaurIykpCVtbW3p6egbNE9Lr9XR1ddHa2kpeXh7vvfcef/3rX3FxcenTT19//TUajYbu7m7a2trQ6XSSsxQVFcUnn3xCbm4ux48f7/NgEARBsoXYVxezRW/6d/EtzmKxSMudPT090oNbHGNTp04F6GNvd3d3qqqq2L17NxkZGXR2dqJUKjl8+DBVVVW0trZK5fe2oVi/3mWJOQuXCrHNXV1d6PX6Qcer2WzG39+fzz//nDNnzpCVlSWNReCCNhUEQRp3arUaBwcHXnnllQFjorCwkBdffBGZTIZarQagqqoKnU4njffp06dz+PBhSbhQ1F4S62g0GrFYLAwZMgSdTse0adMG2MFisdDW1kZZWRmJiYnMnDlTWu5qbGwkJSWFxsZG3n77bcrKysjPz2fhwoVSjpE4bsWyRI2k+fPnS2OgN9W+2HZBOL9bqaqqCoPBwNKlSzlw4ACxsbF9nBCFQoHFYqGqqgq9Xk93d7fUF83NzaxZswZbW1vpIWpra0tVVVWfCLdKpaK5uVmSThD7yWAwMHz4cKkPjx07JuVuCIIgjXlRLkEcS0ajkSlTppCWloZOp0OtVjNnzhycnZ3Zv38/lZWVtLW1ERER0WdOF+n/DQYDpaWlvP7665LoY+9r9p5H9Xq9FOnW6/U0Nzfj7OxMUlISlZWVuLu7ExMTw6effkpubi7Hjh2TXiR6t0Es18vLCzc3N4qKiqipqWHixIm4uLig0WhoaGiQxkdbWxtr167F1tYWvV6PXC6nra2NNWvWSBGb3jbp/dve3h6DwcCmTZs4deoU586dw9vbm6ioKDo6OnB1dSU4OJjPPvuMM2fOcOrUKUaMGMHmzZupqKigvb1dyhe7HHDZ8dC0t7ezf/9+FAoF06dPx87Ojt27d+Pv74+9vT0FBQXMmjWLw4cPS7TVQUFBA94uxDeK2NhYmpub2bJlC6NHjyY/P5/k5GROnTpFcnIyTk5ObN++nbi4ODo7O5k4cSInTpzg7NmzTJs2DX9/f7KysnBwcGDo0KHSMgWcvwkqKyvJz8/Hzc2NUaNG0djYyLFjx5gwYQJhYWF9js3LyyM/P59hw4ZRUFBASkoK6enpRERE4Ovry+bNmwkMDGTWrFmoVCrpzUpM1JswYQIlJSUolUpSUlJQq9Xs2rWLG264AYvFwo033ihRhCcnJyMIAt999x0KhYJ58+ZRUVHBuXPnmDVrVp/ls0OHDvHZZ5+xfPlygoKCGDp0KJs2bcLNzQ2tVoufnx81NTV91Hl7enqktyPx78DAQDIyMkhJSSErKwt/f398fHykJbvp06dTX1/PkSNHWLBgAVlZWRQXFzNjxgxcXFw4cOAASUlJBAUFkZaWhqOjoxTiFe3Y09PD4cOHMZvNEr19VlYWLi4ujBw5krq6OrKyspgxYwY+Pj598lvS0tIoLy/Hz88PpVLJsGHDCAwMHDB2SkpKOHLkCCNHjqSkpISQkBDq6uqYNWsWNjY2bNq0CZlMxvz583F3d+9zjYMHD16SLaZMmUJ6ejoAQUFBZGVlMXfuXDIyMlAqlbi5uXHu3DnmzJmDWq2Wlh5GjRpFQUEBtbW1jBo1iqNHjzJ16lTa29s5cOAAo0ePZvz48VRUVHDgwAGGDBmCQqFg8uTJ0tu8yWQiPT0dtVrNiBEjSE9PZ+LEiZw+fRpfX18SExN/FLNuaWkpR48eZejQoRKNff/xGhcXx+bNm3F3d2f27NmcOXOGlpYWZs6ciUwmk3bkLFiwQLKpxWIhNzdXUuzOy8sbMCbEPBtRIHHUqFEcOXKEnp4enJycaGpqYubMmahUKo4ePdongrtz504iIiLo7u6mqamJiIgIvv32W2xsbCTa+rFjxwLnX7Ruu+02Zs2ahZ+fH4mJiTg5OZGZmYlGo8HHx4eYmBiOHTuGSqXC2dkZFxcXzp07h62tLTExMeTl5dHS0sKMGTM4efIk3d3d+Pr6EhcXx6lTp2hoaCA5OZnU1FSio6MxGAxUVlYyZ84czpw5g1qtZvbs2VgsFtauXcvKlSv7RCsEQWDPnj3Y2NhIy59xcXGkpqaSkpLCyZMnpUhGSEgIFotFcqhEan61Ws3mzZtJTEwkNzeX+Ph42traUKvVTJkyhZ07d0rSBSJBnyCc33F09uxZJk2axLFjxxg6dChms5mamhpmzJjB6dOn6e7uxt/fn+joaKqrq9mzZ4/ksI8fP579+/dLc7qdnR379u1j7NixUoTSzs6OkSNHSjlGlZWVpKWlkZKSwqlTp/Dz88NgMFBTU0NAQABdXV3Mnj1beukKDg6mu7ubrVu3otPpmDdvnhRxFwSBs2fPUlJSQlJSEqmpqUyYMAGlUklOTg6urq6MHDkSW1tbNmzYwJgxY4iOjkYmOy/OKebQdHd3ExoaiqOjI7t372bMmDHEx8ezZ88egoODpeXrESNGcOrUKVJSUjCZTHz//fcMGzaMqVOnolQqqa6uxtbWFh8fH9RqNd999x3Ozs5S3tLWrVtxdHQEYMKECX2eS79nXHYOzaWid7MvRlPdn/66/zn9v7vQ8Re61g/V40J1utB3l1KOxWJh48aNlJWVcd111xEaGorBYGD58uU899xzREZGDmjfxbgc9uzZw+7du3nllVek/IrB7DJYXX+ovv3bdSHb/lLl/xL4oev/kE0vpdyL9Uf/73/MeeIxP8aG/ytcyvi/lPb1P+5Cbf8x96ogCLz77rvY2tqSmJiIWq0mPDxcylExGo3ccsstPProo8TFxV1w/rgQfs680fv4AwcOkJmZycyZMxkzZsyPGps/9t77Kf30Q/UXz/2hefdi8/OlXqd/uZc6J16ozAvVcbDr/tQ6D1bHH7LLb+He/qVweS2g/Qj8UCf2z234MeX8mAHySx37Y8sJDg7uM+FqtVqWLl0qhagvtC7eH4Ig4ObmxtSpU9FqtVKS2S9Z1x977n/L/r9UmT/1mj+17T/lvN/iJPdzxtSPOfdCx/7QNZYvX87Ro0epq6tj9OjRfRStxR1u4rbbS72/LrXOl3K8IAj4+/szbtw4Ro0a9aPHyi9x7/1S7b3Uuv8UO/2Ua19qmT91bF1q+Zd6zd/i/f1L4A8bobHCCiussMIKKy4fXDYRmv5ERSJksr60ziLJUO+lEfG8C9G4i99fytJT/+tZYYUVP4z+9+V/o/yfSxX/v4SYHC2TyX5SnUV7ymSyAfPapcx5/22I17dYLD9JmkGESCD3c8q4EEQbAhct/6eMLfEckTPtvw2xLf8NO/2WcFntclqzZg3vv/8+//rXv3j++ed599132bBhg3TzmM1mvvnmG9avXy+do9VqefbZZ6VtthdCfX09Tz31FKtXr+bcuXP87W9/45tvvpF2FXzyySfs37//v9k8K6y47CBOtOfOnePJJ5/8xaUkxIe3Wq3m5Zdf/lWlKno/IH/ouKKiIrZu3crrr7/+k5lcN2/eTFpa2oDPu7u7ef755wdwBl0KJf+ltuGHYDab+frrr1m7du1POl9k612zZg3btm372fXpjd5t3L179w/O6xaLhddee+2SZB56l71mzRry8vJ+foV/ABaLhU2bNv1kW/+ecFk5NN7e3txyyy0Sl8jKlStxcXHBbDZjMpmQy+WoVCqqqqokemgnJyd6enokXSaLxTLolmk/Pz+amppwc3OT2EyzsrJQKBQolUoMBgMJCQm/RrOtsOJ3CUEQpG3Y7u7ulJSUSPeq6Ij0/n+w8y8k6SB+dvDgQSorK3FxceFPf/qTRBUvPhD7yxGI80LvMvof05uivn99+tPo967f0aNHKS0tveB54rwjCAJr164lOjqa2267TeIK6V1+7/r3proXrwcwa9YsRo8eLZ0rlq9Sqejo6ECj0UifGQwGNmzYIG27H6yNgnCee2nv3r0Drnex/um9jV7c3q5QKHBzc6OysnJAX/SWPhisbJFFWqSfqKuru+B1BrN1/zm+/2ciCaPZbCYlJUXifRpsHAjC+S3xt912Gx4eHtL3/WVmxGNbWlr4/vvvEQSBRYsWERkZCfz/547YhsFkMgZrR+9+6j8WxG3wMpkMBwcHKisrLzh2LxdcNktOAHPmzOlDay1u0dy0aRPNzc0SEVJBQQH/+Mc/UKlU/P3vf5dCfhqNhiNHjlBaWkpYWBhXXHFFnyUkMTwok8m4/vrrWbZsGbW1tRIZWu8tuFZYYcXFYTQaWbVqFcHBwdxwww3o9XrWrVtHTk4Ojz/+uMT+nJuby/z58/sksvaWDSkrK2P69OloNBq2b9+Oh4cHlZWVPPTQQ7z22mtce+21JCQksGHDBh5++GE2bdqEWq3GYDBgNpsJDw8nPT2dhx56iM2bNxMfH49CoeDIkSPcfvvtvP7668THx5ORkcH48eNpaGjAYDDwwAMP9AnhV1dXk5ubi16vZ/r06ZSWllJdXU1eXp5Uzrx58wgMDJRI9wRB4Ny5c2RnZ2MwGHBycmLUqFFkZ2cTHh7OokWL+swp33//PSUlJcyfP58PPviAJ598klOnThEUFIRer6eqqoq8vDzuuusutm3bRkBAADNnzmTz5s1YLBays7O59tprkcvlpKens23bNlJSUggKCuKdd95hyJAhwHl6g8bGRq655hqpjYJwXtogLy+PmJgYiouLJfvPmDGDmJiYPv2ze/dunJ2d0Wg0zJgxg61bt+Lk5ERhYSErVqyQ5tKqqipeeeUVnnzySerr62ltbcXb25vy8nJOnz7N3Xff3UdapKysjLfeekva9l1UVMRLL72Eu7s7f/7zn0lPT6e2tpby8nLuv//+PgSHGo2G77//HmdnZ6qrq7n++uv5+OOPsbe3p6SkhDFjxhAYGMjq1auJiIggPz8fBwcHRo8ezapVq0hMTOTo0aPMnj2b06dPExYWxpw5c3j77be57bbbKCkpwWAwkJmZSUREBImJiZw6dQqNRiPRbOzZs4ehQ4eyZ88ekpOTiYmJYfv27ZJUzPLly1m7di12dnZUVFQwadIkabs1IEWmWltbaW1txdPTk8WLF/Phhx/i7e3NsGHDpGdaUVERV111lUQ++PTTTyOXy3nsscf62OVywWUToZHJBkofwHll5fDwcPR6vcTbERgYyOOPP05BQUEfsrutW7dSXV2Nr68v27dvv2h4OjAwkPHjx/Ptt99y/PhxkpKS/jsNs8KKyxRKpZKQkBBGjx6Nm5sbSqWSRYsW4e7uTk5ODp988glGoxEnJyd27NjR59zGxkbWrVvH7NmzmThxIi+88AIqlYrKykruuusu3NzcOHDggETPHxQURG5uLvD/CfduuukmUlNTmTFjBh4eHuTm5koPcl9fX/Lz8yWRxbi4OIkT6OabbyYjIwOtVtunTnl5eWRkZDBmzBj0ej1r1qzB1dWVuro6SktLJekSe3t76RyLxcKbb77J2LFjueqqq/j000+B89xCEyZMGCCXMnLkSHJycvD396e8vBy1Wk1XVxdubm58/vnnuLq6Ul9fT35+PgaDgdraWurr69mxYwfz58+nsrKSoKAgAOLi4li2bBnbtm3D398fPz8/xowZw4EDB6itrWXMmDF9ri2TyQgLC2PYsGEoFIoB9u+9NGY2myWZmOHDh5OWlkZpaSlz587FwcGBTz75RDo2ODgYhUJBaWkp7e3tRERE8NFHH+Hk5IRarSYzM7NPRMHX1xc/Pz8SEhKQyWQEBgZyxx13cODAAUpLS9mwYQMeHh4UFhZSWloqnScIgsQTtGDBAsrLyzly5AjNzc2EhITwyCOP8OWXX+Ls7ExgYCAjRowAkEj1CgsLmTZtGgEBAbS2tnLjjTeyc+dOKfqh1+sZNWoUsbGxZGdnM2nSJDo6OggNDcVsNnPkyBHCwsIIDw9n2LBhdHR00NLSwoYNG5DL5SxYsIDq6mr279+PRqPB2dmZpUuXsm3btj7tVygUdHd3o1KpePTRR0lNTaWpqYnc3FyuvfZagoKC2LlzJ3PmzGHkyJG88sorWCwW/Pz8eOyxx6isrOTw4cO0trbS3Nz8qy7D/tK4bByaC0GlUlFeXo6NjY10w7m6uuLk5MSIESOkpSaAc+fOERUVxRVXXCFxqlwIcrmcm266iY0bN5Kfny8JLl4oOdkKK6y4OGxtbXFwcMDBwYGuri4qKioYNWoUt99+O/fee2+fY9VqNS0tLchkMoKCgtDpdAAS9fyIESPo7OyUjrexsZHuZxsbG+zt7bG1tZXo81UqFV1dXQO2u4ovSmK5KpVKWrbqn9syYcIE7OzsePTRRykrK8NgMJCcnMxrr70mLf2I5YowmUxUVlZKdffw8JDYivsfK9IteHh48P333xMXF8e6deskttju7m6Sk5N59dVXSUhIQKlUIpPJJLmRjIwMJk+ejJubG3Be2sDBwWHAssby5ctJTU3l7bff7sMQ3bsuarWa1tbWPvbvbQ8bGxtuv/12SbOqqqqqD7OzKIsC5+fSpUuXsn79erRaLfb29mg0GpKSknjuueckVfcLwcnJSSJ9bGpqQi6XM2HCBN555x0iIiL6HCuKrcpkMoYOHUp9fT1KpVKSpPDy8pKY42UymSSLI5fLUSqVkiSJOG5EYVXxOBcXF9544w2WL19OYGAgKpWKiooKFAqFdGz/snvXKTIykvr6ehQKhSRP0t/hEM91dnbG1dWVsLAwjEYj9vb2eHh40N3djVqtlhzQ1tZWLBaLJE8zcuRIamtr+eSTT3j//ff79MXvHZedQyMKfXV2dmKxWPjyyy/R6/VotVqampqA84rLHR0dCIJAQkKCROMuUv2fOnWKI0eO9KG1NplM6PV6qVw4/+YRFRXFsGHDpHXu8vJydu/ebXVqrLDiEmBvby/JK4gU+SLl/pAhQ1i9erWkmN37ngoODsbR0VHS9ZkyZQoqlQqNRkNHRwdVVVVMmzYNOzs7ysvL6ejokCRFRKmCwSRBXFxcKCwspKCgALVajV6v7yMlINbNYDAMcATy8vJYsWIFU/+Pebmuro6tW7dy9OhRSRuroqJCcr4ASfn7yJEjtLW14ePjw5AhQ+jq6pKEY3tDLpdz5ZVXsmvXLu666y4yMzOJjY3Fy8uLhoYGtmzZwtGjRyUHoru7G4vFgq+vLzqdjkmTJkkRqq6uLukYuVwuMQufOXNGEmrt7VyJ/dXY2CjpHxUXF1NZWUlKSkofXSij0UhjYyNvvvkmZWVljB49mpKSElpbW6mqqmLu3LnSvCvOw42NjQQHB+Pu7k5XVxfr168nLS2NsrKyPnVQKBSSIyhKNBiNRok5+ezZs+zZs4cjR45Ic76I6dOnk56eTmdnJ+3t7aSkpCAIAs3NzRKbeUBAAJ2dndTU1Eg2EiUNxL4XPzMYDJJ0R09PDzt37kSr1ZKSkkJaWhrr1q1DrVaj0+loaWmRtNMaGxulsnvXqbW1lZSUlAEyB4MlYre0tNDc3Iy9vT0hISGS1ERkZCQ9PT1UVlZSXl7O7NmzsbOzo6urS8qbmjNnDn/729948sknpWXGywGXHQ9Ne3s7hw8fRqFQMGXKFGpra8nMzCQmJkaScD9+/Dj29vZERkbi6enJgQMH8PHxYcSIEezevRuNRsPcuXPx9vaWoi7Nzc0cOXIEJycnJk+ejKOjI4IgUF5ejpeXl6TnIw6iKVOmWPNprLDiB1BVVUVaWhojR47k7NmzTJgwgbNnz+Lm5sawYcPYtm0bTk5OzJ49G0dHxz40CfX19eTm5kq08pWVlTzxxBPcfffd+Pr6Ehsby9mzZykvLycmJobs7GzGjRsnJUdGRERw/Phxxo0bR1FREc7OzgwdOlTSN+vs7CQ2NpZjx44RFxcnRYVGjhwpnddb+qK8vJy6ujqUSqVUn4MHDxIfH09SUhJFRUUUFhYyb948adlJEAS6urpIS0vD1taWqKgobG1tOXjwIGFhYVI+T2+IuTLDhg3j7NmzxMTESDkSBw4cYOTIkYwZM4Zjx45JkYi1a9eiVCrR6XQkJiYC5zc62NvbU1xczMyZMzl8+LBE5W+xWLCzs2PEiBF9ItVqtZrt27czY8YMLBYLZ8+elezfOyfDbDaTlpaGSqXCzc1NykdpaGjAy8uLmJgYMjMz6ejoYOrUqahUKgoLCwkPD8fW1pbKykp2797NsGHDmDRpUh8RRYvFwt69e3FwcJASY4cOHUpGRgZTp06lvr6e9PR0kpKSGDVqVJ/6m81mMjMz0ev1BAQEEBkZyT/+8Q/8/f0ZPnw4o0aNwsXFhS1bthAZGSlpxA0bNoyTJ0+SnJxMUVERtra2BAYGkpOTw5gxY8jOziYyMlLSPRPHqq+vL8eOHWPEiBFSrteuXbtISEiQxH6TkpLIyspCr9fj7+9PeHg4hw4dkoQ9S0pKmD17dh/7fvjhh1RUVDBt2jRJufvYsWMkJycTEhJCZWUlRUVFuLu7Ex8fj8FgIDU1FTs7O4YOHdpHnPVywmXn0FhhhRV/PAiCQE5ODq+99hqffPKJtNxiBaxfv55z584xc+bMPg9yq33OO0cPP/wwM2bM6JN4+1uGIAi8/fbbKBQK7rrrrl+NS+i3iMtql5MVVljxx4WtrS1Lly5Fo9Hg6en5a1fnN4OFCxeSnp5OS0sLI0aMGFSM94+Knp4eZs2ahZubmyRQ+luHyWQiPj4ei8WCwWCQhEGtsEZorLDCCiussMKKywDWCI0VVljxq6A3Gd7/SpLgYpIAP6fM34qswn+jLj/HZoMxC4u7x36M1MUv1W/imOsvVSPWRyTK652rJf6I1xXJ6nqf17+si7W7vxTPz+mv39LY+y3A6tBYYYUVvwrMZjMbN25ErVZzxx13/E+u2dHRwQsvvMBdd91FaGjozy5PEASys7P57rvv+Oc//9knefV/DUEQSEtL4+DBgzz++OO/2AOuo6ODV199leuuu464uLg+D/cfQkZGBgcOHMDBwYGysjJGjhxJS0sLISEhdHV1sXLlykuuw4svvsidd955yf3W3xERUVdXx6uvvkpYWJhEE3DNNdegVqt56623GDFiBMuWLZOclTfeeIOuri7uu+8+cnNzOXPmDA4ODvT09LBs2TLeffddDAYDYWFh6PV6rrvuOuzt7fnkk0+oqqqSdtLq9XruuOOOPs5STk4OmzZt4qmnnvrRY0cQBEpKSli9ejXPPfdcn11mf1RYs4mssMKKXwUKhQIvLy+qqqoApJwA8U1ZpHEXKel7U8P3h0jf35sGvj/FvcViwd7enpaWFrq6ugaUIVLH969Db7p4kV24tzSDp6cnpaWlF5RH6C9r0Pv/wdoh1qE3Nb+4bbh3W3pv5xXP8/DwoKys7ILyCv3bMlgde0sqmEwmXF1dJeqLmpoadu3a1ce+F9N2kslk3HLLLYwePZqqqipWrFjB2LFjUalU1NTUDJCu6G3f3nBxcZH6rbdtRDv0lpoQ++DYsWMUFBT0KUsmk+Hv709NTQ2xsbEsW7aMr776itTUVHx9fbFYLLz99tu0t7cjCAJVVVXs3LkTk8mEo6Mjq1atYubMmVx33XXI5XIcHBwQBAF7e3tuuOEGysvL+fDDD1GpVLi4uKDValmyZAnLli0bVBrHy8uLc+fODZD8ELeI9/5/MLt4eXlRXl7eZ3yIdrjY/+L1RGqSywXWCI0VVljxq6D38oHBYODgwYPU19fT1dXF4sWLefXVV0lKSiItLY1JkyZRU1ODTCbjvvvu6/PWbTKZWLduHb6+vpSVlXH99dfzyiuvEBUVJTH7RkdHs3nz5j4OVG90d3fz9ttvo1QqqaioYPjw4YwbN05ijp02bRp1dXW4uLhw9uxZli9fjk6n49SpUxL/R2NjI2+//TZ///vf+fTTT5kwYQIhISGcOHGClpYWgoODiY+P58SJE+Tl5TFlyhQmTJjQJ5Ki1WrZvHmzxAlz6623UlNTQ2trK7W1tcTHxxMXF8e2bdtwc3OjtLSU2267jYqKCgoKCmhraxtAxCbKK3zwwQd4enoyZswYZDIZDQ0NtLW1cfPNN7N//34cHR0JDg4mNzcXpVJJXFwcn332Gf/+97+Ry+UIgsCGDRvIz88nJiaGrKwsvLy8EASBlJSUQSNCY8eORS6XU1paKpEUTps2jX379lFeXs5bb71FT08PDz/8MOnp6TQ0NHDu3Dn+8pe/DKDmF0UWy8rKuOqqq5g8eTKpqamUlZUREhLC+PHjSUtLQxAERo4cydtvv83kyZMJCQnBycmpz7iTyWSYzWY0Gg2CIEjbrIcOHUplZSU7duxg+fLlZGRkkJSUhFwuR6FQMGTIEJ544gmeeeYZli1bhkKhQC6XY7FY0Ol06HQ6icxPoVDQ1dVFU1MTFRUVA6RxxHp0dHTw2WefUVBQwNNPP01mZiZGo5HCwkJuu+02cnJykMvl1NfXs2TJkj7kfDKZDJ1Ox/vvv09+fj4PP/wwdnZ25Obmkp2dzZIlS3B1dSUrK4szZ84wd+5c2tvbOXbsmMT38/DDD/+qkcVfEtYIjRVWWPGrIzMzk/T0dLy9vUlPT8disVBVVUViYiJJSUkUFxdzyy23kJaWNiC6YrFYsLW1JSIign379gFQX19PbGwsKSkp7N+/X4oqzJw5k8DAwAHXt7Ozk2jqH3vsMb777jsMBgPFxcXcdddd9PT0kJ+fz5w5c/D392fVqlWsWrWKCRMmMHXqVGxtbXFzc6OsrAyFQoHFYqGpqYlPPvmE0aNHc8UVV+Do6Mhnn32GXq/Hw8ODPXv2DHjjzszMpKysjLlz51JRUYGTkxMffvihtK345Zdf5osvvkAulzN//nyam5vZvHkz77//PtOmTSMlJWXQh5O7uzsFBQWsXLkSR0dHDh06hJeXF1lZWRQWFrJ9+3ZCQkLw9fVFqVRSXV2Nj4+PFD2A/y99EBUVhZeXF5s3b8bBwWEAG29vDJbbIX7m6+vLn//8Z7Kzs8nPz2fTpk24u7tTWlpKRUXFgLJkMhmzZs3iL3/5C++++y5ffvklNTU1+Pj4sGfPHkpLS9m/fz9xcXF4e3sTHBzM2LFjcXR0HFCWIAjs3r2bG2+8kQceeIDk5GTgPMPxbbfdxtq1ayXHV9SRkslkPPHEEyQlJXHjjTeydetWqbzc3FyWL1/OiBEjuPnmm6U2t7S0kJOTw7Fjx9BqtYM6fSqVihtvvBGz2cy5c+ck3puioiLKysrYv38/9fX1FxQ/trW1ZeXKlcyePZu3336bDz74AEEQcHBwYP/+/Xz88ccYDAZcXV0l/p7m5mZuvfVWTp061YdR+/cOq0NjhRVW/Oqorq7Gz8+PGTNm8N577+Hi4oJCocDOzg6VSoWdnR1KpVJaCukNQRDQarW0t7cPkDcQqePz8/Oxs7PrQznfG2L0wNHREU9PT3x9faWHgqurq8T2ChAREUFNTQ0lJSXY2dn1kVXoDbPZTHFxMXK5HF9fXxITEykvL2fUqFHcfPPN/PWvfx1wTlRUFE1NTRQWFrJgwQKMRiMNDQ3A+QerxWLh3LlzEo1+REQE5eXlVFRUYGtrO6iendg2e3t7XF1daW9vx8PDg2nTprFq1SpiY2OZP38+jzzyCGfPnr2k/rK3t+e2227j+eefl5zIHyv74uDggK2tLQqFgoaGBmxsbJg0aRLvvPPOoHkycrkcZ2dnQkJCcHR0pLi4mGHDhnH11Vfz4osvMmLECOLi4vjb3/7WJwo3mBMhl8uZO3cuKSkp7N27tw/rc3JyMiqVijfffJOkpCTpfDEC87e//Y3XXnuNt956S2JSjo+P5+abb2bPnj3o9XqprCFDhjBnzhzuvfdeQkJCJPmH3hDHt0qlkpYBq6qqsLW1xWKxcMstt3DgwAEpV6c/RMmQESNGoNPpqK6uZuzYsdx3332sWLGCyspKxowZw5/+9Cfuuece5HI59vb20v1wsSXD3xusDo0VVljxq0Bkye3u7paWhA4ePMjevXvp7OyUaOW7u7ul3AlRqqD3g7OpqYmNGzdiMpno7OyksbFRoowXy09ISGD79u0UFhbS1NREU1PToHkmTU1NVFVVERoaipubG11dXRiNRhISEqipqaGhoYHy8nIWLVpEREQEmzZtkij99Xo9crmc7OxsqqurUavVxMXF8eKLL7J7926ysrKIjo7mnXfeIT09nfT09AF1MBgM+Pr60tXVRXx8PO7u7oSFhZGVlUVlZSXjxo1j0aJFZGRkSMzFCxcuxNfXl23btnHu3DmpLr3bJdqxp6eHqKgodu/ezf79+9m3bx/FxcUEBwfzyCOPcObMGVxdXTl37hy5ubl0dHSg0WgkiQR7e3vq6+spKyujq6uLl19+mcLCQlpbW/nyyy8HfThaLBY0Gg06nU6SOhClHcQ+8vX15cyZM2zfvp19+/bR0tIyoBylUklHRwfl5eVMnjyZSZMmsXr1atLS0jh06BAFBQVMnjyZ5cuXU1hYiL29PaWlpX0iEKIttFoter2ev/zlL1RXV/PRRx9JzoDRaGTlypWYTCaCg4Pp6Oigs7OT7u5u3njjDRobG/H09CQsLAwbGxs6OzvRarVcddVVxMbG8tRTT9HZ2SnZrqGhgfr6etauXdtHSkKsi2gHMU/ps88+k5bD6uvrJSkK0XHvDXEpsKOjg5KSEpYsWUJwcDDvvfcex44d48yZM0RERPDuu++SlpbGyZMnpXHQW8LhcmFvsfLQWGGFFb8KzGYzWVlZtLW1MWnSJHJycjh37hxTpkzB1dWV1NRURowYQVNTEzqdjqioKE6dOkVycrJE0Q/ndYN27tyJh4cHGo2GoKAgqquriYiIoLOzk87OTsaPH8+RI0fo6urCycmJmJgY/P39+7y9P/3003h4eBAfH8/IkSNpb28nNzeXcePG4evrS0lJCTU1NRJ1f0dHB3v27JHqkpycTHFxMWVlZQwZMgRnZ2cCAgLYuXMnjo6OzJgxA7PZzI4dO1AqlcyYMQMnJ6c+u16ys7PZv38/crmczs5Orr32Wvz9/cnKysLJyYnY2FgcHR3JyclBo9FIrL+NjY3s37+fwMBA5HI5SUlJEuGaIAhUV1dz6tQpEhMTCQgI4OTJk5ID4OnpSW5uLiqViqFDh2JnZ8eOHTsIDg6WHKCsrCwCAgIIDQ1lz549jBs3jqqqKlQqFf7+/jg7O5OWlsasWbMGSDUYDAZOnDhBY2MjSUlJBAYGkp2dTVtbG7GxsZw4cYKxY8eiVqs5ceIEiYmJxMXFDdjeXFpaSklJCR4eHowYMQIbGxvJ+Z05cyYmk4nS0lJsbW2JiYmhoaGB3NzcPrIBotN67NgxvL29SU5Oprm5mczMTCIjIzl37hwxMTGSnpODgwNHjx7FYrEwfvx4cnNzMZlMaDQaxo4di7u7O0eOHEEmkzFp0iQADhw4QGRkJHV1dXR0dODh4SFFVmbMmCEtCQqCQEVFBWfOnCEpKYmcnBwCAwMl3SdPT08pkV2MFkZHR/eJBlosFjIzM1Gr1fj7+xMTE4NGo2HHjh1SFM5oNPL999/j6OjI9OnTKS8vp7a2lri4OE6ePElSUtKAe+H3CqtDY4UVVvzhYbFYePTRR5kwYQJXXnnlr0InLwgCzz//vKRDpdVqiY+Pl3I4rLDCiovD6tBYYYUVf3h0d3dz7NgxnJ2dGT169K9CgS8IAmq1mpMnT0rCkP13xlhhhRUXhtWhscIKK6z4BdCf7+S3WPZ/s44/tg5WR82KXxqXXVJwb6rq/jTXP+fY/uf1JrX6KXX7sRisjj+23r3PscKK3yJMJhOtra3/9Z0XZrOZ1tbWQXdM6fV6Kisr6enp+VFl6vV6cnNzgfNLWNXV1ZSWlva5htFopKSkhLq6ukHvQ7VaTVFREVqttk+yZ1NTk5R8/FPvX0EQaG9vp7i4WOJNEQnZRNI2vV7fZ5eO2Wyms7PzokSA3d3daLXaPiR/Wq1WSgAW58vOzk6JyC8nJ2fQHT+9y21ra7voMfD/E47VajXd3d196mgymWhqakKj0UjJ3Wq1mvb2dulHTFZWq9V9+kkQBDQaDRqN5mePRZHIbrDxZrFY6Ojo6JPEbcVPx2Xl0AiCQH19PU888QSff/45W7Zs4c0337zgPvv29naeeeYZVq9ezbZt23jrrbdobGz8wes0NzfzyCOP0N7efsn1En+++uor2traflS74Pxk+frrr7Nr1y6pzLNnz/L3v/+dysrKC16zP1pbW1m3bt1ltVXPissHlZWV3Hvvvf/1Cb6hoYG77rqrz64TOH+frVmzhkOHDpGTk3PJ5VksFg4cOICfn5/0WXV1Nffffz+tra3A+XsyIyODf/7znwPI70QYjUbeeOMNPvroI+n/zZs309XVRWBgIPv378dkMv24xv4fzGYzBw4cIDAwkNraWlauXMmdd97J7bffzvr16yW+lKNHj5KdnY3JZGLLli2UlJSwcePGAfOoIAgUFxeTmprK/v37+eKLL7BYLOzbt4+zZ8+yfft2amtraW1tZcuWLRLfDYCHhweHDx++qHMmkuj9UJs2bNjArbfe2mdetVgsfPDBB+zevZsNGzbwn//8h6NHj/L222+zdu1aVqxYwY4dO3jmmWfIzc3liiuu6FOfqqoq5s+fz44dO36RubK2tpa77rprwC6l7u5u/v3vf5Oamvqzr2HFZebQyGQyHB0dSU9PZ+rUqSxcuJCUlBRsbW0HPd7R0ZHjx4+TnJzM/PnzmT59upRRfrEfmUxGV1cXLi4ulxQ5qauro7CwEJlMRmxs7CVdo/+NbmtrS25uLgcOHACgs7OT//znP1RWVhIUFDTgurW1tRQWFg4o097entjYWElA7adGqayw4peA+CZeX1+PyWTC19cXo9FIe3s7DQ0NEu9MS0sLarVaetuvra2VohidnZ3SFubGxkZpy3dLSwtms5n29nbq6+ulKIT41i1GC0RYLBbOnDlDbm4uV155JWPHjqW1tVWqmxgNEN/se5/b0tJCa2sr3t7eEoOrnZ0dQ4cOlbaIt7e3U1RUREJCAkOGDBl0ycXGxoarr76atLQ01Go1KpWKlJQUQkJCJN2h4uLin2RrcQeQo6MjgiDw3HPP8dJLL7F48WKmTJnCunXrGDt2LImJiaxbt46ioiJKSkoYPXo09vb20tzTG6LDMXLkSL7//ntaW1vZu3cvY8eOJTIykvXr17N161b8/PxISEjgxIkTNDQ0EBwcTHl5+UVJ3ZYuXUpYWBhGo5GOjg6am5ulMSBCqVQSGhqKQqHos/Otq6uL7du3k5yczIoVK4iNjSUoKIj777+fCRMmYLFYWLp0KStXriQ8PBw/Pz9Wr14tRaLS0tLQ6/XS1mwRFosFtVqNVqulrq4Oo9FIXV0dOp1OGpv19fWSdII4PgFp/JlMJurr61Gr1djb2+Pt7X1BB9eKH4fLg++4F0pKSrC3twfOC6ONHz/+gmu1tbW1kmNy7NgxJk6cSEZGBvX19SxYsICvvvqKFStWkJubi06nIycnh6VLl1JcXMzQoUNpb2/nyy+/5M9//jNHjhwhLCyM4OBg0tLSUCgUFBcXs3jxYh566CHmz5+PVqslKyuLkSNHUl1dTUlJCe3t7SQkJNDS0sKpU6cYPnw4x48f56GHHuqTmNjU1ERoaCiVlZWYzWb27NmD2Wxm+vTpKBQKqqurKS4uprq6mvnz5/PQQw+xcOFCAPbt24ePjw+urq7U19cTFxeH2Wzm5MmTGAwG1Go1s2fPZsuWLdja2hIfH39R9k8rrPglIAjn2Vp1Oh1yuZzCwkLuvPNONBoNGzdu5NixY9xyyy3AeS2fmpoa5s6dy65du3BxcWHr1q08/PDDPP744yQlJeHs7Mzp06dZsGABSUlJbNmyhVGjRtHQ0EBWVhbDhg0jISGBAwcOEBMTQ3Nz84D6VFdX09LSQlNTE8ePH5dYiYuLi7n22mv529/+xsSJEwkPD2fRokXSuUVFRQO2vnZ2djJ8+HDq6uqIjo6mrq6OxsZGJk2adME5qaGhgZiYGEaOHMmBAwcYPXq05CQBhIeHc+bMGWJiYqSXkubm5kGZdf38/AgODpbOPXPmDMOGDUMmk0mfFxUVMWTIEFxcXKitrUWlUmGxWKirq6OoqEiag5ydnTlx4oT0QgfnXyDnzp1LS0sLmzZt4pZbbkGtVqPX61EoFDg7O1NcXIyzs7MktyCW7e/vj4eHB5WVlYwYMWJA3bu6unj66adZsGABMpmMDz/8kLlz50oEc72lDMS69Ia9vT0pKSnceuutPP7448ycOXNQ8cbhw4ej0WiYOXMm33zzDTk5OQwdOhSDwYCfn9+AchsbG7n33nu59tprOXjwIHFxcdjZ2VFaWso//vEPPvroI8aPH8/OnTuZPXs2Hh4eHD16lMjISGnJaevWrbi4uPD111/z1FNPDToOrPhpuKwiNIIgkJqaSkREBIcOHaKzs3OAXHvvY0+ePElgYCCZmZlUV1ejUCgICQkhNTUVuVyOnZ0d6enpZGRkkJCQwNmzZ3F0dOTkyZMkJiZiNpspKSnBbDbz6aefYmtry3vvvYebmxvOzs60tbXh6OiIi4sLV111Fa2trdKbxquvvkp8fDy2trZs2rSJ9vZ28vLycHNzo7i4uM9biCAIFBUVkZycjFqtJisrC3t7e5qampgwYQJ1dXW8+eabODo6UldXJzGCXnXVVQiCwPHjx4mNjSU+Pp7s7GwcHR3ZunUr5eXlJCQk4ObmhlqtZtu2bcTHx+Pq6vq/7DYr/qAwGo18/vnnTJgwgXnz5pGRkUFlZSVOTk6sXLmSBx54gK+++orm5mbWr19PfHw8xcXFnDhxAplMRlBQkMQ0e80113D77bezZMkSjhw5Qk1NDePHj+frr79Gp9NJSshfffUVI0aMYOzYsQO2QysUCiIiIvD19SU0NJQ1a9Ywfvx45s+fL0kuyOVybr75Zq6++uo+5zY3N/e5b3p6elAoFISFhVFVVUVVVRU+Pj5UVFQQHR09qD0EQZCiPNdccw1bt26ltra2Tz1dXV0HOGK2tra4uLgM+On/AG9pacHZ2RlA0h/atWsXUVFRUsQKkHSOekcNZDLZoCy1cN55iIyMlEjb+vexyGrc+zOZTIarqytNTU2Dlung4IC7uzsGgwF/f39J/NFkMv3gUr/odP31r3/lkUce4d///jfvvPNOH2esd7tkMhkuLi4sXbqUTz75pI9+U394enoil8uZOnWq5JguXbqU8vJyTpw4QU1NDUlJScycOZMPP/yQtWvXMnr0aBITE/Hy8qKpqYnvv/8eo9FIVFTUJactWHFpuKwcGrPZTFpaGsuXL+e6664jJSXlgscKgsDhw4e57rrrWLRoEVdeeSWdnZ24ubmh1+spLy8nOTmZL774gnnz5lFUVISrqyt2dnYS+VJBQQFRUVFkZ2djNpvR6/WUlJQQHx/Pvn37GDt2LHV1dXh4eEhvOBMmTOD7778nKioKT09PCgoKGDp0KMePH2fy5Ml0dXURFRXVJzojCAKVlZXExsbS1tbGyZMniYiIwGg0Eh4ezvr163F2dsZisXDffffR2NgoXTMzM5OEhARiY2NxcHCgo6ODwMBAvvvuO6ZMmYKjoyMpKSl4eHgQExPDzp078fT0/F90lxV/cIjJox0dHSiVStzd3YHzD1u5XI63tzfu7u7MmzeP5ORkHnvsMSoqKlAqlcyZM4e//e1vODo6olAosLGxQaFQkJycTE1NDadOnSI8PJy2tjbi4uK48cYbWbBgAQ0NDej1+gu+6PSum0jKp1QqcXNzk8Q0RamD3ucrlco+CZ+tra14enoSEBBAUVGRpGwsl8vx8PCQlh5652eICsgqlYrIyEjs7e3Jzs7uI9JosVgGbCkXE4n7//R3FmxsbPpcr7m5meLiYol639nZWVJldnJyIiAgQHJqdDodfn5+Ur3F5ZTPPvuMmpoaxowZw/79++np6ZGcpe7ubry9vfHz85McHZPJJM0vZrO5j5xF/6Xu3vYV7S1GpS7Wbx0dHeTm5lJeXs4VV1zBu+++y6ZNmyTpCvG43n8LgsCiRYvIycnh9OnThIeHD1q+WAdxLIh/i0rrjY2NWCwW3N3dsbGxobGxUYryyWQySch02rRp/OUvf+mjKyYmD1uX/H86LhuHRhDOS73X1tYSGhqKSqVCp9Oxdu3aQXcytLS0kJ2djY+PD/X19Xz//fdkZ2djY2NDd3c3HR0dBAQE0NraSltbG/v27cPFxYXKykq6urpQq9Xk5ORIDo6/vz/FxcXIZDLy8vI4ffo0NjY2ZGZm4uHhQUdHh7SGbTQa8fHxobKyEoVCQUpKCpWVlcyYMYNTp07h4eEh3XyCcJ7ZUqTbDgwMZN68eRw7dozg4GDs7Ozo6upiyJAh+Pv7o9PpyMrKwtPTk87OTrKysqSwbWVlJQ4ODmg0GuRyOa2trZSWlqJWq9m6dStXX3019fX11hvKiv8JbG1tueGGG9ixYwe5ubmMGjWK8PBwHB0dKSoqIisri5tuuomDBw+SmJjI5MmTGT58OEVFRTzzzDN89dVXaLVatFotzc3NCMJ5NtWZM2fi4uKCSqVixowZ/OMf/+Ddd98lLy+PK664gs8//5xdu3bR3t5OTU2NNN4tFgvNzc10dHRgMplYtmyZVDcxkqnX62lpaRlwjwQHB0u5Mh0dHWzbtg2NRoO3tzcqlQpvb28OHDiAk5OT9HD/5JNPOHToEHD+QZ+WlkZRUREGgwGFQsHSpUvx8fHp82BvamoiODi4z7UDAgKYNm3agJ/IyMhB6yiiubkZg8GATCbD1taW6dOnU1hYSFFREVOnTmXUqFEolUqampqorq5mxowZpKen8/7770tliDlJhYWFxMbGEh0dTWRkJNXV1eTn57NgwQLmz5/PuXPnqKurw9vbm5CQEGkODgoKIisri0ceeaRPBEhcCm9ra0OtVktzrlarpbOzs88urOrqatra2khLS2P//v288MILGAwG3nzzTXJycigtLZXyI81mM1VVVWi1Wqkfq6urOX36NI6OjixdupTJkyejVqvp7Oykurq6z/NDp9PR3d2NRqOho6MDtVpNR0cH3d3dDB06FDc3N06cOEFOTg4333wzV1xxBR9//DF79uyhvb0di8WCSqXi8ccf5+OPP5bKaGtro6KighdffNG6YeNn4LLhoRGXZYqLi0lKSsLX1xeNRkNmZibTpk3rEz4UhPOU0zk5OXh4eGCxWNDr9aSkpODg4EBWVhajR49GoVBw9OhRTCYTAQEBdHR0MHz4cFJTU5kyZQolJSXodDpCQ0M5d+4co0ePZt++fcTExNDU1ERkZCTt7e3o9XpGjRrFwYMHiY+PR6VSkZubK+m0yOVyzp49y+jRozl79ixyubxP4m55eTmVlZUkJyfT1taGu7s7x48fRyaTMW7cOPR6PYcPH2b48OGSSquoBZOdnS1NTK2trWRnZ0sOVElJiUTzfeLECVQqFUFBQX2S66yw4r8Js9lMTU0NJpNJWkISc1i8vb3x9vamra1NoqH39vamtbWVmpoawsLCUCqV1NXV4eLiIuWaiBEYe3t7TCYT586dw8bGhrCwMGQyGWVlZVgsFhwcHPDx8ZE2DVgsFhobG9Hr9fj7+2NnZ0d1dTUWi4WgoCB6enpoaGjAw8MDDw+PPveITqfj66+/5pZbbqG7u5uGhgYcHR3x9vamp6cHmUwmiUz6+fmhUqloaGigpqZGWr6ura3FbDYTGBiIra0tJpNJitjA+Xlrw4YNTJw4kYCAgB9t68bGRg4cOMD111+PTCZDq9VSU1Mj5dUYjUYpyVe0rZhQ7ebmRkBAAHq9noyMDKZNmwacXz6qrKxEq9USERGBs7OztO3d3t6ekJAQZDIZ1dXVaLVagoKCcHZ2xmg0smbNGm666Sa6urrYtGkTy5Ytk5bJjEYjNTU12NraYmtri0ajwcvLi+bmZjw9PaWImWi3rq4ubGxsEAQBuVxOUFAQTU1NdHZ2YjKZiIqKksQea2tr6e7uxsvLC3d3d+m4gIAAKQLX2dlJc3Mz9vb2BAQESHIOOp2O+vp6PD090el0mM1mXF1daWlpkXJuampqcHNzk8RES0tLpfHo4+NDT08PpaWlBAYG4ubmRk1NDUqlEg8PDxobGwkNDbXOvz8Rl41DY4UVVljxa0EQBDIzM3FyciI6OvoHH0hiZEBU976U8uvr6ykoKGD69Ok/6YEnCAIHDx4kKiqKwMDAn1RGR0cHra2tknP4UyDy0CiVSmJjY+nu7kan0+Hp6Wl9kFvxs2B1aKywwgorfgGYzWaampoG3R3TH70pIC7lIS4uPbu7u1+QhuJSYDQaaWtrG7CUdamwWCyXXOeLldHQ0ICfn9+vopllxeULq0NjhRVWWGGFFVb87mF1j62wwgorrLDCit89rA6NFVZYYYUVVljxu4fVobHCCiussMIKK373sDo0VlhhhRVWWGHF7x5Wh8YKK6ywwgorrPjd4w/l0AiCQE9PDxqNBo1GIxEj/dSNXiJVtcgA+VuDxWKR2DDFttfV1Q1QrP2lYDabUavVgzIza7Va9Hr9oOcJgkBXVxc1NTUX1Iv5rUNsg0aj+Z9dz2Aw0NHR8YN92dnZeUHbW2GFFVZcLvhDOTRwXmPlT3/6Ex999BHr16/niSeeGCD29mNQUlLCvffe+5t8EOt0Ot5//30MBgM6nY6vvvqKmpoaPvroowFOxy+B+vp67rzzTtRqdZ/Pu7u7ee6559i/f/+g53V1dbF27Vr27t1LVlbWL16vnwOz2UxPT88PHmexWPjyyy9ZvXr1JZUrOkA/B0eOHOFf//rXRcsXBIH9+/dz8uTJn3wdQRDIzs6mqqrqJ5dhhRVWWPHfxh/KoZHJZBKt/4gRI7jppptoaWnh4MGDCIKATqejsbERg8EgRRXa2towGo2SRktTUxMWi0XSMLG1taWnp2eA2JlGo6GhoUEqq7OzE51OR0NDwwBnwmAw0NTUhFar7VOOxWKhvb0dtVqNxWJBp9Oh1Wqpr6+XBOK0Wi0NDQ3S/11dXTQ2NtLT04OjoyP33HMPNjY2pKenU1NTQ3h4OMuWLZNkFdRqNS0tLX0iTBaLRYpgNTQ0YDQaJUp4QRCk/8XogFgPuVwu1UMUatNqtZKWzWBOnyAInD17ljNnzrBo0SISExPRaDQ0NjZKdhcFDHvbR6x7c3MzZrOZtrY2Ojo6pL9NJlOf+lssFjo7OyVtmJaWFgwGA3q9vo82jNhXbW1tGAwGvv/+e3bv3o3RaMRgMNDY2IhGo+nTblHDJygoCL1eT3d3N62trX3aK/alaOucnBw+++yzPurEg40xsY3i/3q9XrKNqN0ltql3G2pqanjzzTfR6XTMmTOHsWPHYjab0Wg0Eq270WikoaFBGr89PT3U19dLjpBYlslkYuvWrQPGee+2tbW10dbWZtUBs8IKK3412PzaFfhfQ2S4PHfuHCaTCa1Wy6hRo2hpaeHDDz9kzJgxlJaWsnTpUrZs2UJERISkAlxTU0NmZiYjR44kPDyc3NxcvLy86OjokMoXBIGsrCxyc3Px9/fn6NGj3Hvvvfz9739n/PjxnD59miuvvJI5c+ZIWiT/+c9/mDBhAlu3buUf//gHjo6OAGRmZtLU1IRer2fs2LE8/fTTjB49muLiYoYNG8YVV1xBbm4ulZWVqFQqrrzySvbt24e3tzfHjx/n+uuv55lnnmHVqlUUFhZSX1/P6dOneffdd/n44485c+YMer2es2fPMnLkSObMmQOcF8C7//77WbRoEUeOHCEmJkYS4Xz66af59NNPSUhIYM+ePcyaNQtfX18OHDhAVFQULS0tGI1GNm/ejJOTE9999x3PP//8BfvDYrFQUVFBU1MT9fX1kh6Xl5cXx48f54YbbuCJJ54gOTmZ8PBwFi9ejCAInDx5kpaWFo4dO8b48ePJz89HpVJxxx13sG7dOmbNmsWZM2dobm5Gr9czfvx43nzzTeLi4oiOjubrr7/mvffeo7y8HJ1Ox/Tp04Hzgn3btm0jLCwMNzc3Dh48SEBAAElJSaxfv56kpCS+//57rr76alQqFfn5+XR3d6PX6wkJCZH6rbGxkblz50qsrocPH6auro6mpiYmTpwoqfp2dHRI+jWnT5/uM8aCgoLIzc3FaDQybNgwIiIiOHLkCM7OzpLQYm1tLVu3bmX//v288847uLm5AVBUVMTx48e55ppr2Lx5M/7+/iQlJfHEE09w3XXXsXXrVmbOnElTUxMKhYI77riD7du3Y2try+HDh3nxxRext7fHYrFw/Phxzp49S319PWFhYdjY/P9pQxAEtm3bhsViIT8/n8WLFzN8+PCffZ9aYYUVVvxY/KEiNL0hipApFApKS0tRqVSkpKQgk8nIzMykp6eHXbt2YTQaCQ0N5YsvvpCEKFtbW/nss88YP34848aNw9XVVSpXEAQ+/fRTRo8ezcyZM6mpqaG4uBhbW1smTJjA3LlzycnJkY6XyWRMnDgRJycniouL++RgaDQaNm7cSGRkpCRYN3fuXJ566in27t3LZ599RltbG35+fuj1ejZu3EhERASzZs1i+fLlkqCeXC4nNDSUIUOGMHHiRBQKBd3d3WzYsIGUlBRuvfVWRo4cKV1XFN6bOnUqU6ZMkZSHKyoqyMzM5Ny5c4wbN45Zs2axevVqPv/8c+Lj4xk3bhze3t5UVlayf/9+SdSvra3tgv2gUCiIjIzE19eXqKgoPvnkE8aOHcusWbMoKyujsbERQRC4+eabufrqq4HzTtAXX3yBXq8nPDwctVrN1VdfzalTp6ivr2fEiBHs3r2bpqYmvLy8MJlM+Pj4IJfL+fvf/87ixYuJjIyUjh8zZozk6BoMBnbu3InFYmHo0KH4+/sTFRVFUVERDQ0NJCcnM2XKFD788EO++OILkpOTueaaa5g1axYAx48fp6GhgUWLFkmOKcCQIUMYPnw4er2egoICAgMDGTJkiBQxFAShzxhTq9X4+fkxcuRIenp6OHPmDN988w2xsbHMmzePa6+9FplMhr+/PytWrEClUtHY2CiNKX9/f3x9fYmMjJQiOX5+flgsFubNm8fw4cNxdXVl6dKlnD59moyMDE6fPo2trS1ubm5otVrpPvH392fMmDFMnjxZEunrjaioKMLCwlCr1ZSWll7qLWiFFVZY8YviDxehEREeHs7MmTMpKyvjq6++Ijo6mmPHjjFv3jwUCgXu7u488sgjvPnmmzQ1NdHa2kp8fDxRUVE0NTWxZ88eSUW3v66JVqtFrVYjl8vx8vKSNFvkcjkKhaLP8o7RaOTAgQMsWbIEJyenPiH7hIQEdDodTz31FC+++KJUhpOTE+7u7rS3tzN79mwmT55Me3s7b7zxBo2NjSgUCvz8/KTrDgaTySQp4Hp7e0vLLuLxYrsUCgVyuVx66BoMBmmZx93dHYVCQX19fZ+lE3FpatasWcycOROdTid9JwgCFotFKrM/tFot7e3tKBQKyXYKhUJyPuH/L3EkJCQQEhJCa2srbm5uuLi4sGnTJm6++WYOHjxIYmIic+fOpaOjg7a2NqkMhULB8uXLef7551myZIkU1QDw9PTk4Ycf5o033qC9vb1PPzU3NyMIAp6ensjlcurq6mhrayM4OBgPDw8AvLy82LJlCxMnTpQ0fcRlNYvFIvVLb3uIx7S0tEhjrK2tjaysLJqbm/H395eiO83NzSiVSnx8fKiqqkIul0s/vcdVb9v2/7t3v8rlcmkpytHRkfnz5zN9+nRpWVQmk3HmzBlGjBiBRqPB2dl5QP+dOnUKHx8ffH19f5PJ8VZYYcUfA38oh0bMSdBoNBQWFqJQKEhLS2PJkiVSqD84OJjGxkaysrJob2/nz3/+M9XV1UyZMoUnn3ySGTNmEB0dzdy5c3n//feZPn067e3tfWTfb7zxRvbt24efnx9eXl7Exsai1+vRaDS0tbWh1WqxWCwoFAq6uro4evQokZGRqNVqioqK8Pf3RyaTkZGRQUhICFdccQVGoxGj0Uh+fj7l5eUsXLgQd3d3XnrpJXJzcwkJCWHhwoU8+eSTlJeXM3z4cOLj4yXnSmy3RqNBr9djMpkYN24cf/3rX5k4cSLTpk3DxcUFOJ+kK+7Y6ejoQK1W09HRQXd3N+Hh4bi7u3Py5EmqqqpYvnw5Wq2WTz75RGqfnZ0dra2t/Pvf/yYkJIS5c+eiVqtxdXXl9OnTrF27lueeew6VSoXFYqG5uVkq/8Ybb2TPnj14eHjg5+dHaGiolCfi7u6OTCbDxsaGCRMm8PjjjzNlyhTi4uKYOHEiixcvJj09HQ8PD2bMmMGzzz5LaWkpQUFBhISEoNVq0Wq1uLm5ERMTg6+vL0OHDu3zwK+qqqK5uZk///nP1NfX4+Pjw8mTJ7nxxhuxsbEhKyuL/Px8brzxRurq6vjHP/7B7NmzSUhIQK1W4+/vz6hRo3j66ad57rnn8Pb2Bs4/9G1sbDCbzVLUr6SkhHPnzjF06FDkcjlTp07tM8ZOnz5Na2srHh4eVFdXM3/+fF599VXy8vIYMWIEKpWKzs5OOjo6pN+ig+Tk5ERLSwtZWVm0trbS1dVFR0cHPT09aLVaNBqNlJ/V1dVFTEwMq1ev5pVXXsHPz49FixazaWpUAAEAAElEQVRJNuno6CA8PJzy8nJsbW1ZvXo1zz//PA4ODgiCwPHjxwkPD6e+vl6KACmVyv/tzW2FFVb84fGHEqcUky7r6+ulaIebmxteXl6YzWby8/Px9/enqamJwMBA2tvbkclkBAYGIpfLKSwsxMbGhqFDhyIIAoWFhdjZ2aFSqfD395cmcYvFQk1NDV1dXQQFBaFUKqmoqMDV1RWDwYDRaCQkJASFQoEgCJSVlQHn34YdHBykZYj29nba29uxtbXF39+fO+64g+uvv57w8HBCQkKQyWScO3cOo9FIVFQUNjY2VFZW0tnZSXR0NHq9nvr6enx9fdHpdBgMBjw9PSWlW3t7ewoKCvDw8CA4OFh6sOt0Oqqrq/H29kan02EymXB3d6exsZGAgADkcjlVVVW4ublJEYfi4mIA7O3tpSWO0tJSwsLCcHV1paqqChsbG1xdXdm4cSM33HCDlKNRX1+PVqslKCgIe3t7qqur6e7uJjg4GJPJRG1tLV5eXnh5eUl1NBqNFBQUoFKpiIiIQKFQYDAY6OrqwsXFRbKrTqcjOjoajUZDa2srAQEBODk5YTQa+fbbb7n22mulHBY4vyOrrq5O6neDwUBZWRnR0dF0d3dTU1ODu7u7tHxTXFyMXC6XHuhGoxFPT0/q6+vx8/PD3d1dGncVFRUEBgbS0tJCeHg4RUVFhIeH4+joiEwmw2Qy9RljXV1dnDt3TnKyo6KiqKyspKenh6ioKGl8+Pn50dDQgIeHh2Qji8VCYWEhXl5e6HQ6BEHA1dWV5uZm/Pz8pIiVvb29FGXq6OigtraWqKgonJ2dJVt3dnbS2tpKYGAgPT09rF+/nuuuuw5HR0cpsiSOs46ODslBs8IKK6z4X+IP5dD8nqHX67n11lu5++67mTRp0u/2gdHd3Y1Go8Hb2/uCy2H/bWRkZPDee+9x++23M3HixF+tHr9H9PT0oFar8fHxsdrNCius+E3B6tD8TtDd3U1tba0Uwfm9OjS/BYjbqv39/a12tMIKK6y4TGB1aKywwopfDSKXj8lkQqFQYG9vj42NzUWjP2Lelclkws/PT0q0b2pqQiaTXTB6ZDabqampwcnJCU9Pz59cZ6PRiE6nk3KVBtv5NRgMBgPV1dX4+/vj4OAw4Luqqir8/f377I67lPICAgKwt7f/0e2wworLDX/411Nx545ICtbY2HhRFl1BEGhubv7RzMBiQvKPoaAXCfou9ZzeO3EuFeKOIaPR2Odzs9lMXV2dJJ3wY+qs1+v7cPP8r6HT6S5JEuDnwGw2S1vK/5swmUyYTKYfPM5isdDd3f1fk7T4b7XVYrHw/fff86c//Ym0tDQee+yxi27zFwSBrVu3cuzYMcrLy9mxYwdms5nvvvuOgoICtm7dKuWkDXbuu+++S0FBwc+qc1dXF/fddx/r1q27pL4RYTAY+Pe//z3ovWE2m3nhhRdobW295PJ6enr417/+9avea1ZY8VvCH96hAUhPT+fJJ5/EaDTy2WefXXDCER8Y69ato76+/kdfZ9euXZw9e/aSjzcYDDz//PMcOXJE+kwQBLq7uwd1uurr61m3bt2PevCo1Wruvfdeamtr+1xjx44d5OTk8OWXX16STELven311Vd8+umnl1yHHwvRaeq//Vn8LDs7+4IyC78UdDodH3/88QXZj/vXD5CYh/vb02g0DtCAEgSBoqIiXnrpJbKzs3+wPiUlJbz22mv/FadDq9Xy6aefDnB6fwnI5XIaGxtJSEhg+vTpFBUVUV5efsHj29vb+fbbb5k9ezbDhw9n165dZGdnc+LECSZPnoyHhwdpaWkDxoaoe9Xe3k5UVJQUGRJ1rjQaDT09PRiNRokpWdwV2N9pUSgUNDc3M2fOHCmhXGTQVqvVEsN172uITMwODg44OTlJidoienp6sLGxwdHRUWLE7q0PJn5mNpulHYHd3d04OTnh6Og4oDwrrPgj4rLZti3e7Dqdju7ubmxsbPDw8KCrqwuTyUR3dzeenp60tLRI3wmCQFtbG46OjnR1dUmMqSIfjDhxeHp6snPnTgwGA1deeSW33HILdnZ2Ek2+VqvFw8MDW1tbOjs7kcvldHZ24uvr2yccfcUVV0h1FeUM3NzcpGPEibe1tRVHR0dcXFxwd3dHp9NRU1ODp6cnFouFV199lZUrVxIUFIRarcZkMuHl5UVAQAA33ngjcJ6UTyaTodFopHqIdfXx8UGhUKDX6yXivd78ITqdjo0bN/Lggw8yY8YMzGYzTU1NODg44OrqSk9Pj7SjyNvbG7lcjl6vl+rl6+tLY2Mj9fX1ODk54eTkhNlsprm5GXt7e1xdXSXuFbPZjFarlR7yHh4ekqimUqnE1dVV2pEDYGNjQ2NjIxs2bOCee+7ByckJmUxGa2sr//nPf3jggQdITEyUnM+Ojg6p7d7e3jQ3N+Pq6oqDgwMmk4nm5macnZ2lckSqf1HCQKlU4u7uLtVJLpfj6emJk5MTd955J0qlEp1OJ8lFeHt7S6zTDz74IC4uLtKuo6qqKo4fPy4xMotoaWkhPz+fGTNm9Pk8JCQEOzu7H9R8EgSB9PR0du/ezYoVKxgyZIj0MNTpdDg5OWFnZyft2nN0dESj0eDm5kZHRwcODg7Y2tpKZHoGg0HaEeXj44OzszN33HEHCoVCOqajowNvb2+USqU0Zp2cnHBwcEChUFxywrDZbCY1NZUpU6awatUqpkyZQmxs7AWP12q1GAwGbG1tKSkpQafTsX79eon0r6Ghoc+uNVEG4siRI7i4uEj8SXv37kWpVHLixAliY2Opra3F1taWyZMnc+jQIaZOncr27dvx8PBg9uzZ+Pn5SWWWl5djZ2dHQECAdI3q6moOHDhAdHQ0qamp3H///ezYsQMPDw/OnDnDlClT6OzsJDw8nEOHDmFvb8+0adOke7+oqAgbGxv27dvHrl27eP311ykrK6O2tlbaZn/DDTewdetWIiIiSEtLY+zYsURERHDw4EEcHR2ZOnXqJS9/WWHF5YjLxqEB2LNnD59//jkpKSns3buXZ599lo8//lgigIuNjcXZ2Zn169fz4IMPUlVVJe266enpobm5mYceeogXXnhBeiA3NTVha2tLfn4+Hh4eNDY28swzz/DnP/8ZOzs70tPTiYqKYvv27dx777089thjTJ8+nZMnT3LddddJlPoGg4GXX36Z+Ph4YmNjycrKwsnJiVGjRhEcHAycn6w/+ugjJk6cyJYtW1iyZAmCIJCamsqJEyfQ6XT89a9/5dChQ8ydO5dz585RW1tLW1sbY8eOxWKxsGbNGl599VXuv/9+xo8fz9mzZ5k7dy6RkZHk5+dTUVGBSqXiqquuYtOmTSQmJlJXV9fHjmq1mqamJmpqaggICODLL79k4sSJbN26lQULFnDo0CHa29tRqVT8/e9/x9HRkY6ODg4dOsScOXOk6MLmzZs5ffo0b7zxBjt27MDV1ZV169bxr3/9i6CgIACys7N58cUXWbRoETt37uStt94iPT0dBwcHvvnmGx544AHS09OpqKggNDSUqVOnUlxcLPEEOTk5AeejU2lpaVx33XWcOHGCpqYmVqxYwX333ce1117LwYMHiYuLw87OjrKyMp5++mm2bt2Km5sb3377LS+//LJEjrdjxw6+/vprJkyYwP79+3n99dc5dOgQ4eHh5OXl4e7uTkJCAk8//TSrVq3i5ZdfxsXFhe7ubnx9fRk9ejQnTpygpaVF4vZpa2vjm2++wdbWlvr6eqnPRfQnpJPJZKhUqkt6QLW2tuLg4MC8efPYsGEDf/nLX+jp6eGTTz5hwoQJtLa2YmdnJy0FTp48mccee4znnnuOzZs34+LiQlxcHM8//zzLli1jw4YNLFy4kPLycnx8fJg8eTLPPPMM77zzDi+88AJeXl50dnYSHBzMrbfeynvvvUd8fDzvvPMOd955JzNnzrzke7apqYm2tjZmz57N9u3b0Wg0fRyS/vDz8yMuLo4PP/yQgwcPMn/+fHbt2sXdd9+NIAjk5uZy8803Sw5Vd3c3r7/+Og8++CC5ubkkJyeTm5vL8ePHeeyxx9i9ezdxcXEkJCTwz3/+kyuuuILExEQEQSAtLY3XXnutT76N+PnYsWOxtbWVIi/PP/88Dz74IN7e3rz22mvs37+frKws/vnPf3L69Gny8vKoq6ujtbWV0NBQYmJiSEtLo62tjYkTJ5KWlkZcXBxXX30127Zto6ysTOJrWrNmDcOGDePDDz8kPj6exMREwsPD+fzzz2lvb2fOnDnExsZad51Z8YfHZbPkJJPJCAoKwt3dnbvuuouUlBRSU1NxcXEhMTGRO++8UxLYi46Opr6+nq+//popU6Ywbtw47Ozs8PT0RKlUYjQaWbduHRMnTmTZsmXMmDEDf39/IiIiCAoKknhM1qxZw/Dhw5k8eTJGo5EzZ85ga2tLSkoKM2bM4MyZM1L97Ozs8PHxobu7m56eHjZt2iRx4Ig4fvw4arWaxMREJk2axKeffoogCEyZMoV//etfVFZW0tbWhqenJ9HR0URERBATEyPR6YeEhKDX63F0dMTOzo4JEyYwf/58Tp8+zbfffotarSYgIACz2cyGDRsICQkhISFBci5E+Pn54enpyYgRI8jLy6OlpYXExESmTJnCmjVr8PT0JDIykqefflpKbnRzc8PT05Phw4cjl8uJiorilltuoaWlhXPnzrFr1y4MBoPEnyL2mZ+fH0qlkqVLl+Lg4EBxcTGbNm2ip6eH4cOHo9FouO666yguLsbV1ZWIiAgCAgLw9fXtw53j7e2Nl5cXUVFREomel5cXcrmcadOmSduzly5dSllZGXl5eaSmpmI2mwkLC5OWGcVx5OnpyT333ENycjJff/01e/fuZcKECVx99dWsXbsWNzc3jEajFEWKjY3l+uuv5/Tp0/j4+ODt7U1YWJhUP09PTxQKBddddx3BwcFYLBZOnjzJvn37SE1NJTs7m3379nHw4EE6OzsvedyLDq/ZbCYkJITvvvtOiv6dPHmSvLw8Ro0ahVarZf369URFReHh4SEto7i6uqLX66Xk2iuuuIKgoCCCg4NZtGgROTk5BAUFYTAYUCqVkvNz7bXXcubMGTo7Ozl69Chjx47F39+f6OjoS36wCoJATk4OYWFhREZGEh0dTV5eHmazWRJ17Q+lUsnjjz9OeHg4gYGBXHPNNbi7u+Pi4kJNTQ0Ao0aNko7Py8uTkod37dpFYmIi3333HVOmTKGmpkbiW/L29sbGxobTp08TGhqKl5cXiYmJ7Ny5s492lcViITU1lZSUFOC8w7R//346OzuJiIjgzJkzjBo1ioMHDzJ9+nRMJhPFxcUkJCRQVFQk8SpZLBZUKpUUwSssLOTKK6+koKCAIUOGcPbsWaKjozGZTKSmphIWFiZFk0T+rOLiYgwGA/X19ZjNZuuSkxV/eFw2Do0IkZLdy8tLIixTKpWS2nJKSgr3338/oaGhkrJ2f1gsFonW3tHRUdp1IE4Yvd/+WlpapOuJnw8mcdAboaGh3H333bz++uucOHFC+rx38qWHh4e020OhUGBra0tAQIAkdgiQk5MjkbX1r5tIHChS27e1tREZGcnSpUu5/vrraWxslHZqXOwBJC439a+TKCPQ/1yxHuJ3MpmMnp4eenp6mDFjBg899BABAQF9Jl+xz8RjtVotU6ZM4b777pNIDEeNGsXOnTv7LL+IeQb9cTGqf3H5R8xxmT17Nn/961/7yB/0rr+Xlxf29va0trZKCubiuOoN0da9Vap7/xYlI3x9fdFqtchkMsLCwoiJiSEyMpIhQ4YQExNDdHQ0KpVq0L4Ql+h6t1kkAoyPj5eifXv27JEe/KdOneK9994jKSmJq666iqeeeoqKiooB9hPt1F8WoX/+SO9xZbFYcHV1ZcqUKezYsYM5c+b0WZr5IfT09LB3716cnZ3Zt28fBw4c4OGHH6awsPCC+UBarZbU1FRqa2v517/+hYuLC7Nnz+bkyZPs2bOHBx98sI/9FAoFHR0dHDx4kObmZnQ6HXZ2dmRmZpKdnS0txSkUCsLCwqQlvm+//Zbo6Gh8fX372L+srExaCtq7dy/PP/88NjY2BAUFUVZWRlVVFXfffTdxcXF0dHSQlpbG4sWLcXNzw93dnXvuuYctW7ZQVFTE2LFjmTZtmuQwtba2kp2dzV/+8hccHBw4d+4caWlpyOVyNBoNkZGRHD16lBMnTlBXV4eXlxf33HMPmzZtoqio6JLtboUVlysuqyUngLq6OvLy8mhoaGDZsmXk5OTQ1NSEi4sLQUFBPPLII4wYMYLp06czfvx43nnnHUJDQ2lvb6eurk5ar543bx6PPPII06ZNIyUlRaLAHzNmjKQNtHz5cjZs2EB0dDR2dnYkJCSwdu1a6XsxN0ShUGAymWhvb5fC4gA33HBDn0k7OTmZbdu2cerUKfLz81m+fDnZ2dlUVVWRlZUlrZmrVCoOHz5MZmYmgiBIuTtxcXGSFpK4vNDW1oZOp2PmzJm88MILzJ49m8jISBYuXMgbb7yBSqWiubmZ6upqKaIgJjI2NTUxduxYNm3aRGZmJgUFBSxfvpzU1FSJ2bZ3botYr5aWFkkyQafT4ezsjFwu54knniAyMpIFCxZI+SoajYbu7m46OjrQarXY2Njg7+/PI488QlxcHFOnTiUjI4Prr7+eVatW8eWXXzJlyhTq6+vJzc2VRDVVKhUGg4G0tDQaGxvRaDSo1WppSVGsj/hZQEAATU1N/POf/yQkJEQSvhRRXV1NXl4ebW1t3HzzzWi1Wvbu3YuLiwtLlizBYDCg1+tpa2ujs7NTYiLW6/UolUo0Gg2ZmZmMHTtWctTc3NwoLS0lNDQUJycnKTonk8loaWmRcjLg/MOzoaGBxsZGysrKiI+Pp729nS+//JJHH31Ucpj37NmDSqUiOjoagBkzZvDZZ58xduxY0tPTuf3229m3bx/Hjh0jNDSUefPmIQgC/v7+fPvtt7S1tWFra0tbWxs9PT1SAqqYL9TV1UVjYyN6vZ729nZJYsHBwQG9Xo9Wq6WoqAidToe3tzdNTU0EBARcUpTGxsaGxx57DEEQkMvlTJ48WerHP/3pTxcsIyEhgalTp0ocQldffTUtLS1MnjwZe3v7PueNHDmS559/HldXV+Lj4/Hw8CAhIYHW1la8vb2ZOHEiHh4e6PV6kpOTSU5OBmDRokWSblfv8vz9/dm4caPk/MXHx+Pp6UlycjL/j73zDo+qTN//Z9J774WEBEIICTVAAOmCICJWVBSx7vp1dV1XF3tZ29or6uqKDUVsiCK91wBJIL333iaZmulzfn+w5/1lUhTL7grmvq65CDPnvP2c85znfZ77NpvNrFixAhcXF6699lo6Ozvx8vLC19cXm83Gww8/jI+PD48//riQ8QDw8/PjmWeewWq1ComNZcuWMXXqVMLCwpg4caLwPra3txMaGoqbmxsPPfQQPj4+PPbYYw7lDWEIv1ecUzw0+fn5vPrqq9x3332Eh4fj7e1NY2Mjrq6uREZGYjKZqKioIDw8nPDwcMxmM+Xl5fj5+eHu7o6Pjw9tbW34+/sTEBBAZWUlnp6exMbGYjQaqa2tJTY2lo6ODry8vAgLC6O9vR2NRkNUVBSurq40NjaKwFmbzUZMTIx4021qahLuYqVSKR7evYOCdTodzc3NgsbebDZTV1eHi4sLsbGxuLi40NzcLN6OZV6Lzs5OAgICRGCqUqnE19cXq9WK2WwmKiqK+vp6rFYrCQkJODs7ixgi+eHq7u4uDJrW1lZ8fX0JCwtDr9fT1NREYGAgQUFBNDc3I0kS0dHRwqCRJInm5mZsNhsuLi6YzWYCAwNFYKlCoRC6SvJDQg7alR8unZ2d+Pv74+npKeapt4xAd3c3RqORyMhIampqiIiIEC57SZKoqanBy8tLbFkEBgYKLSQ5cNfPz4/Ozk4iIiKwWCzU1dURHx8vApXhtO7SO++8wz333CPqsFgs1NfXi2BQnU5HR0cHwcHBaDQaXF1dcXV1RavVEh0dTX19PUFBQUK5XB4fLy8vAgICHB4+nZ2dVFdXM2XKFPGdJEl0d3ejVquFyKkc9CrLXsieRFldG6C1tRW9Xk9YWBgajQabzUZYWBhmsxmVSoW7uzsRERFoNBoaGxsdAte7urrE2vHw8MDFxQWtVouPjw8ajUb01c3NTfxmt9s5cuQII0aMQK1WYzAYuOyyy86qh+uuXbtQq9VceOGF/QyiIQxhCGcPzhmDRpIk9u3bxwcffMCbb74pgjGHMISfAkmS2LlzJ5999hlr1qwRQcf/6TplsdKzDSdOnOC9995jypQpBAQEMHnyZIYNG3ZWGQXytlrvWJkhDGEIZx/OKYOms7NTvJ32ZeIcwhDOBDJxYk9PD+Hh4UMMrD8C+brr6ekhNDR0yMMxhCEM4X+Gc8agGcIQhjCEIQxhCL9fnJM+VrvdLjhKfojT4teCHPAbEBCAq6vrr1q2TPTW3d0t0pB/zbINBgMtLS0/qgfT09ODwWD4RRo4PwUmk0mQt/033/glSUKv12Oz2fD39/+v1du3DQaDAaPROGCwpyRJtLe3ExAQcEbr22g00tra6hDIbbfbCQ4OHtqaHcIQhnDO4JxK25YfRgaDgXvvvZfy8vKfXIZMfz4QLBYLJpOpX52dnZ386U9/+llyCGeCyspK/vznP/9k/agfg8lk4sMPPyQ7O5vjx4/3+12m8Lfb7ZSWlrJz585ftf4fQlNTE19//fWAv8nzfCaQjYMzkW+A08bw+++/z2effXbGbe197pnobsntH8w5KusVvfbaa4P+/vXXX5/RepMkiWPHjnHq1ClaWlq4++67aWtrIzc3l8zMzB89fwhDGMIQzhacMwaNJEmUlpby6quvYrfb8fT0RK/XU1tbK0i6dDoddXV1g4r4mc1mXnnlFUHupdVqqa+vF+nXH374IQcPHsRisaBWq2lsbMRsNhMaGoq7u/uAvDOynEFXVxdtbW0olUqsViutra0YjUaUSiXd3d1YrVbxd1NTEyaTSbQ1PDwcq9VKZ2enIOWS1YXljCeTyYRKpaKpqWnAh7fMaNrY2IhSqRRGSlZWFvPmzRNEYb3R3d3N008/TWtrK2PGjGHRokVCzFKj0VBXV4fZbKaxsVGIQZpMJurr69FoNAOOsWxgqFQq6urqxHG92yZJErGxsSxfvtwh20fO0srOzubdd99Fr9ej1+tFZs9A0Ov1PPPMM9TX14s08fr6+kENCmdnZ8LDw+np6enXvvr6+kFFLyVJYs+ePWzYsEGQJ/YeB51OJ9Kfs7Oz+de//kVPTw8Wi4WGhgaR0i9zEfn4+GA0Ggfsk0Kh4NprryUiIkJISnR2dg66/ry9vVm2bBne3t5ER0eTnp7Oeeedx8iRIx2O6+zsHNCYlw2w5uZmenp6aG5uFhpI3d3d/dbiEIYwhCH8L3BObTmVl5dTXFyMVqvFZrMJTpJRo0axZMkS9u7dC8DJkyd55pln+m0PmUwmDh06xPjx4/Hy8mLHjh1MnTpVUKcfOXIENzc3GhoaWL9+PZMnT6asrIw//elPP9iuzMxM3nzzTS6++GK2b9/O66+/zvPPP89VV12FSqVi//793HDDDaxevZobbriBjRs3MnfuXMHR8oc//IHu7m42bdrEvn37+OMf/4iHhwcGg4G9e/cyc+ZM6uvrKSoqIiQkhL/85S8EBgY6tMFgMPDPf/6TWbNm8f333zN79mxMJhPd3d10dnYOuJXU3d3NsWPHWL58OQcPHqSwsJC77rqL2267jSuuuILMzExiY2MJCwsjJyeHp556is2bNxMcHMyXX37Jyy+/3G/bxmaz8cYbb9DY2Eh0dDR5eXm8/PLLrF+/XrRt3rx5WCwWvvzyS5599lluv/125s+fT05ODpdccgnNzc2Ul5fT3d3Nxo0bmTNnDidPnmTx4sX9tme0Wi1Hjhxh6dKlmM1mdu/ezaRJk3j55Zd54IEHHIjTeuPUqVMYDAYKCgp49dVX2bt3LyEhIWzYsIEXX3xRyCTIkCSJkydPotVq6erqYt++fWIcnn76ab744gtmzZpFXl4ezc3NlJWVoVarOXjwIKGhoYLmfv/+/URFRVFWVjaocaDValm9ejU333wzBoMBg8GATqfj/PPP70cQqFAomDBhAgqFgpycHPF3cHCwYDt2cXHBZrPx6quvcssttxAfH9+vzubmZh544AHWrFnDAw88wJ133om7uzsNDQ14e3sLxty5c+eyaNGiocDgIQxhCP91nDMeGplGPzw8XHC7zJ07l8svv5yCggJ2794tbr4yB01feHl5ERISwqhRo9i0aROxsbFMnjyZiIgIdu/eTVhYGCNHjiQsLIz58+djsVjIy8v7QcpxmSPE29ub5cuXA6fZTmWJBZnUKywsDBcXFy644AJGjx5NSEgIl112GSUlJdhsNgICArjpppu4/fbb+eSTT/j0008xGAwkJydjsViIiooiNjaWxx57rN9DDU5z9DQ2NjJp0iQWLlzIRx99JCjeR40aNWBsjsyjkpSURHx8PAaDQcRtzJw5k7lz52K327niiitobW0lKyuLnJwc7HY7SUlJA3pNXFxcCA8PZ/To0dxzzz24ubmxYcMGmpubRdvef/99hg0bJjhyPDw8mDlzJueffz5FRUVERkYSFRVFaGgoBQUFHD16lEmTJg04/v7+/qIPX331FfHx8UydOpWgoCB27Ngx6LxNnDiR+++/Hx8fH9avX8+xY8ew2WwkJycLgcbecHJyIiIigri4OFpaWsjKykKSJEaOHIler+fkyZPCoJDbbzQa2bdvH2azmdTUVOrq6tizZw8zZsxg8uTJg8ZL+fr64u/vL0RX169fT3x8vGC07g2Z1VmSJHJycpg8eTIKhQKDwcCuXbvIzMykrKwMZ2dnHnnkEYYNGzZgGcOHDyc0NJTW1lZhwLa0tDBhwgQ+++wzDAYDo0eP7rclO4QhDGEI/y2cUx4amcBMRm+Kdp1OR2BgIEuXLkWtVv/oG6TNZqO1tRVAUP7LaGpqYv/+/SxZsuSMg3Tl4+Q2yv/Kn77U8/LfMtW93JeQkBACAgJoa2tjwoQJxMbG0t7eTmZmppAlGKhvkiTR1tYmgl0Ho9fvO54/9JvcJrkvPT09KBQKLrjgAubMmTNozE/vc4ODg/Hw8KC1tVUQ3/Vtm9ynvqrgzs7O3Hfffbz//vu8/PLLPPPMMz/I5WK1WmlrawNOz+kPHSvT/0dGRgpJg4ULFzJnzpxBt4Lg/2+pOTk5sXDhQmbPno1er+fBBx/kX//6F/X19UybNk1sY5nNZubPn8+8efMoLS2lvb29n9zADyE9PR2bzcbf//53XnjhBcEY3Bd6vZ62tjYSEhIAKCgowGq1kpKSIsQP5bkcbDwuvvhi3n33XWbNmsXmzZsJCAggKCiI7u5uh7X4a0Lm6AHH61u+nux2u8O1M9j58vr5T0DeKpTX9P8Cve8lvwaf0Q+VJ/e3r/SJ3W53mJ++v9ntdof7Rd8Xwd73xt7z+d+Yw5+D3muzd7+sVqvDPXywcwcaw1/alt/aGP23cU71PCAggKamJrKzs9Hr9XR2dqJUKtHpdEyfPp2vvvqKZ599lu3btw940cvaPXv37mXRokUUFxdTVFSEXq9n8eLFREREiK2XkydPClmFyspK9Hq9iIPoDUmSBN2+SqUSkgQjRozg22+/5cSJE2i1WqErpVarBY1+d3c3JpMJq9WKl5cXRUVF5OXlccsttzB79mxWr17NmjVrqK2tpaurC6VSOagRMXbsWMLDw8nMzCQ3N5eVK1fS0dGBRqMZNN5F1o06ePAgra2tQk5AjteRP3I74+LiqKmp4bHHHmP9+vWDzpP0bzXunJwcoqOjWb58OaGhoWRmZpKXl8eqVauEdER3d7eIuZFlBoKCgqioqCA3N5cDBw5w3XXXER4ePuCNwcXFBS8vL/bv388ll1xCXl4excXFmM1mFi5cOGD7goKCaG9vJz8/n8jISK6++mpaW1t5+OGH+eSTTwbdCgoPDyc3Nxdvb29qa2t59NFHWb9+PXq9niNHjnD99dcTHBxMcHAwFRUVGI1G7HY7DzzwAB988AFBQUEkJiby5ptvCrXu3tpVMmSJgs7OTo4cOUJkZKSQkxgIbW1t7NmzR/xtt9tRqVQ4OTnh4uKCWq3GYrHw0EMPcfDgwUH1saZPny7GbdSoUSQlJQkxVnkt1tfXDzrvPwc2m41169bx0EMP0dbWxptvvslTTz2FRqOhq6uL5557jpqamkHP1+l0PPzww1RXV/+q7eqNtrY27r33XvEC9FPQ+6H4S46zWCy89dZbv1rgvsViYc2aNezevbtfO/Ly8rj//vv7xVs1Nzdzzz33sG7dOgdPnUaj4c0332TDhg0899xzVFZW8sYbb/Dee++xevVqXnzxRV555RW++OILXnzxRV599VUHLbR169axevVq8TLy30ZvQ02GJEmUlJSwadMmPvvsMwoLC7HZbGzevJndu3fz2WefDfriY7fb2bx5M++8886v1kaVSsX9998vBFr7tvV3E9smnUOwWq1SRUWFpFQqpdraWqmjo0Nqb2+X6urqJLPZLDU3N0vFxcWSwWCQ7HZ7v/PtdrukVCqlmpoayWazSUqlUqqoqJA0Go1kt9sljUYjVVRUSEajUSotLZW6urqkiooKqaurS6qpqZGUSmW/cu12u9TV1SXV1tZKGo1Gqqmpkbq7uyWj0SiVlJRISqVS6uzslNRqtVRTUyOp1WqpoaFBam5uFuUaDAapo6NDKi0tFXWYzWaptLRUqq+vl6xWq9Tc3Cz6ORDsdrvU09MjVVRUSO3t7ZLNZpPa29ulmpoaSafTDToeDQ0NUktLi9Ta2irV1taKdnZ1dUktLS1SfX29+E6r1UpdXV1SQUGBGLOB8NFHH0nPP/+8VFlZKeZCr9eLttntdqmzs1OMR21traRUKqWWlhapsbFRMplMUllZmaTVaqWmpiaprq5O0uv1g/ZBPk8ut6KiQtJqtYO2z2KxSDU1NVJlZaVkNBolu90udXd3SwUFBZJarR70PJPJJJWWlkpGo9FhHKxWq2hnT0+PZLFYpLKyMkmv10sajUYqKCiQurq6xDgUFhZKTU1NUmdn54B1mUwmqba2Vmpra5NUKpVUV1cntba2DtoulUol1dfXi3PsdrtUUFAgffHFF1JdXZ303nvvSXa7XTpy5Ih07NixQcuRx0G+FiwWiyRJksNatNlsA577c2G326WysjJp9uzZUldXl5SZmSktWLBA0uv1kk6nk9577z3JarVKFotFMpvNA15/f/jDH6S8vDzJZDKJ361Wq/i/fL7dbpcsFotktVolu90umUymQftjs9nE+RaLRVq5cqVUXV0trk25DLkc+f9923b48GEpPz9fstvtDmXK7ZL7lJeXJx09elSy2Wz92ib/f82aNdJHH3004BjabDbJbDZLNpvNoVy5nt5jI5f3+uuvS5988km/Y9rb26UrrrhCMhqN/cbkyiuvlLZv3+7Q123btkmrV6+WrFartH//fikzM1PaunWrZDKZpFtuuUX6/PPPpa6uLmn//v3SW2+9JaWkpEiVlZWSJEmSUqmUli9fLl133XUDzoXcVnn+eo+bPFZ9j7darf3GwGw2O5Qhz6HVapU2btworhsZZrNZuvnmm6Xa2lqprq5Ouu6666T8/HzprrvukqxWq/TOO+9IX3zxxYD1m81mad++fdLq1av7jW3vtsv1911TA61Fg8Eg3XjjjVJZWVm/NXfo0CGpoKBgwLk+13BObTk5OzszYsQIgH5Bm3BaWE7WvBkICoVCxI3IZfQux9fXF19fXwBGjRoFIIJv+wbh9i4zMDBQ/C6fD/TbHpA5QXpzg8jneXh4CDFDAFdXV9EGuW8/BIVCgaenpxgfgNDQUEJDQ3/wnJiYmH7fD8RdMlCbB4LNZhOepJiYGMGj4uXl5dA22ZMxWH1JSUkAPypNIMdWDVTuYHBxcekXGBsQEDBgbFJvuLm5iTlxd3d3GIfewpO92w+Qmpoq/vby8mLMmDE/Wk9cXJz4/4/x5fj7+/c7Jjk5GbPZTFtbG5deeikAiYmJ/cQYe0OhUIgx6L2O+67FXxMKhYKEhAQiIyM5cuQICQkJNDQ0cPLkSby9vRkzZgyVlZWUlpaSk5PDddddx8iRIx36IEkSe/fu5YMPPuDiiy9mwoQJfPXVVwQGBuLu7k5AQAAbN27kqaeeYteuXSQmJtLR0SGEQe+++24HjiatVsvnn38urselS5cKN/+xY8cwmUy0trZy6aWXsmXLFry8vCgsLOTKK690kIUwGo2sWbOGadOmER4eTlZWFo2Njbi5uXHZZZexa9cuvLy8iI2N5V//+hfBwcEkJCSQm5tLa2srarWa2267jV27duHs7Exubi4zZ850GD+r1cqHH34ogsclSSI+Pp5jx47xwAMP0NPTw9GjR3FxccFoNHL11Vezc+dOXFxcyMvLY+7cubS3t3Ps2DFKSkrIyMggNTV10DXSextaRmxsLLt372b06NFcdtlluLq64ubmJo6T19V5551HW1sbc+bM4ZNPPuGRRx7hxIkTpKenU1RU1K9Ou93Otm3bUCgUlJaWsnjxYiorKzl16hTe3t5IksTdd9/toJVXWVnJmjVrSEpKIicnh7/97W8olUq6urqoqqri2muvpba2Fr1eT3t7O9OnT+e1117DxcWFRYsWiUQSOTtR1uZra2tj586dhIeH4+TkxIgRI9i4cSNXXHGFwzo8fvw4bW1tFBcXY7fbUavVfP311wQEBODh4UF4eDgff/wxTz75JHl5ebi5uWEymXBycqKlpYUrrrjCYRz0ej3ffvstISEhIpNz8+bNeHh4UFhYyA033MCaNWuYPn06QUFBbN26VYRQLF269JwL3j+ntpyG8NuHQqHgmmuu4cYbb/xd7/X+r+Hi4sKECRNIT08X5H3h4eG/ST0jZ2dnLrnkEr744gtOnjzJ9ddfzxdffEFhYSGjRo3i3XffxdnZGTc3Nw4ePNjvfIVCQUZGBpdccgnbtm0TVAgRERF8//33jBs3jqamJpRKpYiF+e677wgMDKS8vJyGhgaH8kwmE6GhoURFRbF582YHd/7Bgwepra1l8uTJFBcXc/ToURYuXEhycrLDVgqcfkkZNmwYkyZNIjs7m8LCQsLDwzlw4AC1tbVs2bKF4cOHExkZSXx8POPGjaOuro5Dhw4REhJCVlYW2dnZHDx4kIULF5KWljbg2MljsGrVKo4ePcrixYvx9/enoKCAN998kzFjxrBs2TK2bNnC1q1bOXz4MAsXLhSG9rp169BqtQQHB7Nz586fvH0xevRo1qxZw5dffskf/vAHdDpdvy3/3rGDK1euZOfOndTV1dHS0kJiYuKA5dbV1fH9999zwQUXMHXqVF544QXc3d1RqVTceOONZGZm9tuyDQwMpKamhhUrVrBw4ULee+89enp6SEpKoqmpieLiYg4dOkRVVRXp6emEhIQQERHB5MmTHa4NOVHh008/5ejRo5jNZkwmk3hB8/DwQKvVOsy3Tqfjo48+YsGCBWRkZKBQKMRaDAsL4/vvvyclJQWlUklLSwsdHR0kJyezZ88eWlpa+iU+SJLE9u3bsdvtzJ8/n+joaOC0EZuUlMSpU6fo7u4mNjaW9PR0XFxcCAsLE+teGmBr+WzH0BNlCP9VyNlAsjr5EP53+KFg2t8SFAoFs2fPpqCggMbGRlasWMHhw4dRq9W4ubnR0tLClClTuOeee0QmYW84OTnh7e2Np6cnFosFZ2dnWlpasNls2O12vL29mT9/Pp988gnBwcGoVCo8PT2ZNWsWb775JjExMf2SDVpbW7FYLP3iK1asWMGpU6d46aWXqK+vR6vVAjB8+HC6uroGfYjU1tYybNgwLrjgAt544w0SExO54IIL+Nvf/kZxcbE4rrGxkdDQUObNm8fbb7+NRqPBZDKhUChEzFvfsXNycsLDwwM3Nzfc3d2Fh0TmWrLb7bi5uREREUFeXh5ms1mUJ0kS1dXVjB07luuuu47777//jNeL9G8OKVlJXiar/CHCTICEhATGjx/P008/TWpq6qBGdnd3N11dXcBpL5A81p6enri5ufVLEpGNJnd3dzw9PYWX0snJiZqaGtzd3bHZbFx99dUUFRXx/PPPOxhEvfvt5OTEvffey9KlS1EqlUydOpW0tDTRBo1G00+kVaVS0dbWhouLC66urmJuZP4mu92Oh4cHl19+OWvXrsXJyQl/f39uuOEG9u7dy5tvvtkvRrKkpEQYUa6uriLYuKmpCTc3NwdOMnndyuteHvNzybAZMmj+DUmS0Gq1dHZ2/qIJli9itVo96DEWi0UsrJ9Srl6vp7y8vF/asEyuJrMkt7a2/qQ+SJJEU1PToAzJQ/hlkP5NrvhT5+Wnwm6309zcTG1t7TkXBBgWFsbMmTOZNm0aMTExTJ8+nUmTJuHh4UFwcDDvvPMOhw4doqCgwOE8u90utlvkz/Hjxzl58qQIOtdqtVx00UUcP36clJQU4uPjycrKYsuWLezevZuuri6HB1Nubi7Hjx9Hr9eLYHVZqiI3N5dHHnkEHx8fYmNjUavVNDQ0UFVVxYUXXtjPK+nh4UF1dTXDhw/n008/5eDBg+zevVt8d88991BQUICXlxc1NTWEh4ezZcsW9uzZw65du4iMjKSiooLs7Gzq6+tFJqMMSZIwGo3Cg2AwGDCbzRiNRqxWK+effz779u2jq6sLb29vLr74YhGw39DQQFtbm/CwHD16lCNHjmA0GjEajVgsFof1bLFYBNFlU1MTJ0+e5NNPP2Xv3r3s2LEDu93OsGHDxPa4xWJBp9Oh0Wiw2+3iQazRaLjpppvQarWkpqaiVqvR6XT9HuaJiYk4OztTUVFBdXU1CxcuxG63YzabRR8Huqf19PTQ2dlJUVERF1xwAR9//DFmsxmNRkNzczM5OTk89NBDBAUFoVarcXJyorKy0iHQWZIkYZQUFxdz1113kZGRgUajQa1WU1RUxMUXX+ywboKCgnBychLz29HRweHDhzl16pTDWlywYAGVlZWC+LKgoICnn35aEHT2xqRJk9i8ebPIkKytreWjjz7CZrOh0WhoaWnB09OTqqoq9u/fT3Z2NjqdDpVKRUdHBx999NEZsZufLRgSp/w3bDYb69evp7Kykscff/xnv7VaLBaefvppUlJS+r0t2mw2TCYTnZ2d/PWvf+X9998/Yy0dk8nEu+++S0xMDP7+/sybN0/8ptPp+Otf/8qqVatwcnLirbfe4v333z9jD4jNZmPNmjVcffXVgxLNnQ2QL3gfH5/fjNfBYrFgtVqpqKjg2Wef5aOPPvqPeKYkSSIzM5O2tjZqa2v5wx/+MCAvzdkKSZJoaGggJCQET09PGhoaCA0NxcPDA6VSyfbt2wkPD2fWrFkO+lY9PT3s37+fuLg4bDYbzc3NTJw4kT179pCSkkJlZSVz584lICCAsrIyRo0ahUKhoKCggJycHKZOncro0aMd1pNWq2XLli0kJydTVVXFuHHjKCkpISkpSTxEPTw8SElJob6+nqqqKoKCgkhNTe0399XV1eTn5zN//nyys7Npampi7ty5eHt7U1hYiLu7OyNHjsRgMHDkyBHOP/98SktLKSsrY+bMmQwfPpzc3FxKS0uJjIwkNDSU0aNHC8PJarUKWZPY2FhOnTrFpEmTKC8vx9/fnzFjxpCVleUQq9S7vLCwMOLi4tixYwcKhYL58+fT1dVFXl4e06ZNc9Baa2tr4+jRo7i6ugqm9uHDh+Pp6UljYyM6nQ4/Pz+mTZuGq6uroJvw9fUlIyMDi8XCkSNHiImJITk5mfb2dkJCQjh27BgqlYoZM2Y4xPxJkkRzczMlJSUEBASQmppKWVmZ2KrJyckhIyPD4Z7W3d3NypUrBfnouHHjKCwspKamRhigISEhmM1mMYc5OTno9XpmzpwpvEWy56qlpYW0tDRxH6+urqaxsRFfX1/Gjx/vYMBKkkRdXR2HDx8mNjYWDw8PRowYwc6dOxkzZgwVFRXMmzcPX19fPv74Y6699lrc3Nw4deoUdrsdLy+vfmvRYrGwb98+enp68PX1JTk5mcrKSgwGA15eXnh6ehIcHEx+fj7Tpk1j7969Yt3PmDGDkydPMm/evDOi8TgbcE4ZNNK/U6RVKhXR0dEoFApUKpUgEgsJCaG5uZmoqCiRrirLIkRGRnL48GG2bdvGM888g9FopL29HW9vbwIDA2lvbxc8MJ2dnQQEBAiumtDQUHx8fDCbzbS3t7NhwwaioqK49tprRdvsdjubNm1Cr9dz0UUXccstt/Diiy/i5ORETEwMCoWCjo4OTCYT0dHRDnvMkiSRn5/Piy++yAsvvEBoaChdXV309PQQHh6Ou7s7Dz30EIsWLWLEiBHcc889fPzxx8Dpm4yTkxPh4eG0t7cLt7TRaCQsLIzu7m78/PzQ6/X4+Phgs9kwGo3i5uPv7y/66eLiIoj1enNEmEwm9Hq9cFX7+PjQ3t5ObGwsLi4uQiohJiYGFxcXlEolBoOBiIgInJ2dRexCQEAAXV1dBAUFodVqcXZ2xt/fH5VKhcViwdvbG4VCQXt7OxEREXh6ejq0o7q6mg8++IB7770XX19fwekiEy32DRTV6XQolUoCAgLw9vbGbrfT0dGBh4cHAQEBtLe3i7Y4OTkRFBSEyWQS0gS+vr64uLiIt+KoqCiHG5jNZuODDz4gMjKScePGcc899/Dyyy8jSRLR0dFIkvSDc67T6bBarXR1dREWFibkEFpbWwkKCsLHxweVSoXVakWSJF555RWmT5/O+eefj6urq3Bvh4eHi7fhnp4egoKChL6VzEvU2tpKTEwMbm5uYm1FREQgSRJqtVpsrYSHhwuZBLPZTEREBE5OTrS3twuCx1+DB2UIQ/i1If2bi+uWW25h3bp1BAQE/GZefGRkZmayfft2IUfzW2vfbx2/vQjAnwlJkqioqKCgoID29nZUKhUXXnghjzzyCDfffDMbNmxg8eLFtLW14erqyuTJk/nHP/7BpZdeynfffcdDDz0kFo9Wq+Wdd95hwYIFvPvuuyxevJi8vDy6u7u577772LhxIxdddBH79+8nICCAzZs389xzz7Fu3TrOO+88qqur+2W1SJIkou/NZjN6vZ7Dhw+ze/du/vSnP2E0GjEYDBw4cIDJkydzySWXODysOzs7UavVqFQqCgsL6ejoIDQ0lLfeeovHHnvMoS5ZVXnt2rUi4DAgIICWlhb8/f1JTU1l7dq1vPXWW3z33XfMmzePxx57jPvvv5+cnByOHDlCeno6J0+e5PXXX+ezzz4jMjKSL774ggkTJnDrrbc6vGnu37+fjz/+mCuuuIJPP/2UFStWkJWVxZQpUxg5ciRVVVVUV1fj6urKnDlz2Lt3L2FhYSgUCtLS0sjLyyMgIIDk5GReeeUVbrjhBpqbm8nOzmb58uWsXr2auXPnEhcXh9lsxs3NjcOHD/PCCy84xA00NDRQXFyMRqNhy5YthIaGolQqaWpqcsh2gNN73G+99RaTJ0/mjTfe4PHHH2fv3r3MmDGDL7/8ktWrV/PCCy9w1VVXodFo2LNnD0899RRr1qwhIyODl156iT/+8Y+ChXf79u0sXLiQhQsXinmz2WycOHFCvH12dXVx6NAhNm3axNNPP01jYyNGo5H9+/czdepUli1bJs61WCw888wzwOlMrsrKSp5//nm2b99OaGgoX3zxBXfeeScPP/ww8+fPJyoqSvAu6fV6NmzYwNSpUzl69CiRkZHY7XZ27NhBYmIiY8aMYf369axcuZJPP/2Uyy+/nPLycmJjY5kyZQo7d+4kOjoas9nM6NGjeeKJJ1ixYgUbN25kzZo1nDp1Srja4+Pj8fX1xWq1snv3bmbPns2FF144dCMewm8SVquV2267DZ1O96NZi/8LDBs2jAULFjB16tSha+hn4JyKodm0aRNdXV1EREQQGBgoXNKLFi0iISGBmJgYli1bRnFxMdHR0fj7+7Nq1Squvvpqtm7dKvaDs7KyUKlUjBs3jvnz57Nu3TpBytbQ0EBMTAzFxcWUl5fj7OzMqFGjOHnyJK2trUyYMMEhDVeGLHoYFxeHr68vXl5eLF26lKlTp1JSUsKGDRvo6ekhJSWlX5yFk5MT8fHxgnjtk08+IT09nblz59LY2OgQNCijubmZzMxM0tPTWbZsGZ999pnQQwoNDaW+vp76+noiIiIEbb7sqYqIiOCqq66io6MDg8HAwYMHiYqKYuLEiSQmJjoYMwqFgujoaPz8/Lj44ovx9fVl7NixLFiwgIKCAr788ku0Wi2xsbF4e3sTERFBRkYGBoOBoqIiJOm0crSbmxuRkZG4urpis9kIDg7GYDAQFRWFu7s7//d//4fNZqO+vh5PT0/i4+Md9tQVCgWhoaGCSXnjxo1MmzaNJUuWsGfPHjo6OhzGp62tjdLSUjIyMggNDWXkyJEsWLAAd3d3Ghsb0Wg0uLm5YbVaCQwMxGg00tPTQ2ZmJqmpqcTHx5OYmMjnn3+OwWAgLS2tH8Ovq6sroaGhJCYm4uXlRWBgIJdddhkjRoygpqaGzz//fNA5d3NzIywsjEmTJvG3v/0NrVbL5s2bhdRGSkoK/v7+uLu7c/PNN7N8+XIhzaHVasnOzmbSpEksW7aMTz/9lLCwMMLCwnj44YeZOHEiXl5eXHTRRYSHh5OUlMSFF15IQUEBERERTJ8+HaPRSFFREREREXh5eXHJJZcQEBBAY2MjmzZtYsaMGdx0001MnTpVjEFqaupPigsbwhD+k+gb8CrTUFx00UXExsb+Jg2G6OhozjvvvB/dlpb7dg5tsPwqOGc8NHD6rTs1NZXFixfT2dkpgrjkaPLecgLy93Ca56Q3Z4hCoRABnH5+fnh6ehIZGcnYsWN5++23ueOOO8jNzcXNzY1FixZx3nnncfDgwZ8UUNybzt9ms6FWq0lLSyMxMVHUPdgFZzKZUCqVjBgxgsDAwEGzAGS+Fy8vL3x9fQVXyJYtW7jooot4++23+dOf/tSvnt6ZL05OTlx11VXs27eP+Ph4pk2b9qP9kf+2Wq1oNBri4+M577zzaG9vp6CggIqKCkaOHCn2rf/2t7/x0ksv9aM6l8dSliGQ94mXLFkithL7tkGGHGwoizD23QaJi4sjOTmZvXv3smrVKtzc3Pjmm2+46qqrCAwMHFCewtfXl0WLFrFt2zYuuOACoqKiUKvVTJgwgejo6AGZTHtnWvSdc5VKxdixY0lISBh0zuXMjJCQEHHuwoULmTVrFt3d3QPyfgBCOdvLy0vEFPXedus9v72vi5KSEvLz80lNTaWkpMThWPmY5uZmlEolUVFRIsBw3LhxDBs2jLa2th9cu70hz3Nf/JAEw5mU1Zt3RPbo/RYfXr8GpH9T7cuyJ//puuTr8pem91ssFgwGg8i+ksuXg5e9vLx+dh2yR7u0tJTx48cLziT5OpalTHpLMfTu32DrT07p77t1PZikw49JH8jH9L7H9C7nh9oHUFpailarZdKkSYNu89rtdoxGI+7u7r+LreBzyqC58MILefLJJykoKCAlJYWRI0disVhQqVTodDo6Ojrw9PQUUfrt7e3k5uZSXl7OpZdeSk5ODjqdjrFjx7J582YyMzOpqKhg1apVKBQKVqxYwWuvvUZ0dDRubm689957PP744yQmJrJgwQI+++wz3n//fcrLy4mOjhY3GhlRUVEcOnSIESNGYDQa6erqoru7Gx8fH+bPn899993HjBkzmD59ugMZnLz3q9Vq0Wq13HzzzezcuRMPDw+SkpJISEgQcgl+fn4YDAb8/f2ZNWsWO3bswN3dneuuuw5vb2+WLFlCa2srF110Ea+88gpxcXHo9XoxPnL8RFdXF0ajEY1GQ05OjpA8CA8PJz093eEik49VqVQYDAY6Ojro7u6mp6eHCy64gGeffZY5c+Ywbtw42tvbyc7Oxmq1UltbKxTMr7zyStzd3UlKSuLrr78mOjpaROnL7Zk9ezZ33HEHXV1djBgxgssvv9xh/v38/FAqlZSWlrJ8+XK+//57xowZwwUXXNCPTE+tVlNeXo7NZiM8PJzhw4dTUFAg6s3PzycpKYlvvvmG4cOHo9frUavV5OXlERISQnBwMElJScyaNYt7772XadOmMWvWLId5g9OEhwcPHsTZ2RmDwUBXVxdqtRqtVivmfPr06cyYMaPfuYDgJhk7diwXXnghmzdv5pFHHiExMZHp06eL8XZxcUGlUtHS0sLkyZPJyMhg165dAKxcuVJkXxgMBiFVoVar0ev1dHR0iDizqqoqcnJyUCgU1NfXU1tbK6QudDodRqORhQsX8pe//IUZM2Zw0UUXMXfuXFavXs20adOYOXPmgP0YCFarlc8//5wjR44wZ84cYSzJSt5nAvkmbzKZ+Mc//sEVV1zBuHHjgNOG//PPP89f/vKXMw6+P9vQ1NTECy+8wJNPPvmr9XGwB7vZbGbNmjWMHj2aCy+88BfVoVKpePnllxk5ciQ33XQT8P+3WWNiYli+fPmAhJEDvTAOZDS88847XHXVVYJ4U5IkDhw4gK+vLxMnTiQvLw+NRoNer+e8887DycmJQ4cO4eXlhbe3NxMnTuxnuLzzzjs0NDTg5uaGl5cXt956KydPnsTLywuLxcJ5551HR0cH+fn5ODk5kZCQwPDhwx3Ksdvt1NTUsGPHDpYvX05ISAgqlYqTJ0/S1NRETEwMc+bMIS8vD51Oh06nY+bMmbi6unLgwAG8vLxwdnZm6tSpjBgxgpdffpmAgAAHos7edW3ZskXEHP4etoLPuaDg5uZmTCYTw4YNExpAgYGBaDQaXF1dcXFxQa/XYzQaefjhh3n66acJDg7Gz89PvNWGhYWJ1OqAgAACAwOxWCxUVlbS2dkp2Di7u7vp6Ohg2LBheHh4oNFoaG1tJTAwUFwYvReQyWSisbGRkJAQNBoN/v7+6PV6EXDa0NCAu7u7EETs3S/Z0AgKCsLLy4v29naMRiORkZEiUNbNzQ0XFxd0Oh0hISG4uLjQ3NyMu7s7YWFhODk5YTKZsNvtuLu7o9Pp8PX1xWw2C2NPfkvy9fVFpVLh7u7Opk2bmDx5Mj09PVRUVLBixQqHtyrZeAkMDKS7uxsvLy9sNhsWi0UoNNtsNmJjY7FardTX1xMeHo5SqSQ0NBS1Wo2zs7MYdzmbRX5jUavVBAYG4uPjQ2dnJ93d3cTFxfV787bb7dTX1xMUFIS3tzctLS2CKbhvwG1WVhbNzc3ExMTQ3t6Op6cnKSkpWCwWUa6fn59DW1QqFYcOHRKppF1dXVx88cXU1dWJ7bS+NwyZcTQwMBCtVktgYCBqtVow1NbX1w845wCvvPKKyGiTeXvUajVtbW2CG6Wrqwt/f388PT1pb2/H1dWVkJAQYRx4enqKWCKZEM5gMAix1u7ubsGCazQaCQgIoLm5WcxPQECAOFalUuHj44OPjw91dXX4+voSGhqKzWajrq4OT09PESR8ptfrgQMHeO211/jyyy9FdlFKSop4S5Vd77JnQPY0yeer1Wq2bNnCVVddxf3338+ll15Kenq6WJ/d3d0i+NNiseDk5OTgqZJvfzIPCJwmHZQFBnt7fXq3QT7XbreLNspB1nLdciqy7D3p/Xvfh6W8XalQKLDb7bi4uDgYFWazGRcXl35GhsFg4LrrruOdd97Bz89PtEUOFHd1dRVl9+4XnDYg5NTj3n3SaDRs3ryZq6++GmdnZ8xmM87Ozjg7O/P222/j5+fH1VdfjSSdFq6U2y/3W/aAy56cgR6iFouFdevWUV5ezj/+8Q8AysrKeOaZZ3jqqacGVH232Wzk5uZSXV2N3W4nLi5uwFgTSZK4//77Wb16tXiRUalUPPnkkyxatEiQ8D366KOcOHGCmpoaXF1dxXbrE088wT333OMQY2Oz2fj++++JiYlBrVbT3NxMYmIi5eXlrFy5ktdee40lS5awadMmrr76anx9fXnllVd49NFHHV5q7XY7bW1tPPHEEzz44IOCBXrSpEnEx8dz00038dJLL/HRRx/x6KOPcvToUZqamkRCwpIlS/jHP/7BzTffTEREBG+//TZTpkwhPT293xh0dHTw8MMP89JLL+Hk5CR4dnqvz95e8HPB2DmnPDRyPIeM3lIFvSnyAwICOHnyJGazWWTyKBQKwsLCxDGurq5Cmdhms/Hkk0+iUCi47777xMT3lUYYiGK+N9zd3QXrpXxc77cqub6B+tWXsr9venXvfvfePutNkS+3QYZct7u7+4ASB76+vnR2dgruAjmmo28MTe9xGChVODY2Vvzt7Ows+BXk+nvPjbu7u4MEAjhS+/+QXIMcayRjoD7JaG9vZ8+ePaSmphIWFkZGRsaAchm921JfX8+hQ4fQarWEhoaSkZGBi4tLv/b2hre3N8OHDwcQN8je/R1szmXPoru7O9HR0WLM+0ow9C6rd3+dnZ0dxqL3mMnGlNy+vug7P/J66l1Xb/ZWFxeXQdlcfwjyg9RkMtHc3ExDQwN+fn4UFBRQU1MjtgzT0tJ49dVXiYmJYcyYMQ5vmkeOHGH9+vWCRVUOup4xYwZTpkzh5Zdf5r777qO4uBiFQkFLSwuXX365w0MmOzubdevWER8fT2FhIY888ggbN25k6tSpwOmg9yuuuIK33nqLiIgIIiIiyMnJYcyYMRw/fpw77riDgIAAsrKyhCfuqquuYvPmzSJTbOTIkWRlZVFcXMyUKVOYN2+egxGxbt06LBYLEydOZN26dTz11FNs376dKVOmUF1djVKppLCwkNWrVzvMmTx+Gzdu5OTJk/z1r39Fp9NRWVlJa2srixcv5tixY7S0tPCXv/yFdevWsWDBAvLy8sSLxq233upg8B89epT169czceJEqqqqcHV1JT8/n9tuu01s1xw4cACtVktGRgYnTpygrq6OkJAQxo0bx9tvv82ECRM4deoUTz311ID3xO7ubhISEsjOzhacKXq9Hjc3twGpIyRJ4uDBg7i4uLBkyRJcXV1/0HDu/Z5us9koKysT12lLSwt6vR4XFxdCQ0P58ssvcXZ25uabbxYSEE1NTQ7XmZOTE4sWLcLFxYU9e/aI2DQ5s9HT05O8vDyKiopEJmhdXR16vd6h/zKxaO/78OTJk4UcgUxPYDAYcHZ2JjQ0lG+//ZagoCBmz54tDNqqqirxAjWQT0KSJE6ePElVVRV5eXlERkaSmZmJh4cH3d3dLFu2jBdffJGoqCjc3Nz4v//7v0HH8mzCORUU/FMwYsQI1qxZc0ZcHTIr5P3334+Xl9d/oXW/HQQHB/Pqq69yzTXXcOmll/6gjsvZAoVCwYUXXsjjjz/OpZdeykUXXfSD+lMyxo4dy4svvsjll1/OsmXL/qOBhU5OTtx+++1cccUV57xEhE6no6SkhJMnT6JUKlmzZg0zZ87k0ksv5d1338Vms1FTU8PKlSsd+JcUCgVxcXHExsaK+LD09HSuvfZaduzYQXBwMI2NjZjNZg4fPkx1dTWTJ0/uN2fBwcG0trZy2223kZaWxjfffOOQLl9dXU1QUBA1NTXccsstLF26lMbGRq688kquueYa/vnPf/LGG28wfvx4ISFQUFDA5s2bhYEuyweEhIRw4MABh9ghJycnpkyZQk5ODnFxcVRUVAjaApvNxqZNm/Dz86OlpYXa2tp+4+fi4sLFF19MWloax44dw2KxkJSUhF6v5+TJk8yaNYtTp06J+ouKiigoKCA0NJSsrCwHsjZ5TGUuGCcnJ5KSkigpKRFKzsePH0ehUHDRRRexe/du6urqCAsL49ChQ/j4+NDQ0MDy5cvFvA2EtrY2EhISsNls6PV6mpubMRqNwvPaFxaLhYKCAoKDgwW3T3t7e7/j7HY7RUVFwqMrSRJVVVVERUXh4eEhaCZkI8DJyQm9Xk9PT4+D5683iZ48Lm5ubpSVlWGz2fDx8UGv14tznJyc0Gq1grpC9gj2TRQYCOPHjyc2NpY9e/awfPlyfH19HdrX09PjUJdCoRBkeGlpaeTm5qLRaBzKdHJyIi0tjYiICKZMmcK7775LSkoKF198Mfv27aOuro6mpiZmzZrF9ddf/6NtPFtwTnlozhQKheJHvSkDHf97hEKhwMvL65wz5JycnH5UpLIvfuq6+SWQs+J+DwgJCeH8889n7ty5qFQqmpqasNvtIhtQ1sjx9fV1EInsC3mturu7Y7VaRcA0nI4jevnllzl+/DgvvviieJGRt7Dk4NTk5GRyc3P7GZHyMb6+vmKbycPDg8TERGw2G42NjWLbReb9ufXWW3n++edZsmQJNTU1LFmyhJEjRwqpgt4YNWqUSK0fP34869atY+bMmajValxcXJg9ezazZs0aMIjaxcVF0PmbzWZsNhvV1dV4enpitVqJjY0lJiaGL774grlz53Lo0CFiY2NZvHix4CwaDD09PTQ1NeHh4SG2ktra2tizZw8zZswQulVz5szhwgsvRKvV4ubmhpubm8N2V2/IW6UjR47Ey8tLbDMeOHCASZMmDfiSIG+Fa7Va8ftAho/0b0Z1eSvNarVy6tQpYcjJRqoMs9mMv78/bm5uImtSlsPoHbgrY9u2bcyePRs47bmUjSM5vtDT01NIC7i6ugo5gr7l9G1zYWEhfn5+zJ8/X4hM9m5fYGCgaJ/FYhH3IA8PD8H6PBgkSRLr09XVlejoaCEb4uvr+6MCv2cTzu1XP04vzoqKCrZs2YJKper3u0aj4fDhw/1otX8MZrOZEydO0N7eTmVlJeXl5cDpmInMzEwOHz5MV1cXBw4c4NSpU7+Z9DqLxSLa3Rt2u52qqqrfVFuHcG5DVhuWg43l+J9p06Zx6NAhmpqaGD16tCAClNWie0N2odfU1NDT00NPT484Vk61NxgMnDx5kocffhhfX99+9PFwOkhcqVRSUVHBBRdcQFBQEHl5eRQUFKBSqURAtdFoBE7HG7W1tVFQUMDSpUtZtGgRe/fupbOzE39/f6Kjo9HpdPzjH/+gqqpKbJvt37+fI0eO9OuHrOZ86tQp/vjHP5Kdnc2YMWOIjY2lqKiIr776iu3bt6NUKh3OM5vNol2y9Mnnn3+OSqWiu7tbaFZdfvnlZGVlkZCQwNixY1m/fj27du1i586d/e59ssBjeXk5n3zyCUajEbVaTWNjIz09PcyfPx+73c7777/PmDFjeOedd9i7dy979uzBYDBgMpkwmUz09PQ4eEPg9MNV9jTJhJ/ymObk5DBs2LAB7z9ubm5C12nChAlMnjxZbOX2hrOzM1OmTKGjowOtVouLiwuXXHIJkydPFnFuCQkJIqaturqaOXPmMGfOHGprawXTeExMDN9++y25ubkObS8sLBSEntOmTaOtrQ2z2YxOpyM9PZ2pU6fS2NgoJCNcXV1Zs2aNkMKR4760Wi1KpRKr1UphYSGffPIJtbW1rF27luDgYPz9/dHpdFRXVwtjtqGhAaPRiCRJIgg4KyuLqVOn9tsulyRJXA9Wq5VFixaxZ88elEol7u7ujBkzBoPBQE9Pzzl1vz+ngoIHQnV1NZs3b2b8+PEkJCSIeA6DwYBarUahUHDLLbfwySef/KQ3b5PJxB133MGNN96It7c3NpuNCRMm8M9//pP09HSam5spKCjgkksuobm5mQULFvzs7Qm9Xo9er3d4s/i5MJvN3HXXXaxYsUIEN8Pph8vmzZs5ePAgL7744lm/rTSE3z4sFgs5OTl0dHSQnp4usqN6enrIycnB3d2dUaNGYTAYyMnJYezYsf22+SwWCzt27CAhIYGWlhbCwsJwdXWlrq6OlJQUCgoKSE5ORq/XY7FY8PT0ZNSoUQ4emLq6Ou666y7+8pe/EBYWRnJysiBTHDlyJD09PYSFhVFcXMyECRMICAjgmmuu4c477yQwMJDx48djt9vJzs4GTscXyXF67u7uREVF4e/vz65du3BycmLOnDn9EgbgNKGnzHLeW8unrKyM7OxsJk6c2I/6vrOzk+PHjzN27FgaGhpEgkFhYaFQkJ4/f77QSRo+fDh2u51jx47R0NDA7Nmz+wWzWywWdu7cyahRo1AqlahUKgICArBarSKhICQkhNLSUiZOnEhhYSEqlYq5c+diMpnIy8tj/PjxFBYWMmzYMJKSkkT5drud/Px8Ojs7mTx5MgaDgYCAAGpra4VAZ98+9m5XZmam8OAlJCQI1ereGCgouKuri1OnTuHh4cGECRPo7u6mrq4OZ2dnJkyYgEKh4NSpU1itVmJiYoQhqVQqhUdGkiS2bt3K9OnTCQwMxG63U1hYiF6vF8Sger2e/Px8JEli9OjR+Pv7s23bNsaOHSuMtbKyMurr6wXJaW1tLaWlpcDpGLnZs2fT0tJCQ0ODMOCcnJzIy8vDZDIRFhYmYu9+KChYftFOT08nKCiInJwcbDYbCQkJeHl5cfToUUFfca5sa59TBo3dbqeyslKw6MbFxfHxxx+Tm5vL3/72N+Li4kS2weuvv47JZOKGG27g9ttv55577kGj0TB37lxcXFwoLS0VKr69DR1JkoQI3D//+U9uvvlmYmJiBDX/7bffzo033khsbCwPPPAADz30EFOmTKGpqYmqqipSUlLw8vKiuroajUZDSkoKBoOByspKRo0aRUBAAMXFxbi7u6NUKpkxYwYvvfQSvr6+3HjjjSLI2Wq1iuMqKytJSUmhra0Nq9XKtGnTMBgMlJaWYjQaGTt2LL6+vjQ0NNDS0sJ7773H9ddfT3p6OoWFhRgMBjIyMjh58iRfffUVzz//PJWVlej1emJjYwkJCflfTekQhvAfhSRJlJaW8vDDD7Nu3ToHOY3Bjpf1gNauXUt4ePiQ8f9fRG/OFjlWZSCD5oknnuCGG27op3j9U9DY2Ngv0eTntLegoICRI0f+4HbpzylXkiReeuklLrzwQsaMGfOrlX0249wwyzg9wceOHWPXrl0kJibywgsvUFJSgr+/P76+vg4uOXlv1d/fHw8PD3p6ejAajWzbto2DBw+yc+dOampqKCkp4cUXX3TYt25oaGD9+vXEx8eLFOgtW7bwzTff4OPjg7u7OxEREeJNMTIyktLSUvbv34/BYOBvf/sbeXl5PPjgg3R2dnLq1Cl2796N2Wzm3nvvpa2tjQcffBC1Ws26desoLi7GYrGIyPne/V27di27du3CYDDwwAMP4O7uzttvv019fT0vv/wyHh4emEwmHnnkEerq6li3bh3Dhw/HaDRit9tFZsUXX3zB5s2bRdltbW18/vnn+Pj4iEDAIQzhXIWbmxt//OMf0ev1Z3S8xWLhz3/+c7/A0SH85yGTS8pkm4MZKzfccANlZWVotdqfXVdoaKhQBv8lSEpK+o+IP5aUlDB58uQBOWh+rzhnDBqAnTt3EhUVRXh4OOPGjePQoUP4+fnh7e2Nr6+vAzuql5cXfn5+uLm54e3tTXp6OmPGjKG1tZUdO3bg7OzMpEmTuOiiixzqOHDgACEhIYSGhgohvqioKCRJwsvLCw8PD4KDgwkMDBSu2cOHD2Oz2QgLC+PWW28VbbzssstoaGjAbDYTHBzMbbfdRmBgIIGBgUycOJG4uDjB6yIHrslwdXUlODiYESNGMHHiRNzd3UlLSyMwMJCamhqysrIYPnw4GRkZlJWVsXnzZgICAggNDSUmJgaDwcChQ4cAuPLKKx0sfJl/Ze3atURGRp5Te6xDGEJvKBQKEhMTWbhwoYNy9A8dHx4ezgUXXCA8vr81/N4o8fv2V87UWrBggfBo/xy4u7v/4q0YhUKBh4fHf2SdpKSkMHv27B+VSfg94ZwyaCIiIsRepEKhGJCcSYa89dQ3a0B2ZZpMJtLT0/Hx8XE4xt3dnYqKCux2O1artZ92zUA3Eg8PD9ra2hg/fjwxMTEOb3aenp60trYybtw4YmNj+731ye3p29bBbliSJAkCpY6ODux2O+Hh4QQEBFBZWYnNZhMR8VqtloCAANLT0x1uChaLhQcffJDo6Gj+9a9/CYLC39NNcgj/PVitVlQqlQh4tFqtInDyl645ubzeWxUy4Z1ctiRJ4vr6JfXJZVsslkG5QTQaTb9+ycHRer3+F9efn5/PyZMnRV/69qv3ePRuc+/xGKxffX+Xv+9drnyc/Old10Dl9j6n9/dye+S/+85X7zltbW1l3759/YK9+xIGysHTP3WMZWkaOYDWbrej0WjQaDQDZp39FPQdp4HGbaB5HGi7TX4m9f3IJKe/h/v3OWPQKBQKli9fDsCePXvw9/dn7ty5tLW1YTQa+7ke09LSyMnJITc3V7DKylkCq1atYsOGDTzxxBOYTCYH0qkFCxZgNBp57bXXhNEgs7A2Nzfj6upKfX09TU1NeHp6Ul9fz5IlS6iqquL++++nqqpKCB92dHSwaNEimpubue+++ygrKxMsvW1tbYJwKi0tjaNHj9Lc3CzaYTabsdvt6HQ62tvbcXd3p6WlBTc3N0wmE/fccw9bt27lwIED3HrrrSxduhRJknj11VexWCz09PRw/fXX8/e//5033ngDb29v2tvbUSgUtLW1sW/fPlJSUpg5cybHjx/n888//+9M5BB+d9DpdLz55ps89dRT2O12enp6ePfdd8nKyvpFDwybzcaxY8c4fvw4a9euRavV0tXVxY4dOzh48CCFhYVIksSpU6c4cuQI27Zt68fn8VNgt9v57LPP+PDDDwf9/dChQ0LtGU4/rE6cOME999xDfX39oGX3NsYGezB1d3fz6aefMmbMGJycnGhoaODYsWPs2bNH8PHs3LmTEydOsH//fhEovHv3bvbu3Ut1dXW/su12O+Xl5bzxxht0dXU5tEfO5pSk08zbx44dY9u2bYKJffv27Rw7doyjR48OaAyVlZWxZ88e3nrrLWpra8UxnZ2dfPTRR8DpLf49e/awZ88eamtrsdvtHD58mBMnTrBt2zaRLm2xWBy2zfvWVVBQwNatW3nppZewWCyDjvNAMJvNfPHFF2Kb0WKxsGHDBnbu3PmTy+rbru7ubrZu3cru3bspKSnBYDCIccvMzBTBvfv372fHjh20tLQMWl5OTg5PP/00b731Fn/4wx/48MMPefzxx3n33Xd/P/dv6RyD1WqVdDqdZLPZJLvdLpnNZslkMkk2m83hOJvNJvX09EhWq1UymUyS1WqVzGazZDabJbvdLhkMBsloNEp2u93hPLlMvV4vmc1myWaziTrkssxms/jbYrGIc3p6eiS73S5ZLBZxfO/y7HZ7v/ZYLBbR1t5tkc+Tj+ldv1ynwWCQDAaDZLfbRb1yu+Xvev+/bz/kOuW29B2LIQzh14DVapW+//576eqrr5Zqamokm80m7du3r981+1OhVqulv/71r1JPT4901113Sfv27ZPWrFkjFRYWSj09PdL9998vtbW1SQ899JBkNpulzMxM6YMPPhjwmjeZTOJastlsksVikYxGo7iG5evt22+/lZ566qkBrxW73S5lZWVJN910k1RXVyfZ7XZJrVZLb7/9tvTKK68Meo5arZa2bdsmffbZZ9KGDRukhoaGAfvb0tIiPfDAA6KN77zzjqTX6yWlUimtX79eOnjwoPTpp59KdrtdeuGFF6Ty8nLpqaeekpqbm6XOzk7pkUcekaxWq0OZNptNamlpkW677TapqanJYWwfeeQR6csvv5SUSqX06KOPShaLRdqzZ4+0YcMG6bvvvpO2bt0q2Ww26fHHH5daW1sdyu3p6ZEefPBBSaPRSHv27JH+/Oc/SzabTbJardLXX38t/fWvf5UsFov0xBNPSK2trVJ7e7v06KOPSrW1tdIzzzwj2Ww26euvv5Z27Ngh2e12KTs7W3rzzTcHHBer1So98MADUk5OjqRWq8X9tPd9UJ7PweZgz5490h//+Efp6NGjkt1ulw4dOiQZDIYB6ztTWK1W6emnn5YqKyulvLw86cCBA9I333wjbd++XbJardJjjz0mNTU1SQ8//LCkUqmkmpoa6dlnnx30PpyXlyc1NjZKOTk50sKFCyWDwSAdOnRI+vDDD6WXX35ZMhgM4tkgr1f5/+cKzjliPWdnZwf238H2F2W6avmc3v8CgwZxKRQKQZjUu6ze9Q/0d+9z+qrI9v5N1kzpe37fCHm5HX3xQ31wcXHpV3dvwrze5fXtx+9BqXUI/xvIUhKLFy/mm2++4dZbbx007sBoNA6oau/p6UlwcLDDOb6+vjzxxBO0tbXh7u5OcnIyn332GZdcconIIiwqKkKn0+Hi4kJISAinTp3i+uuvd9iqKCoqEmysM2bMoKKigry8PEGy97e//Y29e/cKbqofCgDV6XSMGDGC1tZWoqOjqa+vp6Ojg7lz5w54vNls5rvvvmPmzJlC72uwa1GS/r/OlCw2es8997Bw4ULS0tLYt2+fkKhwd3cnPz+f8vJyfHx8cHNzo7a2Fr1e7yDH4uTkJBIcZNjtdkpLS0UwamNjo/BkBwcHs2PHDjw9PR0kKmpqahyIIl1dXVm4cCFubm5CTgGgqqqKyMhI4WmuqKjA29sbZ2dnqquryc/PFwHBwcHBZGZmCkoMqVf2U2+0tLRQWFhIQkICMTEx7NixAz8/P3Jzc1myZAm5ubkiq/PBBx/sxxpusVhwdnZmxYoVfPHFF0yaNAmbzTaggKrNZhPe9d6QiTJ7z11XVxc5OTnMnz8fvV5Peno6L7zwApdccglOTk6C9bihoQEvLy/sdjslJSVCb64vZBZ3mWPMycmJ6dOn09raytGjR3n11VexWq3cd999HD58mO7ubrKysrjvvvscZB7OZpwzW05DGMIQzk60t7cTHh7OokWLyMzMpKqqatAAXb1eT2VlZb/PQK54ma5eDqw3Go0YjUbxAmK32x3iWWRK+b7GkiRJjBo1Smxh+fr6YrVaueWWW8jOzqapqYlvvvmGBQsWMGXKlEEDQE0mEy4uLsTFxdHU1CREB+vr6xk1atSA59XU1Iit5YqKikEzdwwGA0eOHBGaVgAzZswgPDycN954g7a2Ngd6fycnJ3Q6naDqB86Iql+SJGpqaggLCxMvQ0ajUZTh7OwsCN16P7xl8jwZMvuxzWbjyJEj/OlPf0Kn06HVaomOjhZxIxaLxUFKQKfTOSR3yJlp0dHRdHd309DQ0G/+oqOjiY6OZurUqSiVSg4ePMiCBQvIyMjgtddeE2Khjz322KC6U4GBgUyePJm2tjbB6jsQrFYrtbW1/dZnXV1dPyNHJtcbOXIkCoWC999/v58Eg06nw2q1ijH4oRiggbK+5LUeHx/PbbfdRm5uLjU1NXz55Zd4eXkJcsFzBUMGzRCGMIT/GSRJorOzU4iOjhkzhq+++mpQEklZNbjvZyBvZUdHB3v37mXcuHH4+fmxYcMGYYzIb/K9DSeZZr5vZovRaKSyslIQaMptkOvs7Oykq6sLJyenfh7Q3ujs7CQ4OJhhw4ZRWloqgoBdXFwG1RKTAz1lY0ymihhoHGVVbEDoOD388MO8++67bN26FV9fX8EKbDKZCAkJcaDqlyULZE/HYO05deoUFRUVnDp1isLCQoeEBZPJREBAgJAFgNMPeT8/v37lWiwW9u/fz1VXXUVoaCj5+fm0tLRw+PBhamtrqa+vx9PTUwQtu7u7ExoaKubPZDIJOg5ZLbqv0dAXMoOyJElER0fT09ODk5MT3t7eeHt7D5jV1NraSkREBB4eHixatIh33313UP4h2Yjuuz77KqzDaWHY6OhofH19CQkJoaCgAH9/fzFuNpuNkJAQ3NzcRDKH3MbB5mcweHh4CDkIORZqzpw5vPjiiyQlJZ0zAcO/a4Pmhy7cIfw0/JKxHJqH3yckSaK9vZ2ioiLxBnrFFVeg1+sHFY11d3cnLCys32egN2ulUsnx48cxGo0olUqSkpKYNWsWtbW1KJVKYmNjSUtLIzg4GLVaTVVV1YBbP99//z1tbW10dnbS3NwsMrJkiYXg4GBMJhM7d+6ksrKStra2fsGiPT09HD16FGdnZyIiIujo6CAqKopjx44RFhY26PpPSEjA1dWViIgIxo8fz/jx4wfsq5eXF3PnziUrKws4/YC3Wq3o9XoCAwOZMmUKM2fOpKWlRbR90qRJgvSzubmZsWPH4uzszKuvviqCo6V/B65qtVo6OjqQJIlly5Yxbdo0/P39CQ8PJzk5GR8fH3Q6HTU1NUJKoK6uTngcEhMT2bBhAwUFBcBpw+jTTz8lOzub3bt3s3nzZjIyMjj//PMZNmwY3t7exMTEMGnSJOHNGj9+PBMmTBAZS3V1dYLtXGbejY+P72c4yJ4drVZLcnIyAOXl5ZSXl7NkyRKMRiM6nW7AjFeVSkV2djYuLi4oFAoWLVqExWLpJzUgw8nJiZCQkH7rMzg4uJ+xFBwcTHJyMm1tbdTX1zN9+nTmz58vtv5cXV1JS0sTpKn19fVkZGTQ0dHBRx99NKBha7fb6e7uFoKbkiQJcUtZCkQW+Fy3bh1btmwZUOTzbMU5xRR8ppAkidraWrKysrjooovw8vKira2N4ODgAd+wdDodWVlZQpBu5MiR+Pn5/WxuAbvdTllZGV1dXWRkZAhV3ePHj+Pr6yuyFH5J/1paWigqKsLT05Pw8HBBo/6fgEajYdOmTSxZskRQjUvSaZXbqqoqgoKC8PHxISkpqd/+v16v57vvvmPWrFk/yMhpNBqpqakhOTn5jMddkiSampooKirC399fPDSmTJnyq8UEmc1muru7CQ4OJjc3F7VajZ+fHxEREcTExKBUKsnNzSU9PZ2AgADhsm9oaGDy5Ml0d3dTXFyMp6cnERERhIeHk5WVhZubGx4eHiQlJQmWZ5koUvYuTJo0CaVSeday1UrSaZZembYgKChIZA7+Gn2SmcNbW1tRKBRkZGRgt9vJy8vDZrMxYsQIQkJCaG1tpa6uDhcXF8aNG9fP29PY2EhOTg7JyclCWVqv1zN8+HCKioqYOHEiOp2OnJwcoqOj8fHxITk52WGNdXR0UFRURGRkJPHx8ajValxdXYUxN378+AGNOEk6rf1z+PBhNBoNCoWCmTNnEhMT0+/Y1tZWXn/9dZ5++mnx//Lycjw8PEhJScHb25vi4mJ0Oh0hISEkJibS09NDQUEBkiSRkpKCj48PO3bsYNy4cURHRwstvObmZvz8/EhJScHT0xO1Wk1BQQFOTk6MHTsWlUpFY2Mjrq6ujBs3DoVCQUFBAUajkcjISIYNG0ZhYSFarZbp06djtVo5cuSIMJySkpJISkrCZrORn5+PWq0mOTkZX19fkY02ZswYfH19qa2tpa2tDU9PT1JTU3F2diYnJ4fjx49z++239xsXtVpNdnY2ERERJCcn09HRQXl5OYGBgSQlJVFYWEhPTw/p6ekOsYqSJFFXV0d1dTWjRo0SW2Gyx+aXrk9JklAqlRQUFODq6kp6ejqurq7k5+djMpmIiooiNjYWtVpNSUkJCoWCsWPHYjabKSwsZPr06f2eE1arlby8PNrb20lNTSU6Opq8vDwMBgPx8fEUFhYyYcIE9Ho9x48fJy0tbVCpibMSv2KA8VkDu90ulZeXS1deeaVkNBolSZKk48ePS2VlZQMebzAYpCVLlkibNm2SPvvsM+nmm2+WLBbLGdWj1+ul7u7uft9/88030sSJEyWtVivZ7Xbp6NGjUmJiooii/yUwGAzSrbfeKpWVlUnbt2+Xtm7d+ovK+zFotVpp5cqVklqtFt/Z7XZp//790oIFC6TKykrp6quvlvbu3dvvXKPRKN100039siDkMjo7OyWj0Sh1dHRI77///k/KfLHb7VJlZaWUkZEhFRcXSzU1NdLatWvPaO7OBBaLRdq0aZPU1tYm2Ww26bnnnpPuvfdeKTc3V1q0aJHU1NQktba2SmlpaaLvXV1d0uLFi6UHH3xQ0mq10q233iqVl5dLW7dulXbs2CH19PRIF1xwgfT9999Ln3zyifSHP/xBslgsUmlpqTRt2jSptLRUqq6ult5//33JYrFIX331lWQymX6V/gzhtws5c0nO3hwISqVSuu+++ySTyfSz7yF2u13Kzc0V98VfC3a7XaqtrZVaWlp+1XLlsnft2iWtW7fuVy97CGcXzrkspzNFXV0daWlpIlrcz8+PkydPDkgjrdFo6OnpYdq0aZw4cUK48tRqNU1NTdhsNlJSUlAoFDQ1NdHV1UVISAheXl48+eSTnH/++SxevNjBClYqlUJR1Wg0snfvXuECl6TTcu9NTU2MGTMGb29v6uvr0el0mEwmhg8fTllZGQkJCRQWFjJjxgyHzAqbzUZNTQ1KpZKZM2disVjEd2azGRcXF0JDQ6moqGDixIkiA8Hb25va2loMBgNWq5Xk5GTy8vKIjIykqamJSZMm0dDQQGdnJxMnTkShUFBdXU1tbS0hISH9ZOgLCgqYPXu2YBuWyZ3a29upq6tj+PDh2Gw23NzcCA4OFl4Yk8mEv78/FouFZ555hgceeACj0cjcuXMpLy8X8hLl5eWMHTsWvV5PTU0Nzs7ODm8bCoWCqqoqRo8ezfDhw+ns7OSaa65BqVRSX1+Pt7e3kL6QyQ/HjRtHW1sbGo2G4cOH4+7uLsagvb2d9PR0kVFx6NAhNBoNoaGhQqhu1apVBAUFCXmJ1tZWRowYQXNzMzabje3bt2MwGFi4cCGA2P6YNWsWNpsNlUqFxWIhIyODw4cPYzAYAKisrCQtLY24uDiUSiVXX301zs7OqFQqWltbiYyMJCsri8mTJw8xh56DGEizqC8CAgK4/PLLycvLY+LEiT/bCymrRP/aiIyM/I+U29raKrbChvD7xu82hubw4cNMnz7dQQW2pqZmwGPz8/NRKBQ8+eSTFBcX8/zzz9PR0cHrr79OQEAAn332GcXFxWzdupWDBw/i5+fH008/jclkoqWlpZ9wmMFgEGmHSqWSnTt3EhAQwNixY/Hy8mLbtm0cPnyYAwcOUFVVxbZt28jPz0ej0bBx40a6u7t5+eWXKSws5L333qOzs9OhfC8vL/76179y3333UVNTg6+vL1999RUVFRU0NTXx8ccfo1Qqef/99zGbzTz99NOo1Wo2btxITU0NjY2N7Ny5E4VCwb/+9S8OHjxIWVkZ7733HnV1dXz55ZeoVCpRTmFhIZMmTXK44dpsNg4dOkRXVxd33303K1asYO7cuRw/fpxNmzZRUFBATk4ORUVFjBo1CqvVyrvvvgvA1q1bqampQa1W4+3tTWhoKJ988gn19fW0trby2WefYbPZ2LVrF+3t7bzxxhuoVCq++uqrfoyj+/fvx9vbm1deeYWamho8PDxob2/n9ddfJy8vj46ODrZs2cKmTZsoKioSWlg9PT2sWbMGm83GmjVrOHbsGDU1NaJ8u93Oxo0bhdpvd3c3lZWV7N69m9dee43nnnuOqKgoKioqmDFjBs3NzeTk5IjAPNlQveuuu1i9ejX19fX4+vqSm5uLs7Mzjz32GBUVFTz77LM4Ozuzf/9+vLy8eOmll0SwJJzOlOjs7MRisXDs2DER9DmE3x+cnJxIT08nPT39Z29Zy0Gtv/YWxH+qXDjNEH/++ef/IpmDIZwb+F0aNEajkfLyctLS0sR3ra2tA94EJEniwIEDXHvttQQGBhIUFER0dDT//Oc/mTRpEuHh4XR1ddHZ2cm6detYsmQJCoUCjUaDUqkkLi6O2NhYhwu5rq6O2NhYQkND+f7774mNjaWkpIQ5c+bQ0dHBP//5T7y9vRkzZgwhISF8++23nH/++VRWVooguaCgIFJTU/Hz83NQw5b+nep3wQUXcNNNN/HCCy9QWlrK0aNHmT9/Pvn5+UybNo3a2lpSU1PJz88XAWQHDhxg1qxZlJWVMXnyZIxGo+DIGD16NAcPHkSr1XLZZZeJ+Jj09HRaWlqYOnVqPw+UUqnkpptuoqamhunTp2M2m3nhhRcIDg7G39+fWbNmkZWVxZQpU9i5cycAiYmJtLS0kJKSQnV1NTNnzsTb25uOjg5SUlJISkqira0NlUrFwoULWbt2LQBNTU3ceOONDm0wGo3k5eVx++23c9111wnvV3V1NdHR0Sxfvpz09HQqKirIyMhg6tSpVFRUcN555wmadK1Wi06nY8GCBVx55ZVijRiNRjQajdDxKiwsZMSIEcyfP5/29nYmTpyI1WpFp9ORkpJCYWEhJSUl+Pr6EhsbS2BgIDqdjsWLF3PjjTfywgsvYLfbOXDgACtXrsTX11eI4xkMBgoKCvi///s/Vq5cSUpKiggIlOMWPD09ufvuux14hc5lSP9mzTWZTD9IrW8wGM444Fwu02w2/0eC1M+0PbI3s++nr7Heu++92w6ckeEg9UqNljOHZOkGmTZ/oL9/ztgMNl9yfXL/5L4PVr/8/97nDCQBIP8u908+9r8tA9B3Xv5b6Ls+fi/43Rk00r8Dvby8vESkuiRJ5ObmMnHixH7HG41GcnJyyMjI4NJLL2Xr1q309PRQV1dHamoqNTU1hISEEBQURGBgIL6+vuzbt4+rr76a/Px8Ro4c6ZDxIEmS+D48PByNRsOYMWMoKSlh4sSJ1NfXExYWxvz584WwZGhoKHq9nh07dhAbG8uRI0e4/PLLaWpqIiwszEH/yWKx8NVXX2G328nIyMDNzY2srCzi4+NFSuSECRPIzs4mPj6e48ePk5KSwvHjx4mLi6Ozs5PDhw8TERFBWVkZo0ePJjIykoKCAqZMmcKCBQtITk7m4MGDpKWlUVFRQV1dHb6+vg4319zcXGJiYhg1ahSjR4/mwIEDKJVKJEliwYIFzJ07V2xZBQcHc+DAAcaPH09ubi5dXV04Oztz8uRJEhISaGxsxMPDA2dnZzw9PUUAYnJyMkVFRSxatIgLL7yQoKAgB0K0uro6LBYLw4YNIyIigoKCApqamsjMzOTSSy/FycmJrq4uzGYzkyZNoqqqSmRBHD16lGuuuUYYgIGBgQ43TUmSCA4OFuRa+/fvZ/r06UydOpXOzk5qamqor68nMDCQiIgI6urqWLhwIYcPH+a8887DarXy9ddfi3lycXGhp6eHU6dOMWXKFC677DI2b96M2WymuroahUJBTEwMERER5Obm0traisFgoLu7m+joaMrKygRnx+8F+/fv57HHHhNpvX2zPrRaLc8///xPeph89913rFmz5tduKnA6AP655577UZXusrIy/v73v/Pee++xatUqPvroIx599FHUarU4pq6ujrvuustBQmHPnj08/vjjP6lNVVVVrFq1ii1btgiD64UXXiArK4s9e/awcuVKcnNzOXr0KNdddx3Hjx//2Q/JAwcO8OSTTzp819nZyV133cW7774rjJUPP/yQ77//nvz8fFauXMnu3bspLS3lxhtv5LvvvqOgoIBVq1axbdu2AdO0NRoNq1ev5v3336epqYk//elPfP7559hsNgoKCnjxxRfPWFn9l0KSJL777jveeOON/1p9VquVxsZG7rrrrh+U8pCPPZfwuzNoenp6yM7OJi0tTZA96XQ6uru7mTZtmsOxkiRRXl5OfHw8cFrddNy4cZSWlnLNNddQUlJCXV0d99xzDyNHjmTSpElkZWUxatQoFi9ejN1u75c5pdFoqK2txWw2s3jxYu68805OnTrFqFGjMJlMJCUlERMTw/fffy9YRc1mM3l5eYwbN06kBU6ZMgW73U5YWJhD+WazGaVSybfffsuuXbu44447hFcnOzubyMhIQkNDCQgIQKPRkJKSQlRUFOPGjaOjo4OKigqRpaDT6bj00ktRKBTMmzcPpVLJvn37sNvtTJ8+nZKSElpbW0lKSnJ4mJjNZioqKkhNTcVisbBy5UoqKirw8fEhIyODTZs2CeHMqKgonJycmDVrFocPH8ZsNpOcnIzdbsfX11fEBslCo+7u7kydOpWxY8fi6urKVVddxa5duygsLHQYB7vdTkFBAcnJyezYsYMPP/yQw4cPExgYSFxcnGDV1Gg0LF26FA8PD6ZPn46/vz8nTpzg0ksvJSUlBZPJxNKlS/sZCh4eHsTExCBJEl1dXRgMBsLDw/Hy8uLaa6/l+PHjZGdno1AoiIqK4tlnn8VsNou0Sdn79e2337J7927uvPNOKisrBZNrWloaqampFBUVUVBQQFJSEtu3b+fDDz/k2LFjBAcHU1hYSEZGBt7e3nzxxRe/SIfof4HeHgGZMMxkMgkvhiRJGI1GzGYzVqtVcLDI8UlxcXG0traK7b/eW4IAPj4+3HXXXYJbxWAwDOjpsNlsGAwGcb3K+m+9vQZyqrP0b74XWdPHaDRis9kwm80O3o6BCPq8vb35y1/+4sAr0tPT088Qc3Fx4aabbuKCCy6goaGBq666iiuuuEKkYdtsNkJDQ+ns7BTjpVAoiI2Npb29XYyrXq8XHhW5j70NAIVCwYgRI4iPjxfZVq6urvj4+DBx4kSSkpJoaGhg9OjRjBo1irq6OkaPHu3gyZbHo/ecWSwWTCaT8HTJ8xoUFERbW5vDuISGhorYRJkFWZIkZs6cSVJSEk1NTSQlJTFy5EgRj5acnExTUxOjRo0aMCvV39+fKVOmiKyzUaNGUVRUJBiGp02bhpeXl8P49F2XMueP3H6DwSD603vuBjKopH+nSptMJhQKhVhTvdefPB9y/fLcylpd8toymUxijVksFtGG3u3pXW9eXh67d+8mKCiIjo6OfnPT+/pqbW1l/fr1P9vr9lvE7y5te6DFe+jQISIiIkhKSur3Fv5DsNvt/S5uSZLEdwO5Q89kuOVjesf3yGRKfR+sfeuQ22Cz2VAoFCIw0G63s3XrVurr67n99tsd6pDLGKwe+Ri73e5Ql3y8/G/vNvxQ33qPUe8x7Ft/73YM9HfvdsllnEkb+var9999x773sX37sWvXLqZMmTIgL8iZYKB5OlPIInYLFy7E29t7wLX2W4fZbGbNmjVia2/58uU0NDRw/PhxLr30UnGTb2tr47zzzuPZZ5/liSee4P3332fChAkkJibyxBNP8PLLL7NixQpuvPFGIWsgSadF/eR4pszMTNzd3WlsbOTKK68UD0KLxcKXX35JXFwcBoMBDw8P1qxZQ0ZGBo2NjTz99NMcOXIErVZLS0sLMTExdHZ2olarmTVrFu+88w7PPfccO3fuJCkpibVr1zJu3Diys7O54YYbHJiDq6ureemll3jyySd57733cHZ2prOzk9TUVFasWNFv7dbV1bFy5Uq2bduGk5MTa9euJSkpiZqaGlauXMmVV17JnDlzyM3N5Y477iA4OJjnn3+ef/7zn+zZswetVktubi6rV69my5YtxMXF0dPTw4IFCxzmobS0lD/84Q9s3LiRlpYW6uvrufDCC6mrq+O6665j586d6HQ6li1bxtatWx1IAMvLyzl8+DB2u50xY8YgSRLr168nLS2N4uJinn32WQoLC6mrq6Ozs5PMzEw++OADh/tmV1cXl112GWvWrCEsLIwtW7Zwww03oNfrufDCC/nwww+Jjo5m6dKlvPjii4wcOZLFixfz7rvvDpjEAadDCK688krWr1/Pjh07eOedd9i2bRu7d+9m3rx5ZGVlodVqKS0t5f777xcvTpIkcfToUTQaDVlZWVxxxRWUlZXh6enJyZMnufXWW1m/fj0mkwmtVkt8fDw33XSTmDubzcbGjRvx8/MjPz+fefPmYTAYePnll0lNTaW1tZVnnnmGPXv2MGzYMLq6ujjvvPPYt28fbW1t2O12xo0bx6effkpQUBDjx4/n22+/5cUXX6S8vFzIMKhUKg4ePMi9995LRESEqPvRRx/FarVy5513ctttt3HxxRcL8VOA48ePYzKZGD9+PNXV1WzcuJHnnnuO4cOHn1X3jsHwu/PQ9JZdlx9iaWlp/YyZgY7t+5HZKeWP/IbR+/8/tUz5PPnc3vX0/m6wOuTvXF1dBRmUQnGa0r29vZ3ExERhAMjnyv8OVo9crvyG07dd8r9n0se+YzTQ333bNdjfvdvV92F+JuPc+7jeYzfQOA+0js477zyhOfRzPgPN05l+ABYtWoS3t/ega+23DjlAOiIiggcffJB169ZhtVoJDg4mOzubwsJCjh8/ztSpUwkMDKS9vR1nZ2fc3NwclJ99fX2JjIxk0qRJDvo64eHhNDc3Y7fbycrKEh6t3uNksVjYvn07ZrOZlJQU4HSQ6W233UZDQwP19fWsXbuW+fPnc8kll/D222+TmppKXl4eMTEx1NXV0d3djZOTE8nJyeh0OiZPnsy8efM4evSoQ3/Dw8NpaWkRaywyMpJVq1axb98+BwN8sDWXkJBAeHg4+/btE3o+N9xwAzfddBNvv/228PTIiQTe3t50d3dTV1fHjh07MJlM/RIUAEaOHElycjKbN2/mxIkTPyjf0BcKhYLRo0fj7OzM4cOHCQkJQaVScf3119PZ2UlTUxNr165l3rx5zJw5c8Asp8DAQJYuXcq6des4duwYkydPPqO6fwhhYWGkpqbyzTffIEmS6J+sML5v3z58fX1paWmho6NDnKfVatm4cSPz589n+fLl+Pj44OfnR2JiIlVVVTQ3N+Pm5kZQUBA33HADBw8edPCw1dfXs2vXLubPn8+8efN46aWXsNlsxMTE8PDDD4ukj507d6JWqxk/fjw7d+6koqKCkJAQioqK8PHxobOzk9WrV3PRRRcREBBAZWWlyLz89NNPxbVTXl4u6nZ2diYuLo6UlBRCQkJwdXVl2bJlYudAniuFQsHRo0eJi4sjLi6OhISEs+7eMRh+dwZNXzg7OzvEXpyr8PHxEW7sX0LaN4T/Dy8vrx8UIvxP1+3l5XVWr1vZGPXy8sLd3Z3W1lYyMjK48847ueaaa7j44osBeOihh1CpVD+57N5G3vXXX099fT1PPfWUSIWH01uYf/7zn/nXv/4lsuTc3d2FIKvBYKC9vV3Q97u7uxMTE4O7uzvfffcdGRkZfPjhh0RERODq6ioMLjc3t35swXKb5L/d3d1xcXHBZrP9qEfRYrFQUVEB/H99HldXV9zd3UlISBBbNXDa4+Hh4cH555/Pyy+/THx8PH/+859Zu3atiK/rDScnJ2688UbeffddjEajIMf8MchbK+Xl5fj6+gpZCDc3NzF+er2exsZGhxeZgeZq+fLlHDx4kKysrAFfLn+oDQONnUKh4LLLLuOtt94iKSmJq666ildeeYVRo0bR2dmJn58fCxcu5NVXX3Xor06no7GxEbvdTlJSEiEhIbS1tdHa2oq3t7fwUsvSFwPFbcmxghEREZjNZux2O56enri6ujJ69Gjsdjt33XUXGzZs4OOPP6ahoYHhw4dz8cUX8+yzz+Lu7o6HhwdeXl64ublx5ZVX8umnnwKnn1dGo5E5c+bw9NNPM378+EHXjrwW3d3dsVgsYq78/Px+VCLibMXQk+13gh/zOAxhCP9tyKKLWq0WZ2dnRo4cyXPPPcf3339PXl4ex44dY9WqVaSmpqLT6fDw8ODAgQPU1tbS3t6ORqMRcQgeHh4UFBSIYE85DkIWSszMzOTuu+8mPDycnp4e0Qa9Xk99fT1PPvkk7e3taLVaQROv1+vx8PBg0qRJHD58mOrqajIyMggNDeWiiy6isLCQW265hZKSEsaMGSPiIgwGAzqdrl9Gk1ym3CaZkr5vbIvcfqVSiV6vR6PRUFdXx759+2hvb6e7u5uOjg6cnZ2FdMTll1+OwWCgp6eH2NhYcnNzWbduHZs2bRLbSE8++WS/GBY4fW+YOHEiw4cPFx4sOTasp6eH7u5uuru76enpoaury+H8PXv2UFtbS2trK83NzSiVSqE7JcseDBs2jM8//5zCwkI6OjoGDMiNjo5mzpw5pKam4urqiiRJIsOwq6sLnU6HTqdDqVSi0WjQ6/V0dnbS3t7OO++80y+4VaFQMHnyZCZOnMi4ceOYMWMGY8eOZdSoUSQnJ3Po0CHWr1/Pt99+69AeWd/qH//4B19++SX19fV89dVXYgxqa2vR6XQOc9e77oSEBPz9/SkoKKCkpISlS5fi7e2NXq+nvb0dnU7HjBkzKCsr48knn0StVjNx4kTWrl3Lpk2b2L59u+ifHIOTnp6OSqUSzNZGo5E333yTTZs20dzc7NBvPz8/KioqaGhoEGtMXtM7d+4U1BfNzc04OzvT1tbWL/bsbMbvLoZmCEMYwm8DFouFU6dO4ezszNixYzGZTBw8eBB/f38mT55MY2MjGo0GT09PRowYQW1tLVVVVQwbNgwPDw+sVitNTU1MmDCBtrY22traRMaYJP1/naiUlBTUajVGoxEfHx8HF7vFYqGgoABnZ2ciIyPp6uoSCsjFxcUkJSXh7+9Pfn4+Hh4ejBo1Ck9PTwwGAyqVSmSwxcfHYzQayc7OJiYmRhBmTpgwQcTryO0ZOXIkLS0teHh44OPjQ0NDA+np6Q4p9zabjdLSUpqbmxkxYgRRUVEcPHiQqKgo2tvbSUlJQaVS0dbWRnh4OImJidTW1orxaG9vJz8/nwkTJhATE0NhYaHY5hpIyVySTpN5hoeHi6DlsrIympqaBAFmfX09cXFxjBgxQsS6dXZ2cuLECZKTk6mpqSEuLo6WlhaSk5MpKSkhMTERLy8vjh49KspOSUkZcOupvb0dT09PkTEpq1ZHRkbi4+NDZWUlISEh+Pn5UVVVRXh4OHFxcRQVFZGRkdHP8yxJEg0NDcTGxgKnKQ5kyYiysjJKSkpIT08nJiZGjIdsSB45coQRI0YwevRoCgsL6erqIigoSKiTOzs7ExgYSG1tLenp6UK2QjYES0tL8ff3Z9SoUULQ0263k5iYKAL6nZycCAsLIzQ0lJycHNrb25k2bZowYMeOHUtQUBCSJFFfX090dLQwQjIzMxk5ciQpKSkO/VapVBw7dozRo0dTXV1NUlKSoCSJiIjg5MmTJCcnU1tby9SpU8nMzGTMmDFERkaeEy+6QwbNEIYwhCEMYQhDOOsxtOU0hCEMYQhDGMIQznoMGTRDGMIQhjCEIQzhrMfvVpxyCEMYwm8D8q63yWTC3d39J2W4/NRzBoPdbhfBnQqFQsS9yKRnfbN0bDYbNptNUBnIJGcuLi44OTk5BIrKafnAgJlPvX//LUCWDPg54ypzK8nn/xhkokRPT8+fNO8WiwUnJ6cBifXkYHM5w+en9EGSTosO22w2AgMDHTiyuru7cXZ2xt/f/1eZr94cWnIgtsVicVgP8njK3/Vej7+UqkG+ftzc3M6ZzNdzoxdDGMIQzkrIsgVlZWXcddddP0mmoLKykrvuusshDfvHYLVa+6Xawul03ccff5xHH32U9evX88wzz9Da2kpPTw+vvPIKq1evFkaKJEl8+OGH3HrrrbS0tJCbm8vrr7/Oxo0beeyxx9BoNHz00Uf88Y9/5Ouvv+app54iPz8fi8XCunXr+OMf/8h3333H119/zcsvv9zPyPlfIysri40bN/6sc+12Ox999BHr1q1z+H6wcW9ra+O1114b8LfBYDAYeOSRR8jOzu73mywBsWvXLt57772fnL0jSRLFxcXccssttLa2iu9KSkq48847qa+v/1UygtRqNTt37uTjjz8GTo/PsWPHhLitSqVCq9WyZcsWjh8/LsgLT5w4wbFjx9iyZQvd3d2/qA2NjY3ccccdDpIaZzuGPDRD+Fno6OgQdOnnoibI2QJnZ2cmT56Mj4/P/7opPxmSJHHw4EEUitPq40qlEpPJhM1mE2ri8v8H4twJDw8XacLAgOfA6Qeg7F1Zt24dF154IREREQ7lyTIbPj4+XHfddTzyyCN88MEHPPDAA8TGxvLhhx+yYsUKJk2ahEql4vDhw+h0OiIjI1mzZg1Tp05l2bJlfPvtt9jtdmJiYjCZTFx11VV88803PPnkk3z++efExsZiMBi49NJLgdMsvQNl5xiNRpFJJHPjyN4oq9WK2WwWYyLLLnh4eAxYltVqxWg04uXlJZi9e3p68PDwEB4OWXrC09OT9PR0xo4dCyCyenrX1xdyea6uroJ0rrKyEp1OJ7w069atY9GiRURGRjrMT3h4OLfffjtOTk7CsDOZTKIuORW+d9tlr0ZfQ1Yetw0bNvD4448THBws0r9788bI0gUDeYWcnJwICAggJiaGlpYWoqKi6OnpoaCggISEBNLS0n4V74y7uzsBAQEcPHgQgIaGBvbs2cMDDzxAWVkZX3/9NZ6ensTGxjJt2jQee+wxYmJi+Pbbb/n73/9OXl4eX331FbfeeqtDe3rLKnh6euLs7Cw03uQ5hNNeOB8fH1Qq1U8yJn/rGDJohvCzUFxcjJubG9HR0f/rpvyuIROZnY2wWq1s2LCB+Ph4hg0bhk6n48svv+Tw4cM8/vjjmEwmSktLycnJYc6cOcyZM6cfG7RWq+W9997j5MmT3H333fj6+lJZWcmJEydYvHgxzs7O6HQ6WltbmT59Oh9//LFQuu+9XSHzM5lMJlpbW2ltbWXhwoXAaVLKyy+/XEguZGVlMXnyZPbt2wfA1KlTeeaZZ5AkiYULF4oHpdVqRalUUlFRwciRI0UdOp2Oqqoqmpqa8PLy6id7YbPZePbZZxk/fjxubm4cPnyYRx99lE8++YT58+ezefNmIWOwaNEitm7dSkhICDExMaSmpjqUpdPpWLt2LaNGjaKxsZEbbrhB0Op/++23rFq1CkmSaGtro76+nqVLl5Kfn09jYyOrVq3igQceYOrUqRw7dozbbrutn4Cv2Wzms88+IyYmhszMTC677DIA8vLyeOqpp9DpdNx99918/PHHREVF0dHRwcmTJwX9vpubGx9//DGPP/44Tz31FImJiZSUlLBkyRIyMjLYu3cvHR0dGI1GbrnlFr744gvi4+MpKysT89MbtbW1VFdXk5uby7Rp01i/fj0JCQkcOnSIa6+9li+++EJoIT3yyCP9tsXktO2xY8eKFHhZtmHSpEkDrmNJkgRPTV8EBwcPuEXl4eHhIJkSFhbG4sWLxbpxd3cnLy+PtLQ0YeydPHkSnU6Hi4sLAQEBFBQU9KvPbDbzzjvvkJqaSm5uLqtWreLhhx9m8uTJHD9+nFtuuYWoqCh27NhBVFQUnZ2dA/bpbMWQQTOEnwWF4rToYlxc3P+6Kb879NXV+i3FX/wUuLi4MGzYMNLS0vD398fT05OrrrqK5uZmCgsLOXz4MOPHjyciIoK8vDzmzJnTrwxPT09uvfVWMjMz+ec//0lgYCDTp08nPDyc/Px8vLy8qKqqYuXKlYSHhxMeHs7EiRMHjL2A0/wkN9xwAytXruTyyy8X43vZZZdx5513UlRURFNTEyNHjhQGjUye9o9//IMDBw7w7LPPAqd5Ve655x6io6N5+OGHxTyZTCba2tpoamoiNja23/w5Ozszc+ZM9u3bx/XXX8/bb7+NWq0mIiICf39/xowZg7OzM1u2bGHatGls3bpV1DMQRo8eTWBgIOvWrWPp0qVs376dtWvX8t1335GYmEhbWxuBgYEUFRVRVFREREQE2dnZ+Pn5odPpmDlzJi4uLpw4caKfQVNYWEh+fj7XX389kiTxxhtvsGDBAkaNGsXq1av5wx/+QEVFhRh3lUrFmDFjOHnyJMeOHWPFihU0Nzfj4eGBzWYjLS2NUaNGsXfvXjQaDTU1NSQmJnLw4EEOHDhAV1cXq1atYseOHf36qVAoSEhIICwsjIyMDE6cOEFXVxc333wzjY2NfPrpp7i7uxMXF8cll1wy4IuAHP8zYsQIiouLaW1tFWR1F1544aDXmlarFVtUveHp6XlGWm/e3t5MnDiRhoYGjEYjy5YtIysry8Hj1lt8VvZQDaTvN3LkSMLDwzl8+DC33norPT09TJs2DQ8PD44fP45Wq2X27NlMnDiRf/3rXz/atrMJQzE0QxjCWQRJkqiqqiIzM5OPP/4YrVb7v27SrwYXFxexbWE2m2lpaWHy5Mnceuut3HDDDQPGLri6uuLq6irE9dra2pg6dSr/93//xzXXXMOSJUsICAjggQceQKlUivMGi4MYO3Ysd999N59//rmgsJckifDwcBYtWsRjjz1GSkqKMIisVivl5eUsWLCATz/9lBMnTlBSUgJAVFQUDz74IMePH6e0tFTUERwczPTp07nmmmtITU3tF8OgUCiYOnUqtbW1ZGZmMmrUKD755BOSk5MxGo0UFxcLIreIiAiuv/56nnjiCY4dO9avP2azmeLiYtzd3XFycsLPz4/Ro0ezc+dOZs+eTWhoKO3t7dTW1uLv7y/kC2TIemOurq4DbitrNBq6u7uRJImoqCixDeTp6YmbmxsjRoxwOE+lUlFeXo6/v38/UVuFQiHqstlstLS0EB8fz7Jly3j66adpaWkR21JnIuaqUqnE2MbGxqLT6XB2dsbT03PQIGS1Wo2fnx+RkZHU1tbS1dWFj48PWq2WyMjIQeuSFdf7fs5UYkCSJFQqFTk5Odxxxx1YLBZB5CeXL5MCysG8gYGB/fpgtVrFWpMZl/vOYVVVlYNe3rmEc6s3QzhrIXsden9+TkCfnG1yNkPOdhisH1VVVaSkpKDX6ykrK/svt+7XhZ+fH0VFRTQ2NqLX69Hr9Wi1WgwGA+PGjePpp5/mm2++IScnp9+5cqZHa2srRUVFrFixgtTUVHFObm4uR48e5YorriA9PV246/Py8voFH1ssFrq6uujq6mL+/PnMnj2bBx98kI6ODqqrq2ltbWXFihX4+vqSmppKR0cHGo0Go9HIBx98QHZ2Nj09PQwfPpzQ0FA6OzvRarXExsZyzz338MADD1BZWSmUuisqKigpKeHdd991MLRk+Pj4MHXqVPR6PStXrqSiooL4+HgKCwvJycmhsbGRzs5OcnNz8fDw4O6776apqalfOVVVVRw5coTGxka6urpoaWnBYDAIj4/BYGDLli00NzcLw6azs1NQ78tyCjJ9ft94i7Fjx4qg7uLiYq644gq8vLxQqVQ0NTXh5uZGRkYGrq6u5OXlsXXrVmpra2lpaaGxsVFIIfSWg9BqtfT09DBp0iTef/99vvjiC7Zt28aECRM4cuQIhw8fpqGhgaamJof2yFtJspzGjBkzaGxspK6ujqqqKpYtW4ZWq0WtVg8YN2I0Gjlw4ABWq5WQkBB6enqIiooiKyvLgcV5IAQHBzNixIh+n4CAgAGPN5lMNDc3o1KphMTEiy++SHFxMe+88w65ubnMnz+fmpoaVCoVfn5+TJ48WbBZV1ZWMm/evH7ltra2smvXLlF2fX29wxz29PQwa9Ys1q1bR3Z2Np2dnQPKYZytGGIKHsLPwoEDBxg2bBjDhw//xWVJksTOnTtxcXEhKyuLqVOn0t7ezuWXXz7o1sBAZRQWFtLa2sr555/fL1AOOGu2ZqxWK1999RXnn38+wcHBA/als7OTjRs3cu21156VAcEyOjs7yc/PZ8SIEdTV1TFq1Cjq6+vx9vYmLi6OzMxMPDw8SE9PHzDeoaKigo6ODiIiIhg+fDhGo5GjR4/i7e3NpEmTaG1tRaVS4eHhQWJiIqWlpSJ+o/fbqV6vJz8/H4VCIeIWTp48SVhYGEqlEn9/f0aOHIlKpcLf35+ioiI0Gg2jR48WBoBKpWLkyJFERUVRWFiIVqslJSWFwMBATp06haenJ2azGbVajbe3txClTE9PH1AKoLu7G4VCgbe3N+3t7cIDcuTIERISEmhoaCApKYn29nZcXFwYPny48NzIMJlMHD58mNjYWFpaWggJCeHAgQMEBASg0WgIDw9n0qRJwmBqbW0lICAAnU7HyJEjKSkpYdiwYahUKrEl1Ns7IkmSMPqCgoJITEzEbDaLsRw5ciQBAQGUlJRgMBiIiYkhLy+PpKQk6urqGDZsGI2NjSQlJVFbW0twcDB2ux21Wk1aWhrFxcW0t7czdepU/P39KS4uFv2IjIwkPDzcIcW5s7OT0tJShg0bxrBhw2hqahKSDlFRURQUFODq6sqYMWP63Vs0Gg0lJSX4+/uTmJiISqXCy8uLsrIyLBYLycnJZ7R9dCbQ6XRUVlZiNBqJjIwkJCSE7OxsrFYrCoWCcePGERAQQHFxMUajkdjYWBEEX1NTg6urK6mpqf36YLVaOXr0KMHBwajVasLDw2lrayMmJgaNRoPVamXUqFFkZ2fj7OyMh4cHSUlJeHt7nzX3xx/CkEEzhJ+F/fv3ExcX96sYNCaTicrKSqKjo3nuued48sknKS0t7adT8kPo6Ohg7dq13HHHHQ4Xp91u59ixY8TGxjpotvyWIUkSdXV1bNq0idtvv91hr18O4Ny+fTsJCQkkJiYOBWYP4Yyxe/duvv/+e84//3ycnJwYOXIkI0aMOCuuiyEM4ccwtOU0hJ+FX0rq1BuyYF1tbS3Dhg3D2dmZsLAwCgsLz0gJ1maziXTcvm8azc3NeHp60tDQ8Jvj+xgMCoWCuLg4IiIi2LdvX7/+NzU1YbFYqKmp+VFX+BCG0Btz5szhj3/8I+Hh4UydOnXImBnCOYWhLKch/OqQH8AajQY3NzfBDzIYZC6b3NxcwX+xZ88eFixYIFJLB9t6kiSJ8vJy1Go1Y8aMEWXV19dTUlJCSEgIEyZMQJIksrKyhIv/1051lrc/KisriY2NJTU1Vbh/vb29GT9+/Blvn8HpMVm8eDGPPfYYM2bMENtKCoWCSZMmDZpCOoQh/BBcXFwYPXr0/7oZQxjCfwRDHpoh/EeQm5vLo48+ekZBq/IWi06nE1kl8l7ymRD2ff3111xwwQUO21PBwcF8+eWXVFZWCir6srKyfvwjvyZCQ0P517/+JbgdjEajSFk9k6yMvvDz8yMhIYEjR46cM0F7A8FsNtPa2irm32Kx0N7eLrKMfirkLBA5RkUmG+sdMC4HkMvHDASbzYbRaHT4mEwmrFarSJn9T0EeB5PJ1O97mcjul9YvSRJ6vR61Wv2bWl82m42uri5MJtMZtUuey97zI5/Xew7l+Zd/k0n2BitTr9cPEYaeZRjy0AzhP4Jx48YRFhZ2xmmLcXFx/OlPfxL/nzt3LrW1tZx33nk/aAxoNBoqKir4y1/+4pD+6e3tzVVXXcXGjRu57LLLyM7OZsmSJYSEhPxHXOwKhYKAgAAuvfRS9u/fz/Tp08nJyWH58uX4+vr+rDoVCgXz5s3jvffeY8GCBefs1oDRaOT9999HqVTy/PPPYzKZWLduHVOnTmXatGk/yRi02WwcOnQIDw8PTpw4wYoVK5AkiWPHjuHn54e7uztTp07lxIkTmEwmuru7mTZtGmFhYf3KOnHiBNu3bycoKIijR49ywQUXUF5eztixY6mvr+e+++77NYfBAVarlddff53Y2FiWL1/u8FtRURGvvPIKa9as+UVbjhUVFeTm5lJaWspdd931qwW8/hLY7XY2btwoDM2rr776jM7ZvXs37733HldddRXd3d24uLgIHqG///3v+Pr6EhcXR1dXF7fccgsBAQFkZWXR2NjIZZddNuC19cEHH7Bw4UKSkpL+E10dwn8AQx6aIfws/FBa9U/lN+hNECd/oqOjmThxIikpKT/4IC8uLiY6OrpfdofM5dHQ0MD27duJi4sjJCQEcOQg6f3GJv890Nu3/NtAn951nn/++Zw4cYK9e/eSkpKCr6/vz64TICEhgdbWVrq6us5oLM9GyKRibW1tVFdX4+XlxaRJk5g+ffpP9mwZDAZ27txJWloaTU1N5OTk8PnnnxMXF8e0adP45ptvaG1tZdOmTWRkZBAfH8+XX3454Nj7+/vzpz/9iVmzZqFWq7n22mu54oorCA0NRalUotfrhWSA3W5Ho9H08yrI3kaz2YxWq8VsNmMymRy8kUajUQgiymUZjUbBfyJ7nDQaDZIkERcX18+rItcjHyen/VutVtRqtfh/71Tsb7/9Fi8vL+655x58fHzQaDRiHdrtdtFOWTrBaDSi1+ux2WxoNBrRXqPRiEajEfcEo9Eo6pD72NPTg16vF/+XU/T79kGpVLJ582YWLFjAZZddhsVicWi/7LXq7aFydnYmISGB7u5uli5dyk033YTFYuHee+8V9P+xsbGsWrWK8vJyvv76a+x2O3v37uWDDz4YVAvshhtuID4+XrS/r1dM7ofBYMBut4uxkTPZ5PGR29vT0yPGU/Yo9b4P2Gw2TCYTWq1WrCNJktDpdA5jJ5ffe45NJlO/8fw9YshDM4SfhV+Doba0tJQNGzYQFBQkHvwy5JuXVqtFp9MRFRXFjTfe6JDeKr95jxs3bsC2+Pr6MmzYMLy8vAQp1cmTJ/H19RVU9HV1dXh5eYk39JqaGtavX89f//pXh7ffpqYmMjMzHcp3dnZm3rx5DlwToaGh+Pv7ExkZSWhoKHA6xT0pKYmoqCgAkQorx8WUlJTw7bff8re//a3fdpinpydhYWGUl5cLg+xcg0ajISQkhIsuuoiNGzdy++239yM+kySJlpYWdDpdv/N7G7Te3t48+uijNDQ04OLiwoQJE/jmm29YtmwZLi4uaLVaSktL0Wg0uLq64u/vT2Fh4YCMq3KsiczvIscvHT16lNraWt5//30qKip47rnnBH/Jnj17eOSRRwgODgZOe4zee+892tra8Pf3p7a2lhn/j73zjo+qSv//e2bSe+8VAqkk9N6kVymiIjZWLGvZXcsWsa6iq+6q6666ihWlF5GO9BYghCQkhPTee5uZlKn39we/e74JCQi23WXn83rNCzJz77mn33Oe8zyfz7hxHDp0iJUrV6LT6UhKSsLDw4Pc3Fwef/xxdu3aha+vLxkZGQwbNozKykpSUlKEf9acOXN61YHRaOS9994T2jx6vZ4XXniB48eP09nZSX5+Pvfccw+vvfaaYA7OyclBoVAwePBgjh49SmBgIMnJyUydOpX6+nq+++47HB0dGTVqlIiMOnjwIHPnzhW0/OPGjePbb7/F29sbOzs7IiIi+Nvf/sbkyZM5deoUf/vb38jLy6O1tZULFy4ISoWGhgaOHTvG/fffz6BBg0Q5CgoKqKioICsri/79+7Nnzx769+/PqVOneOCBB3j33XfFWH7mmWd6SWEoFApsbGxYunQpX375JSUlJUITSl5EeXl5UVlZSUxMDDk5OZw5c4apU6f2SKu9vZ3XXnuNRYsWYW9vT3l5Oe3t7UyfPh0PDw/Rtv/6179wcnLiV7/6FWvWrGHu3LmkpaVRXV2NtbU1Q4YM4eOPPyY4OJjIyEj27NnDu+++S0VFBa2trUybNk307w0bNpCSkoKXlxelpaW8/fbbHD9+HDs7OyGfMXjwYHbv3k14eDinT5/mkUce4YMPPsDJyQm9Xs/zzz//sx2p/zfAYqGx4Afh+xYz17NT8PDw4LvvvuPgwYPMmjWLxYsX9/gsWrSIefPm0b9/f7799ltycnJ67ZBycnIYMGBAn+nLu50hQ4agUChobm5m69at1NbWinRaW1tpb28X94SFhaFSqXrl39/fn/nz5/f4zJkzBxcXlx7XyeRtkZGRKBQKampq+Pbbb6mvrxdpNjQ09CB2i4iIuOrRnMzlkZmZedPuvurq6vD19WX69OmkpKSQl5eHj49Prz6m1WoF+Z38aWlp6RG9plAosLKywmw2Y2NjQ2NjIzqdrkcYf/edtqzf1Bf6WrTLfwcGBvLII48IkdadO3eiUqmwtrbuQXCnUqlwc3PD2dmZX/3qV5SUlDBz5kwiIiLIzMzkk08+ISEhgTlz5pCdnc327duFdtWQIUMA2LhxIy0tLfj5+ZGfn99nX7GyssLR0ZF+/frx0ksvUVpayoYNG0hNTcXJyYmWlhbs7OxobW3lvvvuY/ny5YSGhhIfH49GoyEpKYlbbrmFqVOn8v777+Ph4YHZbOaVV15h5MiRaDQabrvtNtzc3PDz82PJkiUkJiZiZWVFQkICtra2JCYm4u3tjVarZenSpYLDZcuWLUyZMoV7770XDw8PvvzyS0wmEz4+Pj0YlBUKBZGRkXh5eTFy5Ei+/fZbPDw8mDp1KiqVimPHjmEwGBg/fjy//e1vrzkH2dnZYWtrK6wvaWlpbN26lbvuuovZs2dz/Phx3N3dGTFiBF9++WUvXxlHR0esrKzo6OigsrKSAwcOEB8f32OTo1KpmDt3LufPn6exsZGoqCgOHjxIdXU13t7elJaW4uHhgUaj4Y9//CNLlizB19eXrKwsampqxLwkl13WfVq5ciW2trbs27ePb775hltuuYUFCxbw4Ycf8sUXX+Dr68u0adMwGo2cOHEChUJBfHw8zz777A/y1buZ8L+7lPuFIHOsyIJjzs7O+Pn5CcGxwsJC7OzsBHW7JEmo1WpKS0sJDAzEy8sLSZKora2lqqoKNzc3bG1thZpvYWGhmLxdXFyEA2pXVxelpaWYzWahMqtQKOjfv//P3unlyKOmpiYuXbpEWFiY2LF2h7e3N++++y633347W7Zs4Te/+U2voyofHx8iIyOZOXMmFy5c6LGbk51J+/J/0Ol0HD16FIVCQWtrKx4eHjg5OeHm5sbIkSOpqqoiMzNTHHHk5uYyatSoqzJ7qlSq76239vZ2Dhw4gL29PS0tLdjb2+Pk5ERwcDCxsbEUFhaSn58vCABl3aHvi7gKDAzsZR36JSDv9G1tbYWCr2xV6ujooL6+Hl9fX+zt7YW6741a7WQytNDQUOzs7Bg6dChbtmzhpZde6nWtfOTRHVc6jtfV1XHixAmWLFmCn58fmzZtwt3dXTiEKhQKfH19hZK1Xq/Hzc1NjD05ze+DjY0NSqUSlUpFa2srANOmTWPatGniKKH7EapKpRIU9EqlEisrK/R6vTi6UqlU+Pn5UV5eLjR7VCoVZrOZmpoaJk+ezLBhw9BqtX2y3MrHvNbW1tja2grhQS8vL+bMmcO0adPQ6XTi9+59rqOjg8bGRsxmMz4+PsIhX14UyH1fTl/+Wz4OunTpEtHR0aLM8u8qlYr29nZBmRAUFERTUxNNTU2MHz+eOXPmiKO3vupcHkMAQUFBdHR0COmCK0kWu0MOMrCzsyMsLAyAESNGsGLFChQKBS0tLbS1tWFtbc2gQYPYunUrOTk5vZS05f+PGzeO2tpa/vSnP/Huu+8SEREhfh84cCC2trbs2rWL2267jcTERMaMGcPUqVOZM2cOjY2N2NjYiHq85557+Mc//sG8efOEpUdOS7YuyUdker2ehoYGTCYT7u7uWFlZ0djYKPpHUFAQ7e3tKJVK7OzsRJterT7/F2Cx0PzMUCgU7Nu3jyNHjmBtbc2RI0f44osvxG8ZGRn8+c9/FpOyyWRi3bp1bN++XVgKcnJyOHnyJP3796esrIx9+/aJ+7/99lsSExOxtrZm3759bNiwQXTorVu3sn//fpRKJSaTia1bt171vPinRmhoKKtWrWLJkiVXdTZUKBSMHj2alStX8tZbb101mkehUODt7c2kSZN6fC+f6195XAWXdUwmT57MRx99JEzUGo0GR0dHSkpKCAgIYNq0aYwaNYrx48czdepU3NzcaG9vR6vV/qDIDzs7O+bPn8/f//53cdzU2NiIu7s7JSUl9OvXj+nTpzNs2DAm/3/1aEdHR3GsJvtLXAl/f3+qq6t/UQuNyWTi5MmTtLS0oNPp+PLLL/nwww+By1atjz/+mNdff51HH30UrVZLfn4+eXl5N5RHSZKoqKggPT0dk8mEQqEQ7NB9ObsGBQURHR3d4xMVFdWDKbmzs5OCggI0Gg3V1dUMGTKE6dOnU1BQQH19PcHBwcTExBASEkJ9fT35+flMmzaNrKwsvv76617PNJvN1NXVodFohA+LRqMRNPJarRYPDw/q6ur4xz/+webNm3uoLpvNZiHlIB95yH2so6ODxYsXc+jQISorK3FwcODOO++kvr6evXv3UlBQQHV1NaNGjeKtt95i06ZNnDp1SvhtXOl3Jb/ICwoK8PHx4bbbbmPv3r2sWbOGHTt2oFarhbSEvBhpaWkhIiICV1dX0tPTuXTpEkuWLKG9vR21Wo1er0er1QrfG/l7OQ/JyclkZmZSXl5OTU0NtbW14jpZOykkJITnn3+e9evX09TUJOaGzZs3U1BQ0CP/ct1qtVrmz59PSkoKlZWVNDU1MWXKFLRardCO6n5fY2MjWq2WnJwcjh8/zieffMKLL76Ig4MDLS0tNDQ0YDQaMZlM7Nq1i6ioKEaPHs2ECRMYPXo0n332WQ+rqdFoFG1+7tw5xo4dy5w5c3opaqtUKhYuXEhNTQ3e3t6MGzeOf/zjH2zYsIFDhw6JsnR0dAhLioODA2FhYX0uOqqqqigtLUWv1zN37lwGDRpEYmIiWVlZzJo1i2XLlnHu3Dmqqqpoa2vjlltuQaPRCN2qt99++6oWx/8FWJiCf2aYzWaeeeYZ7rvvPgYPHsyRI0fYvn07H3zwAQqFgsOHD7N582b+/ve/4+TkRGlpKatXr2bOnDlMnDgRSZJ46623mDZtGsOGDaO+vp7GxkZiY2Mxm8088cQT/OY3vyEqKoo9e/Zw4sQJ/va3vyFJEk888QQPP/wwERERKJVK8vLyiIuL65Nm/UbxU0ofdHZ28uCDD5Kfn8+uXbvw8/O7rh1GZWUlv/rVr9ixY0cvp+C+YDQaKS4uxt/fHycnJxQKBTk5OTg7OxMUFCQmxsrKSry8vH4SZmGdTicIA+3s7FAoFCQnJxMZGYmrq6uwvtXW1vaicpeRlZXF888/z7Zt236R83FJkjh9+jRtbW3CX+PQoUOcOHGC1157jaKiInGMct999/Hiiy8SFRXFRx99xLJly/q0xl3tOSUlJTQ3NzNgwAAhjtjW1tan8N71pllRUUF1dTVWVlYMHjwYuOynpNfrCQ8Px93dndbWVoqLi7G2tiYmJoa2tjbq6uqIjo7u8VxZfLK1tZXw8HD8/PwoKSmhpaWF0NBQioqKCAsLw2w2c/HiRSIjIwkNDRVpmEwmcUzk4+NDcXEx4eHh1NXVYWtrS79+/QQFflhYGG5ublRWVpKXl4efnx9ubm74+Phw4cIFDAYDw4YNEyrUAwYMwMPDQzzr448/pq2tjWnTphERESFUosvKyhgyZAhKpZL8/HxCQkJwd3cnOzsbe3t7BgwYIKj43dzcCA8Pp6ysjObmZqKjo9FoNFRUVIgNlYuLC9bW1jQ0NBAWFkZmZib9+/envLyckJAQampq6N+/P6WlpXh7e+Pu7s758+cJDAxk4MCBtLe3k5ycjLe3t1AOl9uupqaG8vJy+vXrJ45tmpubCQwMFLpfbm5u9O/fX1hzzWYzhYWFNDQ0YG9vj52dHUFBQTg7O9Pe3k52drbg37G2tubSpUu4uroSFhYmxBw7OjqIi4sT84gsAurk5ISzszNNTU2iva601Mr6UV5eXpjNZjIyMlCr1QwfPpy2tjYhUeHu7o7ZbGbjxo0sXry416J9//79HDx4kLvvvlsEMXR0dJCbm4u9vT0RERFYW1uL/hcUFISbmxtZWVnY29sTFhZGVlbWDXNe3VSQLPhZodFopEWLFknp6enSyZMnpQ8//FAqKSmRzGaz1NnZKZ04cUJ6+umnpfLycqmtrU3KysqSHnroIamurk6SJEkym83Spk2bpMmTJ0t/+tOfpKysLMlkMkmSJEmtra3SwoULpczMTOn48ePSBx98IFVUVEhms1lqa2uTZs2aJR04cEB6//33pebmZsloNEpms1mkm5WVJZ08ebLH59SpU1Jzc/P3luvYsWNScXHxT1JHZrNZKisrkwYNGiQ99NBDUldX13XdV1RUJE2fPl3q6Oj4wc82Go2iPn8pGAyGG3pmQUGBNG/evOuulx8Lk8kkPfXUUz3a99ChQ9Jzzz0nmc1m8dHpdNLTTz8tNTY2SmazWVqzZo20detW0ccs+OVgNBqlN998U3rvvfckg8Hw786OBX0gNTVVevTRR6VDhw71GiNms1nauHGj9Mwzz0g6ne7flMP/fvyPLuN+OZSXl+Pt7Y2fnx8XL17E1dVV7OCamprw8vLCx8eHmpoa2tracHZ2RqlU9vCkX7x4MQkJCRw8eJDXX3+dNWvWoFQqKS4uJigoCG9vb86dO4evry+BgYEoFAqKi4sJCQlh6NChdHV14ejoiFar7XH84+np2SeLb18igBqNpgefypXSBwaDgcTExO890howYEAvJ16FQkFwcDBvv/029913H9OmTeP222//3h26yWQSZ/vdITP09hUR858OLy8vRowY0aPssr9VX74TPweMRiOVlZVXtXrJ/iZpaWnceuutwkrg4+NDWlraL5JHC3pCkiTmzZsH8Iv1EwtuDAMHDuSJJ55g4MCBveY2SZIYNmwY0dHRGI3Gn5zJ/H8FlgXNz4y0tDRGjBiBj48PLi4upKSkYDabUSqV1NbWEhYWRlBQEMnJydx+++1cunSJgQMHCrNmamoqAwYMICoqCn9/f2pra8UL5fz584IUzMHBgYKCApF2SkoKI0eOxNPTk1mzZlFRUYHZbO6xoCkuLu4RkQGXX55jx47tYQ7t6Ohg1apVvPzyy1dVdlapVMTFxX0vkd7V7lcoFEyYMEFENV0PlEpln3w3SqWSmJiY/0qWz76cHeU6/aUc/WSH0GuxqBYVFSFJEuPGjUOr1eLk5ITBYLBoS/2bYGVlRWxs7L87GxZcA05OTsTExPT5mywUasGPg2VB8zNB+v9n+adOnWLatGkiXNTW1paMjAzc3Nw4cOAAy5YtE+foZrOZI0eOEBERgdFoxMrKirNnz9LW1kZkZCSpqaksWbIElUpFWVkZZ8+eZfHixZjNZjo7O7GxsRFpnzp1ilmzZpGRkUF9fT2pqak89dRTPfI4dOhQEhISeuW9+0tVkv5PliA7O5uRI0f2WV6lUomXl1efL0HZmnOtF7LstHzHHXcwdOjQ63p5W1tbCxKwvvLyQyD9f/4b+L/IJpPJhMlkElEtv3QEgdFo7NMS9XPBysqKESNGUFVVhZ+fHyaTSXDAdHZ2Ul9fz4svvoiHhwfr169nwYIFTJs2jerqasaMGfOL5NECCyyw4EpYnIJ/JkiSRFtbG+3t7VhZWeHt7Y1Op6O4uFgcC2k0GlxcXMSxj8y4qVKp8Pb2RqlU0tzcjF6vp7GxEV9fXxE909raSkdHB9bW1nh5edHV1UVJSQlBQUGYzWYRYQCXTdAODg4/yNFSLkdSUhJpaWmsXLkShULByZMnCQkJEWGRknRZnPHw4cOEhYWxZ88epk+fjkajwdPTkzlz5lz12ZIksW/fPtra2rjzzjv7DI/u3k3ldOrq6rjnnnvYvn17n5FOPwRdXV1s3ryZxMREXnnlFQICAgQB4LJlywQh3y+JjIwMERXyS/FMNDQ0cOLECRYsWIBSqaSxsRGj0YiXlxdGo5GmpiZxra+vL1qtln379rF06dKfxOncAgsssOBGYbHQ/EyQtX2685rY29v3MAtfGc5sbW3d68UsR4z4+/v3+N7d3R13d3fxt4ODQ4+0u//2YyCXY8SIEaxbt46Ojo6r+lbU1dVx77330tnZyddff82sWbPo7OykqKjoqunLR2elpaU89NBDV31hm0wmLl68yODBg8WCwsHBQdCS/1QLGltbW+bMmcOmTZuwtrbGYDBQV1fH448//rPpQH0fWlpaxAL3l4KXlxdjx46ltraW4OBgfH19xW+2trY9+oAkSdTX17N48eL/3egKCyyw4N8OCw+NBdcFd3d3/Pz8erH1dsegQYNwcnIiNzeX8PBw7OzsepwbX3mfJEkUFhZy9OhR7r//fnGE1P1jMpmor6/nn//8J+Xl5T0WFHZ2dri6uvbihvgxkBk74+LiOH78OKmpqQwcOPC6FjNynn9qVFdXExwc/JOney0oFAr8/f0Fh8/3ISoqCkdHx/9ZQi8LLLDg3w/LdsqC64JCoWDq1KkcOnSIYcOG9fm7m5sbkiRx7tw5Ro4cKWjo8/Pz0el0hIaG4unpKV56TU1NPPXUU/j6+rJq1aoe6cl+QdXV1UJr5+TJkz1emFZWVoSHh1NVVUVUVNRPWtYFCxbw17/+lb///e/4+fkBlxcW/v7+PRidXVxcBFttbm4u1tbWREZG/mR5kR1wr6ZX9XPiep9nWcRYYIEF/wmwLGgsuC4oFApGjBjB+vXrrxmabTKZyMrK4rbbbgOgoqKCCxcuMHPmTLZt28YjjzwirlUqlfz+97+/rjBTNze3HsceMoYNG8alS5eYMmXKT/ZilRdiMTEx9OvXD7jM4PnVV1/x+9//HltbW8xmM4mJiYJ4Tvr/BHktLS0/6YLGZDJRUlLC0qVLf7I0LbDAAgtuRlgWNBZcNzw8PPDx8SEnJ6fP3w0GAxUVFbS1tQlrTXV1NSqVCicnJyoqKjCZTMLPwsPDg8mTJ//g/CgUCkaOHMlbb73VI90fAznKKTExkbi4OHGEJDOUGgwGtFoter2e5uZm6uvrsbW1xdXVFS8vr5/0+Asu+8+YTKZf/MjJAgsssOC/DRYfGguuGwqFgilTpnDkyJE+fUXq6uooKiri9ttvp6ioCJPJhJOTkwh7dnR07OXYajKZaG5uvqrviU6nIzMzk8bGxj5/DwkJEZw+PxWqqqoIDQ3Fz88Pg8GAUqmkra2NsWPHApetTkVFRdTU1FBYWCieLZfhp/KjkSSJ1NRUoWZsgQUWWGDB1WGx0Fhw3VAoFIwaNYpNmzYRHx/f6/egoKBeTqQDBgygpKSEixcvMmvWrB7HQmazmfz8fE6ePMmDDz7Y68hI9k05d+4cCQkJfXLLWFlZMWPGDI4ePcq99977o4+dFAoF4eHhvTSq3N3dhcrt4MGDMZlMtLS0MHbsWBQKBTqdTlin1Gr1VQU5bwRms5ljx46xYsWKH52WBRZYYMHNDsuCxoIbgqenJ15eXuTl5V2XI66trS1z587FZDJhbW3dY8FhMpk4deoUrq6uaDQakpKSBKkdQEREBFFRUdckyVMoFEybNo1Vq1YJXp+fA1OnTkWSJGFhUigUPcLkbWxsmDdvXo9rfgwkSSIvLw9XV1f69+9vcby1wAILLPgeWI6cLLghyNFOR44cue7rVSoVNjY2vV7KVlZWGAwGxo0bh7W1Nb6+vvj7+4vPldwyVzvKcXZ2Zs6cOezdu/dnCZsGerEEK5VKwsPDxd8KheInZRLW6/Xs3LmTe++99xcj07PAAgss+G+GZUFjwQ1BPnaqqan50SJ4kiTh7u5OdXU1CoUCb2/vHh8nJye0Wi0NDQ1UVFSg1+uvmqcJEybQ2dlJXl7ez7ao6eu5PwckSeLAgQOMGDHiunlgLLDAAgv+12GRPrDghmEymfjDH/7AE088IcKafwhk4jzgqpYNs9ksrrGysrqmfEJnZyelpaVER0f/Vx/RGI1GsrKyiIuLs1hnLLDAAguuE5YFjQU3DEmSyM3Nxd/fv4e0gwUWWGCBBRb8u2BZ0FhggQUWWGCBBf/1sPjQWGCBBf9WSJKEXq+nrKyMvLw8ampq0Ol017y+pqaGM2fO0NraKjS01Go1WVlZ4l6TyURpaSlFRUVX9fcymUzk5ORQUVHxo8pgNBqprq6mqqrqqr5efUGn03HhwgVaW1t7/abX60lLS6O5ufmG8pGZmUl9ff1132OBBTcLLAsaCyyw4N8GSZKorKzkjTfeIDs7m9bWVl588cWrEinCZWLDb7/9lq1bt/L000+LUP+amhqeffZZdDodJpOJjRs3UllZyZkzZ8jIyLhqehs3biQ/P/9HlcNsNvPuu+/y8ccfYzQar/s+SZL44IMPBMfRlfjkk09uiH1akiQ+++wz6urqrvseCyy4WWDhobHAAgv+bdBoNPz+97/n0UcfZdKkSQA4ODjg4eFx1XtcXV156KGH6Ozs5Fe/+hVdXV3Y2NjQ1dVF//79cXJyIiMjg7S0NO666y60Wi1nzpxh8ODBwllcdiJvbW2lpqaGmJgYzGYzLS0tmM1mVCoVBoMBBwcHbGxs6OjowM3NjcbGRkwmEx4eHtjY2Ig8KZVKKisrefLJJ3FwcBDP0Ol0tLa24uzsjIODA5Ik0draitFoxMvLi87OTqytrXF0dKSxsREPDw/BY9TV1YXZbMbW1pa6ujp8fHwAUKvV6HQ6lEolnp6eGI1GWlpasLW1FfXg4+NDQ0MDHh4eFsdyC/5nYFnQWGCBBf8WSJLEsWPH6OrqEozLAHFxcde8z9XVFUmSyMzMZM6cOTg5OSFJEmfPnhUq79u3b2f8+PEolUqqqqpEpJz83NLSUk6cOEFQUBA6nQ5PT092796Nm5sbKSkphIWFodfrqaqq4q677mLnzp3MmjWLjRs3MnDgQEaPHt1DX6uhoYGmpiZBNilJEsXFxRw5coQxY8bwz3/+k+eff54dO3YQEhJCdnY2kZGR2NvbExoaSnJyMra2tmJRB1BYWIgkSVy4cIENGzbwzjvvUFpaSkNDA0ajkZKSElasWME333zDsGHD2L9/P4sXL8bNzY28vDza2tqYNWuWZUFjwf8MLEdOFlhgwb8NGRkZREZGYm1tLb5TKBTXDLuXJIny8nKqq6u55557UCqVmEwmUlJSGDlyJGazmZKSEgYNGoQkSaSnpzN06FCRZldXF++++y5TpkzBbDYzevRoMjMzSUtLY+zYsTQ1NZGQkMDs2bPJycnBysqKCRMmYG9vT0pKCkOGDMHf379HntLT0+nXrx8uLi7CMvPmm28yceJEQkJCyMvL48iRI2RnZ4vFW3FxMefOnaO8vJyAgABuueUWTp8+zWeffUZeXh5nz55l8ODBzJo1C1tbW8rLy9mwYQPTp0+nubmZ+Ph4Pv/8c0JDQ0lISOCJJ57gwoUL1NbWolKpmDt3bo96tcCCmx2WBY0FFljwb0NISAgNDQ2YzWZxHNPW1kZ7e3ufviiyQ/DFixeZMWOGcApubGyko6ODoKAgFAoFnp6eODs7U1lZiSRJDBs2TKRRUFCATqfDx8eHXbt2MXToUPbs2cO4ceOoqKiguLgYb29vXFxccHNzIzk5mZCQEBQKBYsXL2bbtm09rB6SJHH8+HEmTpyIQqFAr9ezb98+NBoN/fr1Iz09nREjRpCSksL48eMxGo1kZ2czYcIEcnJyCAoK4tKlSxgMBkJDQxkxYgSenp5cunSJ+fPnc+nSJUJCQigtLWXAgAHo9XqOHj1KeHg4Fy5cYODAgUiShJ2dHRcuXCA2NpaUlBSMRuMvRjJpgQX/CbhpFjRypMP3fW70+ivv/b50u6cvT9K/RJn/U3E9ebzaNddbPrmuf0x9/5L12L1c35fnX6If/Rj8mD6oUChYuHAh7u7ufPrpp+zbt4/U1FRsbGx47733yM7O7nVPV1cXf/3rX9m4cSO//e1vycjIQKfTcezYMfz8/DCZTCLds2fPcuzYMZ566ins7OxEGo6OjnR1dXHo0CG6urro6OjA19eXM2fOkJOTg7OzM21tbSgUCiIjI3F1dcVkMrFt2za8vLx6HYnV1taSm5tLXV0du3fv5q233sJsNhMVFUVOTg51dXU89thjjB49mpaWFs6cOcMdd9yBl5cX3t7ePPjggyQlJYmFS0JCAg4ODtjZ2VFbW0tOTg5PPfUUPj4+5Ofnk5SUJFi0R44cyaFDhzh79ixtbW1YW1vzwAMPUFpayrlz5268QS2w4L8YNw0PjSRJXLp0ie3btxMREYHZbCYuLo6EhAQKCgrYsmULixYtIjY2Vqgjf/TRR/j7+7NgwQLOnTtHcXExVlZWBAUFERkZydq1a/H09MTe3h4/Pz8mTpyIWq1m3bp1QntIq9USFBTElClTeilJf/rpp8ydO/dno69vamris88+Y8WKFdcUcPx3Qq1W88knn7B06dIe9dC9250/f57a2lrmz5/fow5ra2v5/PPPefTRR6/qJCofP6SlpVFVVcWvf/1rrKyu3zVMfhkfOnQIrVbL4sWLf1aWYbPZzNGjR2lsbGTevHl8+umnPProoz1euHK+4HIY7gcffMADDzzwoxW85TR/ivLJaZWUlLB9+3Z+97vf/aDjDUmSMBqN1NfXY2Njg6enJwqFgvb2duzs7Hq1pdlspqOjA0mSUCgU2NnZoVQq6ezsBMDOzg6VSoUkSWi1WmxtbXuJokqSRFtbG3Z2duj1euGsK4ubdnZ2Cufa7OxsRowYgUKhoLOzE7PZjJOTUw8BUoPB0CvM3M7ODqPRiE6nE9fLz7CxscHOzg5JkjAYDNjY2KDVarG3txfllZ2W9Xo9zs7OKJVKzGYzbW1tODs709XVhaOjo/jOyckJGxsbdDodtra2dHR0YGNjc012bQssuNlw01hoAIKDg9m/fz+jRo1i2LBh/OY3v6GqqoqQkBDOnj3Le++9h8lkEg6Fq1evxsrKiq6uLj744AMWLFjA6NGjKSwsxMfHh9TUVHx9fZk+fTrvvvsuiYmJeHh4UFxcjEKhYN68ecyfP7/PiVyhUDB79my8vLyuac35MZYiJycn0tPTxWR+I+l+3/U3mt7VPvb29uTk5KDRaHp8bzKZOHz4MCaTicjISEaNGtXrOW5ubqSnp9PV1XXNvK1Zs4bIyEhuvfVW8eK4nnKbTCaOHj2KXq8X/B1yHq7MS1/4IW0mv1wyMjKwt7dn4cKF2NjY9Lq+pKSE7OxsVCoVCxcuxNHR8Ue3aWtrK6dPn/7B/a37p6ioiNzcXBwdHTl79qywIt3o/kihUGBtbU1gYCDe3t4olUoUCgVOTk59LkyVSiVOTk44OzuLa5RKJY6Ojjg6OoqjIIVCgbOzc5+iqAqFAjc3N+zs7HBxccHKygpra2s8PDywsrLC2dmZ5ORkDh48yODBg7GyskKlUuHk5ISLi0svNXVra2ucnJx6fKysrLCzs8PV1VXIeiiVSlxdXbG3txd/29rairx2L69CocDBwQE3Nzdxv0qlwsPDA2tra7HIsbKywtPTU6RjZ2eHQqHA0dGx10LOAgtudtw0UU7ygFepVFhbW2NtbY3ZbMZoNKJUKpkyZQo7duwgPz+fyMhILl68yKBBg8S1bW1tfPHFFzz00EMsWrRITBZWVlZi56TT6cQELO8sS0pKiIuL6zVxdHR0sG/fPmbMmIFKpaKwsBC1Ws3s2bN77MbNZjPffPMNfn5+uLu7c/ToUR599FESExNRqVTY2dkRHR3NmTNnyMvLY9KkSURFRXHo0CFUKtVV+TpaWlo4fvw4er2eQYMG4eHhwaVLl2hvb2fq1Km9lKybmpo4deoUnZ2dDBs2jIEDB/YoU1tbG++//z4LFiwgICCA48ePM3bsWM6dO0d7eztjxoyhra2NM2fOoNFoWLJkCQUFBSiVSqqrq5EkiaSkJIxGI11dXXh6evL666/j7OxMR0cHbW1tLFiwgKSkJFpaWqirq+POO+/EysoKo9HIpk2biIqKIiEhoUe+ysrKOHfuHMHBwSxYsIDExEQ0Gg01NTXceuut5OXlkZ2dTUNDAw8//LAIfYXLvhSvvvoqr732GlZWVlRXV/PPf/4TR0dHli9fTmZmJvn5+ajVau6///4eYbpms5kjR45QUFBAW1sbjo6O3HPPPWzatAlHR0eMRiPDhg0jPz+flpYW7r//fkpKSigoKCA7OxtJkqitrWXXrl08+OCDNDc3k5mZSUVFBbfccgvvvfce/v7+LFu2jN27d3P77bezZcsWhgwZwtChQ9m9ezczZswgNTWVvLw8Jk+eTHx8fC8r4bFjx7CyssLR0ZFz586RlpaGm5sbx48fx8nJCZVKxbBhw8jOzqa2tpa7776b7du34+LiQlZWFnPnzmXo0KEcPnxYhDhPmTKFDz74gPDwcO644w70ej1btmyhsLCQJ598End39xscvf95GDduHECvxYsFFljwn4ubbrQaDAY+//xznnnmGV577TVCQ0MB8PLyYt68eXz11VeUlJTg6+uLo6MjcJn34h//+Afnzp1j6dKlNDQ0AJd3pQcPHuT+++/n9ttv55ZbbhHPyczMZO/evWzYsKFPVlMbGxvS09Npamri+PHjlJeXEx0d3WsHK5vLc3Nz8fb25syZM2i1WjZu3EhgYCDu7u5s2rSJlpYWHBwc+OKLL9izZw8dHR1MmTLlqscQVVVVgt/i22+/5ezZs+Tm5hIbG9vn9RUVFdjZ2WFnZ8fOnTt7/e7q6oq1tTXp6ek0Nzfj7e1NRUWF2Cnu27cPg8FAYmIid911F3l5eVRWVjJ16lS8vLwwm81s3LgRJycnAgMDCQgIwNfXlyFDhtDV1UVKSgq5ubmkpKQwffp0lEql4OE4deoUMTExvRYzcNkqFxAQwJgxY2hubmbHjh3MnDkTNzc33nvvPdRqNRcuXOCee+7pVVd+fn54e3szZMgQlEol9vb2LFu2jP3795Ofn8/GjRvx8/PjwIEDFBQU9LhXoVDQ1tZGUVERDz74IKdOnSI7O5vTp08TFBTE0KFDWbduHX5+fhw9epTU1FQ+/vhjJk+ezODBg4HLvhzHjh2jo6ODTz75hNGjRxMfH09TUxPBwcEkJCTg7+/PmTNnMJvNBAQEcPr0adrb2/Hx8WH79u09+sWVbLjt7e1s3ryZ4OBg3NzcCA0NJSIigoiICE6cOEFERATDhw/n888/x8fHh+TkZLGIA5g4cSJbtmyhsrKSnTt3MmPGDE6dOoWvry/BwcEMHjxYhFBPnz6drq4uLly40Gf/+m+DUqm0LGYssOC/DDfdiLW2tmbRokUYjUbh2Cdj6dKlnDx5kp07d4ojDrjs5xEaGsrXX3/NhAkTePnll4Vz4eTJkwkJCaGsrKxHZMOwYcO4/fbbee6557C3t+/BcwEIkzPAjBkzSEpK4oMPPuhFi949RFX+18HBgdmzZ/Pkk0+Sn5/PxYsXiYiIYNGiRbz00kucP39emMft7e37rAcfHx/a29vRarXodDomTpxIQUEBf/3rX3stwCRJws/PD7VaTUdHR58LNIVCwdKlS9m3bx8XLlwgPj6egIAAmpub6ezspKurSxCihYaGkp2dLczlDg4OKJVK7rjjDv785z+LYw9AHBcA5Obmija87777cHJywmg0sm3bNgwGQ5/hvN3rrrKykvb2dlQqFXFxcRQXF2Nvb4+XlxehoaHY2tr2Wffyi8vd3V0cQZSVlSFJErGxsXz00UeEh4f3WIwqFArs7e3x9PTEy8uLhIQEWltbcXR0xN/fX1iWYmJi+OCDD/Dw8KCmpgZ7e3scHBxQKBTCv0Or1VJUVISDgwOjR48WXCoKhQIbGxthGZo+fTqXLl0iKSmJ+Ph4Ll26xIABA1i0aBEvvvhir7pxdHRk+vTp/O53vxPHpHKacj5tbW3RaDRER0fzzjvvMHToUGxtbfHw8MDNzQ2dToeLiwt6vZ68vDzGjBmDu7u7SEuuB3d3d9zc3Po8/rTAAgss+CVw0yxo5CMgg8GAi4sLr732Gu+99x6pqakYDAaampoIDAxkwoQJmM1m3N3d0el06HQ6ysvLWbt2LQqFgnHjxuHg4IDZbEav16NQKHjhhRdISkpi69atGI1G9Ho9nZ2ddHR0YDab2bJlS6+Fipwfk8lESUkJr7/+Om5ubhQWFvbKu729PdXV1ZSXl6PVamlpacHf35+//vWvHDp0iKCgILZv3059fT3p6en079+fQ4cO0dDQIMJcr7T87Nixg+rqavHCzM/P5/nnn6dfv35kZWX1ysO2bduor69HpVKh1Wr7DJkNCQkhIiKC8vJy3Nzc2LhxI21tbSiVSrRaLV1dXRgMBsxmMwMGDODo0aPU19fT2tpKU1MTBoOBjz/+mOTkZAwGAwaDgcbGRuHDEhERwb59+zh37hwnT55Eo9GgUqn4zW9+w9tvv01FRUWfPkgGgwGj0ciAAQMEwVlNTQ0TJ07EbDYLavwroVQqMRqNPfIgp+fj40NGRgZ5eXnk5eVdVRtHLndnZyeDBg3CYDBgMpnw9vYmOzub7OxsCgoK6OzspL29XejstLW1odfrMRgMYuG7ceNGcYRkbW1NU1MTWq1WpOnu7s6QIUPIycnB29ub4OBg0S8yMjJ6WWi0Wi2hoaG8+eabfPfdd1hZWdHW1kZra2uPNMvLy0lPT6e4uJiKigpMJhNmsxmTyYTRaESlUhETE4NGo+G2224Tx7pNTU1oNBqMRqOoZ0uosAUWWPDvwk2zoAGorq7mjjvuoLa2lujoaF599VVycnLIycnBycmJ+vp6fvvb33LfffdRV1fH2LFjUalUeHl54eTkxHfffUdJSQmrVq2iubmZSZMmIUkSzs7OvP3227S2tpKXl0dcXBy2trYcOXKEAwcO4Ovr2ytKRavVMmTIEIxGIzY2Nly6dIlp06b1eeQjHx21tLQIorCGhgaam5v59a9/zSOPPIKvry/Hjh0jLi6Ou+++mwEDBnDw4EGWLFkiLBzdMXXqVIxGI97e3vTr1w9JksjKymL06NEMHz681/UzZsygq6sLf39/goOD+7TiKBQKbrnlFqZNmwbA3LlzaWtrIzw8HF9fXzQaDSNGjKC1tZXZs2czZswY9u7dy9y5c/H09KSzs5OSkhIeeeQR/P39mTdvHrm5uSiVSmJjYwkPD+eJJ57gyJEjeHh4YDKZuOWWW/Dz82PJkiVcuHCh18KxpaWFMWPGoFar8fPz48UXX+T06dMA3HfffahUKqKjo9FoNL3K7ODgwG233UZubi52dnZERUVRV1fH1KlTcXNz49lnn+XIkSMYjUaCg4P7dLCsra3l6NGj3H777bi4uDB69Gi0Wi3e3t688MILHD9+nM7OTuLj43nllVc4duwYADNnzqShoYFp06bR0dHBqlWrqKqqoqysjP79+zNv3jza29upra1l0qRJQutn2bJl3H777SiVSh566CF8fHw4duwYsbGxvRhhra2tqauro7W1lV//+tcMHz4cPz8/SkpKGD9+PG1tbbi6uvLKK69w9uxZmpqaCAkJITY2FltbW8xmMyNGjKCxsZGamhoOHz7MP//5T86ePcv8+fNpa2ujqamJadOmUVdXR0REBM7OzpYFjQUWWPBvwU0Ttv1j0FcV/JTRAVceVVzP86/My5Vp/BRpft/13a9Rq9Xs3LmTjo4Oli9fjq2t7ffec71d62r3/JA2uJH7b6QO+6qvXbt2ceHCBV566aU+j8Ou1WZXpnm97St/312T6Fp5vBpupM7XrVsnfLaamppwdnYmOjr6mvdYYIEFFvzSuKksND8U3f0Bvo92/cemf73Pv/Ke7/v7h6T5fdd3h0qlwmQyMXv2bOHT8X33fF8evu+eH4Ibuf9G6vBKmM1mHB0diYyMRKvVfm+dXvn395X7+667njz+VHUu6yWdO3cOV1dXoqKifraxYoEFFljwQ2Gx0Fhw3ejOo2LB/xautAxZYIEFFvyn4X9+QfNjjzh+yHPkhcG1nve/9gK5Vnl/ymOoa6GvY5zrRV9HSlem2xfMZvMPtnTcaL38FH39lxovfT3v+/BT5aevMv6c5f53zEHfdzx65XdX3vNDx8j3pft9+LnrSg7mUCqVwidNJow0mUyoVKofFM4vSZIIVoiKihJkrN3nvRvtd993LH21dLVaLQUFBURERODk5HRTvWNuGmK9G4XcyM3NzXz00Uc89thjV6XX/ynQ0tLCxx9/zAMPPMDJkycZMGAAQ4YM6TNPAIcOHcLKyoopU6b8bHn6T4HZbGbv3r1MnToVBweHHr/JrLnr16/n6aef7tMB+nrSP3LkCKdOnSI2Npauri4mTZpEaGgo6enpbN++nQcffJCwsDDgsr/Qe++9x9ChQ5k6dSq7d+9Gp9PR2dnJpEmTsLKyYu3atYSFhaFUKunXrx9jx46lpKSEtWvXEh4eLpzQp0yZQlRUVJ/5kiSJHTt2EBwczIgRI264XE1NTaJPBQQEXPU6uV+VlpayefNmnnrqqV4h7Fdee7XJMycnh/379/O73/3uhiQmfihMJhPffPMNFRUV3H333ezbt4+Ojg7uu+8+jEYjGzduZM6cOYSHh/9kz6yqquLzzz/nd7/7HW5ubkjSZcXuEydO8Jvf/KaXMCVcv79WX/VqNBr56quviIuLY/To0T9ZOa6ETqfjs88+Y8KECSQkJIjvzWYzx48fp7CwkIceekh8r9VqWb9+PTqdjnvuuQdPT09x/XfffUddXR2SJBEVFUVnZyd5eXkYjUY6Ojrw9vbG2toaPz8/zp49K+oSLnNeffzxx8yfP59Ro0Zd9wtVkiSKi4vZsGHDD54Lvg+tra188cUXODo68sgjj4gFwfr166mvr+fuu+/Gz8/vhtPV6/W89957PProoz3kLY4cOUJCQgJeXl5cvHiR9vZ2rK2tGTp0KFqtloyMDACioqLw9vbuUVdGo5FPPvlEsGV7e3szY8YM0tLSAAgICCA8PJyqqipKSkqQpMsirTJT9aeffspTTz31E9Tafw5uOh8aSeop5nc1GvfGxkYSExNxcnIiMzMTnU7X676+RAG7p9Wd7v3Kv69My9nZmezsbDo6Opg8eTIDBgzo9Ry9Xs93332HJEkMHz5cLHj6etbV8nc9139fnq/3+hv5XAv19fX89a9/5cyZM31e6+PjQ1paGgaDQeThevIhQ6lU4uvrS2JiInPmzMHHx4cnnngCnU5HWFgYhw8f5rPPPhNpnjlzhq+//hpvb29yc3M5fvw4S5cuZeDAgVRUVBAUFERiYiKDBg1i6tSpvPDCC2RnZxMUFMTp06eJjY1lwYIFzJkz55r9B2DChAlERkZedzt3/9vNzY2CggLa29t73dv9eRcvXqSkpARXV1eSk5NFmHVfbVpdXc358+d7pXHp0iWKiopwdnbm/PnzN9QffwzkKLV9+/bh6OhIv379+O6777C1tcXBwQGVSkVISAjANcfr9fRt+ePp6UlmZiZ6vV6k5+LiQkpKSo/0ZSbm7rIeV0v37Nmz1NfX92hHg8HAgQMHgMuL0+rq6uuag65Vru7fX5mWtbU1tbW1Ih/ydQrFZemFixcv9niGk5MT7e3t1NfX99js1dXV8dVXX3HHHXcwatQoSktL6ezsZPHixTg6OlJZWckdd9yBg4MDwcHBrF27lj179ohn7tmzh2+++aZX5OD3zTvwf3OBTA9wrXnmeuawK+Hq6irUz+XrampqOHPmDBMmTMDX17fXMzQaDSdPnmTXrl3s3bsXrVbbK12DwUBnZyc+Pj5ikVRXV8eXX35Je3s7FRUVHDt2jJEjR3Lx4kUuXbrEunXr8Pf3JzIyks8//7wXLUNnZye2trb4+fmhUCjo6uri4MGDGI1Ghg4dytq1a1Gr1axZs4b4+HgcHR3Ztm0bSqWS0NBQamtrf/Lx+u/GTWWh6ejo4MCBAyiVStRqNQsWLGDr1q0olUoaGxt56qmnsLKyQpIk1q5dS05ODkFBQUiSxIEDB8jJyeHBBx8kICCAxMRE8vPzSUhIYMKECWLgyYKQfn5+pKenc++999La2kpubi4NDQ3cd999grtk7NixxMfHc+DAAWxsbKirqxOLloEDBzJ48GBOnDghQord3d35+9//jp+fH2VlZdja2jJt2jQOHTokXjZLlizh1KlT1NXV0dLSQlhYGLfffnsPE2VhYSEFBQUUFRWxcOFC9Ho9Fy9eRK1WM3DgQMLCwvjiiy8YMGAAGRkZPP3001RXV9PU1ERLSwvz588nLS2NoqIi2traWLhwIV9++SUDBgwgNTWVJ554AkmSyM7ORqvVMmrUKL766iseeOABjh8/jpubGyNHjuTcuXMYDAZGjRrVayKQIUkSKSkp/OpXv2LdunVMmjSphzaWQvF/khYA+fn55OTk0NLSQmxsLDU1NTQ2NjJ69Gg2b97Mk08+SUpKCv369SMiIqJXGiqVChsbGzFZKZVKFixYwO7du3n00Ufx9PSkvLycfv36YWVlhaurK+fPn2fHjh3MnTu3hzla5mNRKBSYzWbxvV6vp6Ojg9LSUkGhL6OxsZGUlBS6uroYM2YMR48eJTg4mDFjxnDw4EE0Gg1lZWXMmjWLrKwsdDodVVVVDBkyhFtuuYU9e/Zga2uLvb09U6ZMEbu97Oxs6uvraWlp4dZbbxXfd3Z28pe//IWJEyeyaNEidDodW7duJT8/n6effpra2lpKS0upqalh4cKFfPzxx7S3txMaGiraTKfT8Ze//IXRo0czZ84c2tvbWbduHWVlZfz+979HrVaTlpZGTk4Od9xxB6GhoT+ZGVuhUBAVFYWbmxvnz58nODiYoqIiLl26hJ2dHXFxcdTU1HDhwgVyc3NZtmxZDxHUjo4OTpw4gVqtxtnZmfHjx3P48GGUSiUGg4GZM2eyevVqli5dSnp6OgaDgfnz54v+VllZyfnz52lra+vFZ1RdXc0rr7zCs88+y/Dhwzl//jy5ublMmzYNDw8PsrOzBZXBn//8Zx544AEWLVokrGPp6em89dZbeHp6YmVlRVZWFllZWcTHxzNv3jySk5MpKipCr9dz3333iTbV6/WsW7cOBwcHSkpKRH0XFRXx1FNPUVlZKcZIXFwcQ4cO5ejRo5hMJgoLC5k0aRIVFRWkp6eTm5vL3Xff3aeIpUJxWeblyt/s7Oyoqanh888/Z/ny5fj7++Pk5CRkZGTdq1tvvZXKykruvPNO1q1bx6JFiwR3l5eXl9CokqHVajl06BAKxWXx4Hnz5rF27Vrc3d3Jyspi3rx5DBo0CJVKhU6n45133iEhIYFx48axfft2brvtNkE0Ki8im5ubqaqqYubMmZSVlZGRkYFCcZnUcsWKFb2oDjo7O/Hy8kKv1wuKiIaGBnQ6HREREb3qSKvVsmPHDkaOHEl8fDxKpbKXlVmGJP2fD6LBYKCkpET01YsXLwqJHU9PT06fPk1aWhp33303Dg4O5OXl0d7ejouLi0jP3t6eu+++G5VKxbFjx5g4cSIvvfQSK1aswN7entbWViGn4uTkhL+/P59++in33HNPn/m7GXDTWGgkSeLbb7+lvb2duXPncuHCBU6cOEF+fj62trbcddddPYTrwsPDGThwIIGBgQCMGDGC2NhYvvvuO3bu3ElFRQWenp589dVXPUjmZDG+4cOHM2fOHN59913UajV5eXmsWLGC7777jqamJlxcXPj666/Zu3cvHR0dTJ06VQjLySR6Bw4cwGg0MnPmTHQ6nZAEGDRoEBqNhtzcXJKTk0lPT2fWrFl0dXWxdu1a2tvbxS7om2++6THRdnV1sWbNGiZOnEhCQgL19fW88847jBgxgpkzZ/LGG2+g1+tJSkoS5t7U1FR2794tGGMrKyvZtm0bfn5+HDlyhLq6Os6fP09MTAze3t6cOnWKpKQkcnJyiI+Px8nJiezsbHQ6HUqlktzcXHJzczl9+jRxcXHXPJro6OigurqaWbNmUVVVRU5OzlV3DXq9nnfffZcxY8Ywbdo0Xn/9dXx8fDhz5gwBAQFCqLO9vR1/f/9e9zc3N/Pyyy+zceNG/vnPf4rJLzQ0lFGjRrFlyxYuXbpETEyMyHN4eDh/+9vfWL16NY8//riY5EwmE1u2bOHhhx/mT3/6E3FxccDlXXtSUhL79+8Xu+/uyMzM5Pz58wwaNAhbW1saGhooKSmhsrKSXbt2MWvWLJKTk/Hz86OpqQm1Ws3ChQvZsmWLYHLu168fGzdu7LFj279/Py0tLcTExPR4nr29PYGBgQwfPlyY6GfOnIlKpSI1NZXCwkL69etHaWkpFy9eJCwsjJiYmB6aV7a2tgQFBTF8+HBx5r5w4UKam5vJysrik08+ES+ZHTt2XLWtfyisra2ZP38+W7Zs4cKFCyxatIht27aRmZlJdHQ0H3/8Mba2tmi1WmEJgMtzwjfffIOLiwvTp09Ho9Gwdu1abG1tmTdvHgcPHuTixYsUFRWhVquxs7PrYaUwmUx88MEHDBs2jHHjxvV6+Xl5eeHj48OIESPYsGGDUAhft24dSUlJZGdnEx8fj5ubG35+fowcObKHHlhQUBB+fn7i+Mff358777yTLVu2kJ+fz+7du/H392fPnj1UVVWJ+2TdMb1eL+pi7ty51NTUkJ2d3WOMvPbaayQlJXHu3DmmT59OaGgoJpOJ1atXY2tri1qtZu/evTfUHm5ubrz//vscO3aMu+++m+bm5j6FMOWyjho1CltbW06cOEFycjKjR4/u8+ht48aNKBSXRX+PHz/O+fPnKSkpwcbGhilTprBt2zZxva2tLeHh4Zw6dQqtVouTk1MPLrDa2lq+/vprpk+fTlRUFK+//jomk4ns7GzuvPNODhw4QHt7e6+yNTY2EhgYKBi0S0tLsbOzw8HBoU+ZmdTUVMLDwwWrtizX0h3yQjIgIECI5+bm5hIRESHmGbVaLfqXlZUVLS0tdHR0CP+6vhTdra2tsbOzIzk5GR8fH5Hn7r4/zc3NgsVeJk01m81C5qaqquqmstLcNAsauEydLyvjRkdHU1pair29PT4+PgQFBfUIiZUh795lqveOjg6ys7Pp378/s2bN4q9//WsPJzArKytB8Z+QkIDBYMDGxgYPDw8CAwMpKChgwIABLFy4kFdffZXU1FTc3NyEFIJCoRAv0vT0dBwcHHBycmLp0qU9VHnlVX5RUZGwCgwaNIjCwkJsbGxwdXUVJGbdO2RbWxvl5eXY2NgwYcIEIiIiKCoqwsbGBi8vL6ytrcXE6+LigrOzMzqdTogfrlu3jtLSUlQqFQkJCXz88ceEh4dja2uLm5sbLi4udHV1MXnyZEpKSnjjjTfo6urqNUHFx8djbW3NypUre5jku0OSJNLS0ujs7OTMmTNERESwbt26qw4wmZjP2tpaWA/8/f1RKpWcOHGCsLAwtm3bhoODQ5+7JA8PD5YtWyZM5DKUSiXLly/nm2++4cyZM0JrCS4fh02YMIFt27ZhNBr5/PPPgcuTw7x583Bzc6O8vFyUX6VSMWXKFO68805+97vfodPpepRn+PDh6PV6XnrpJdrb20VfcHNzw2QykZWVJeQFbGxsRJ0bjUaxg2tsbBTHbzIWL17Mzp07+fLLL/tkRpbzZ29vj6urKy4uLnR2dhIcHExxcXEPRuVrhYHD5UW9rHYtyzbExMTw+OOPs3z58j7b7sdAoVAwbdo0kpOTKS4u5oEHHuDw4cM0NjZib29PWVkZsbGxPPnkkyxdurTHvenp6Tg6OuLh4cGSJUvIycnBxsYGKysrIiIiKCsr63NegMubg5ycHFxcXLC1te21oJHrSX5BRUZGsnTpUp599lkmTZpEWVkZb7zxBp2dnVcNy+/+XA8PD1xcXDCZTGLuGjJkCJ9++mkPC6dSqcTGxgYXFxecnJyEnIatrS3Nzc29xsi5c+ewsbERemV6vZ7S0lJiY2N5+umnueOOO667LXQ6HXV1dURGRrJ+/Xri4+N58803r/lStLOzY/ny5Xz44YfU1tYKfb0rkZOTIyxCAwcOpKysDBsbGyFJ0v0IEC5LgeTm5nLs2LFeOm8yMalKpRIM4vLRmpOTE0qlspdcjXwM5O/vj5ubm1CTLywsJDY2tk9nYLVaTVlZGenp6aSnpwv3he6QLeTh4eEoFArUajVZWVkUFRVRWVlJdnY2dnZ2YuOs1+txcXHBzs5OHB3KbX7lPGo2m9m3bx8BAQEoFJdV1uVymc1mXF1dRb5NJhP29vYolUqsra3x8fEhJyen1wLsvxk31YJGVn/W6XS0trYyevRoIVVw5YCztbWlpaWFlpYWjEajkE0wGo3079+fzZs3U1lZSUpKSi8ZAKPRiFarpbS0lDFjxogVNFze0cuCfqmpqQwcOJADBw5QU1NDS0sLra2tgmY/Li6Ozz77jNTUVE6ePIlCoaCjo0NQ8RsMBoYMGUJeXh5arZbq6mpuueUWIalgMpmE1IAMV1dXOjo62LJlC2fPnqWmpobIyEjS0tJobW0lNDQUb2/vXnT1FRUV/POf/6StrQ0bGxtSU1O5dOmSoOqXnyfnvbCwkJUrVxIREUFubi62traUlJRQVVWFWq2muLiY+++/n1mzZnHu3DlSU1MpLi4W+ZQkCZ1Ox5kzZ3jwwQe57bbb+OMf/8ixY8coLy/vscuWy2lvb094eDgXL16kubmZiIgI/Pz8mD17Njt27OC3v/0tBw8eJDY2ttfZvE6nQ6/XM3DgQH7729/yxz/+kaqqKnQ6HY2NjURHRxMREYGPjw/W1tbo9Xq6uro4ffo0hw8fxsHBgWHDhuHg4CD6lLW1NatWrWLLli0cOnRI7KK0Wi1arRaNRtNjVwlQXFzMI488wsSJE0lNTRXtrFKpiIyMRKPRsHDhQiHLIMsPGI1G0U/s7e2FJUr+raysjPfeew+dTkdtbW2PZ8qMwd1lFGST+gcffIC9vT1Go5HW1lYhaaBWq3uMGfnIVKPRiD4nW6ucnZ3ZvXs3ZWVlZGZm3vC4vR4EBAQwZMgQhgwZQnh4OHFxccLKZWtry969eyktLSU7O7vHfbIWV1paGqdPn2bMmDGcPn1a7KaHDx+Ovb09JSUllJeXi/IZjUasrKzw8PDgxIkT1NbW0tbWRldXl0hbXszU19fj6+vL1q1bqays5MKFCxQUFPDss88SGRkp2LDr6up6vOxkq1Z9fb3oB3Jfl/208vLyyMjIoLm5WdwnSZIYv/L1cj/pa4yMGjWKc+fOUVlZKeQ0bGxs2Ldvn6gzuR9d6Tsly3qo1Wrq6+tZt24daWlpbN68GaVSyYgRI3B2du5xvbyIlyQJrVZLa2srU6dOxWAwMGjQIEwmkxiP3Z83btw4zp49i06no729neHDh4sxILdJ9znP2dmZOXPmcOrUKYKDg3u0e3BwMCaTicrKSmpqahgzZoxYxMjzXvd5U5IkOjs7KS8vx97enoCAAIqLi/Hx8SEpKYlBgwb12S8HDx6Mvb09kydPZsaMGUydOrWXs7K9vT2LFi3izJkzmEwmnJycmDlzpghC8Pb2Jj4+Ho1Gg8lkor6+ntGjRxMXF0dTUxOtra2EhYXh4ODArl27aGlpEWnLwsaydWrkyJHU1NQIbb24uDih61ddXc2wYcOEVE1paSnTpk3rtVD/b8ZNFbZtNBo5fvy40NKJjIxk165deHl5MXny5B6+GQ0NDWzevJkJEyaQnJzM2LFjqampoa2tjWnTprFlyxZ0Oh0LFy4kMDCwx9nn0qVLWbJkCe7u7owZM4a0tDSqq6u59dZbkaTLHvGyWd7FxYUtW7aIcL+hQ4eSnZ2No6MjEyZMEIrJS5YswdPTk7Vr1zJo0CCam5vp6Ohgzpw5pKen09LSgpubG8OGDSMxMRGNRkNsbCzHjh1j0aJFeHt7A5cHZn5+Ptu3b2fw4MFMmzYNtVrNiRMnBCmalZUVu3btYsqUKVy4cAEvLy/h/6FQKBgxYgSJiYkkJSUxceJEoqKi+Pbbbxk/fjwFBQVYW1sTEBBAU1MTKpWKESNGkJOTw9mzZ4mOjkahUBASEkJxcTE2NjYkJCSQkpKCj48P8fHxIp+lpaWcOnWK+fPn4+bmRkVFBTt37mTw4MGMGzdOmGcrKirYv38/M2bMwNHRkcTERFxdXYmJiRGimiUlJcTHx5OcnMyIESN6DFKTyURiYiIZGRnMnj2b8PBwduzYgZOTE+7u7uTk5DB//nzUajWenp40Njayb98+oqKiCAgIIDs7WywwZs+eTUNDA3v37iU2NpaJEyeSmZlJRkYG8fHxnD59WkhpNDY2Eh8fz9ChQ0X/KSwspLS0FGtra2JjY0lMTMTa2pr4+Hj+9a9/oVQq0el0zJ07l46ODlQqFcHBwSQlJTF16lR27NjB4MGDycnJYfz48SQnJzN06FA6OzuFg+eoUaN6HPNduHCB9PR0Ro0axalTp5g5cyYpKSl4eHigUCgoKCggLCyM1tZWRo0axZ49e7jtttvw9fUVL+3MzExSUlJISEggNTWV2bNnc/LkSfr160dwcDCbNm3C39+fW2+99WcJBZUkiZycHIKDg3FyciI3N5egoCCcnJyoqKhg06ZNhISEMH/+/B4vlPb2djZu3IjBYGDJkiW4urpy5MgRrKys8PX1JS4ujry8PA4fPkxcXBx6vZ6YmBj279/PpEmTsLW1ZefOnYSEhGAwGATJoJynPXv2AJePVTZt2oSDgwMLFy6krq6O+vp68dJPTk6mrq6OhQsXCh8ao9HI+vXrCQ8Pp6GhAWdnZ4KDgzl16hSLFy8mNTWVCxcuMGXKFIYPHy522jqdjj179uDh4YGvry+nT59m1qxZnDhxgoEDBxIaGsrp06fFGPH29mbfvn3U1NTg7u5OREQEHh4ebNmyhZCQEObNm8elS5fIy8tj0aJFwk9Dq9Wya9cuNBoNfn5+tLe3Y2Njw9ixY0lMTMTGxob29nZmzJiBl5cX7e3t7N+/H7Vazbx58/D09OTEiRM0NTUxd+5c4ZuWnZ3NiRMnGD9+PEOHDhXl0uv1HD16VDjxR0ZGsnPnToKCgrC2tiY/P5/hw4dz4sQJZs6cKURwKysrmTFjRq9NTEFBAdnZ2bi6ujJs2DBycnIoKipi3LhxfPfdd8yZM0cshGRrcU5ODlOnTkWn0+Hs7ExdXR3Jycn0799fSOVc2S+zsrLIyMjA2toaGxsbpk6dKhZ5MrRaLS+//DJvvfWWGJtFRUViDMnvEXmhNW7cOFpbW7lw4QIKhYLIyEgCAwPZtm0b0dHRYoHV2dnJl19+yYoVK7C1taWjo4PTp09jbW2Nt7c3MTExlJSUUFZWJvwZXVxcaG1t5Y033uDNN9+8qVTlb6oFDVw/T8H3XXe133U6HcuWLePdd98lJCTkmtwA8r19fXe167/vmh/KN3Kj5b2R628EP8WL7lp5/Smfc+Wzfop0r9Y2GzZsoK2tjalTp9Lc3Iybm1sPeYG+8nJlGlfL4/X+drW0f0z6PyXk53Rv8xvtq31dfyPj83rq71rPuJFx/WPq9XrGyE+R9g9N43rSvlaaer2erVu30tTUxH333SdCwn9IWj8WcuSUbKWXj/e6o7OzkxdffJHXXnutl+7f1fLbVz3n5+fj7e0tos6uZ67vK526ujr++c9/8tprr1kWNP/LqK+vZ8+ePQwbNoxBgwbdVJ3Bgn8PJEmira2NQ4cOYTabGTp0KP3797f0LQssuAqMRiM7d+4kOjpaWIX/kyFJEseOHUOhUDBu3LgezuE3kobsAP1jyqvRaDh+/Djh4eG9juf/22FZ0FhgwX8I+trFW2CBBX3jv228yH5FV3O6/1/Lx8+B//ktoOx02tHRIUiSruX1La+Sr/SQv5lgMpnQarW9HHM7Ozuv66hJNsHKdfrvgJwHmXTuWtd1b/+fAkajka6urh4fvV6PTqfrMwJJxn/6BCNJUg9+jv+UZ0uSRGtrK3V1dcKxVa1WU1dX973tf61ndSdvM5vNPSLWJEkSjqo/tt+YzWa0Wm0PJ1k5/e5/d3Z20tXVddXnyfd0d1z+IZDLf2W0ztWuvdZY/yF9RpIkOjo6rmuONRgMP1t/vNF57/sgR7D+XGNcbv/u7SZJEu3t7T3q8ufOx78T//MLGoDExERefPFFDAYD7733Xg8vchnyGakkSaxevZqKiop/Q05/GVRVVfHRRx/1iKY4cOAAf/3rX7/3XjnyIjs7m6effvqaL/Afgisn+mshKytLMAJfC0ePHmXVqlW90ryRZ3XHhg0b+Oijj1i1ahXPPPMMa9as4a9//SvPPfccFy5cuKG0/pNgMBh46623OHLkyC/+bL1ez1/+8heOHTvW5++VlZU8+uijnD59GoDy8nK+/PJL2trabug5kiSRlJREWloaq1evpqSkBL1ez549e0hPT2ffvn0YjUZycnI4ceIEBw8eFLTyPwRms5kdO3bw1FNPsWLFCmprazGbzaSmpvLcc8+J/nf27Fl27tzJtm3byMjI6PN5ZrOZL7/8krVr1/6gvMjo6urizTff7JNHqS9kZ2fzxBNP9LmQMhgMvPHGGxw/fvy6n19fX8/jjz/eg3unL+j1et544w2OHj163WnfKPbs2cPf/va3ny39nxJms5kPP/yQb7/9tkf/+PTTTyktLf33ZewXxE2zoJF3Cl1dXWi1WrFjkFes7e3tmM1mNBqN+E3eCXh4eNDc3IyVlVUP/Zauri5hsTl9+jRHjx7FbDbzyCOPEBgYiCRdDgdua2sTOzWdTodOp+sV9nplXvV6PW1tbZhMph5/yxOYvPNQq9XiGqPRKK6R2WjlEHI5rFoOG5atLFfb0cn36fV62tvb6ezsFDosgYGBPPzww5jNZtatW0dJSQne3t40NzeLa6+2Gztw4ABJSUni+s7OTmHtka1b19o1yzuK7vfI9Wk2mykqKmLDhg2iTU0mE+3t7WJ329nZKXhvgoODxf+v9iy5/ZuamgDEs0wmE2q1mo8++kj0ne796loYOHAgjz32GOHh4Tg6OvLQQw9xyy23YG1tTUdHR492l9v0ajt9ufzy7txkMtHW1ibKK98v/20wGER4dl/9U94ty2W8Wt+U+4b8r16vx2w24+joiFarFWnK18h98kZ3s1fWq9FoRKfTifBpeZes0+mws7Ojo6OjVxoKhQInJycWL17Mli1bMJlMeHh4cOuttwp+juuFyWRi//79hIeHY2VlxYEDB0hOTkatVjNy5EgKCgrIyspi7dq1jBw5ksGDB/P111/3KLNcDrVa3We7dEdLSwvBwcG8//77BAcHc+DAARQKBd7e3hQXF4s2/PTTT5kyZQrjx4/nww8/7EUlIVtJXF1daWtrE31bnr/keVG2OOj1ekwmExqNplf4tJ2dHZ6enmIMGgyGHnOTPI5liv/g4GDRn9rb23vMOdbW1jg6OqLRaHr1w+5/m81m1Go1Op0OLy8vzGazmI/6GnNyn7C3txe/y+Hl8qZKniM6OzsxGAzityuvleuno6ND9F95HpLfDXB5oSfPBd3bWr6/ex13L5tcf93HrDxPX2tOlN9B8jPlNu1eDpl+QK1Wo1QqcXNzQ61Wi3YFePDBBwkODhZl+ndazn9u3FTSB4cOHWLLli1ER0eTmprKqlWr+PLLL3FyckKn0zF+/Hh0Oh0HDx7kqaeeoqamhqqqKtra2gQfxEsvvcSLL74omHpra2vx9vbm7NmzuLm5ERUVxXvvvcf999+Pi4sLx44dIyAggNTUVB544AH+/Oc/M3LkSJKTk3nggQf6FJurra3l+PHjqFQqamtrufPOO9mxYwcRERGcOHGCxx57jM8++wwbGxtaW1sZMGAAS5YsESHoiYmJ3HXXXfzrX/9i1apVvPbaa9xzzz2kpaWRn5+P2Wxm1KhR+Pj4UFFRwZIlS3qJEba2tvLkk0/y5z//ma+++oohQ4Ywfvx49u7dS2RkJFu2bOHFF19kw4YNuLq64ufnR0VFBV9//TUXL17knXfe6cW3YDAY2LJlCwkJCQQFBdHa2srWrVs5deoUr7/+OlVVVTQ2NnL06FGWL19OXFxcr5fNyZMnqa2tpaKigrlz52Jra0tKSop4kRmNRhITExk5ciTvvfceL7zwAvv378fOzo4RI0Zw6tQpVCoV/fr1Y+jQoVftK5IkcerUKZqamqitrcVoNFJXV8fmzZsJDg7GYDAQERHBzp07BYmgUqkkIyODX/3qV4Jhui+MGjWqx99KpZIxY8awe/duEhMT2b9/P1FRUdxxxx0cOHCAzs5OCgoKeO6553o5C1ZXV/PCCy8QHR2Nl5cXHh4eGAwGsYP/4osvGDx4MElJSTz55JNs3ryZyMhI2traiI2NZf/+/YSEhHDu3DkeffRR3nnnHQICAigrK2Ps2LEsXry4VxtkZ2fz/vvv8+qrr/LCCy/w+9//Hp1OJxZ9KSkpXLp0CVdXVxYsWMALL7zAO++8Q11dHbW1tcyaNeuqdXNlG+zevVvU63333cf+/fuprKzE3t4elUrFk08+yaZNmwgLCyMrK0voXl2ZTl1dHTNnzuTAgQMUFxfT3t4uNJ66X3fx4kUqKyvFd3Z2dowfP16MD5VKxQsvvEBdXR01NTXce++9HDhwQOgO2dvbC5oGW1tbITmg1+tF9IrJZOJf//qXCON/+umn2bp1K1FRUTQ0NDBnzhxR5+7u7gwbNgyFQkFAQIAgnbOzsxNO4Q0NDTQ0NODq6oqdnR2VlZW0trYKqgaz2cy3336Li4sLaWlpeHl5iTB7vV5PfHw8nZ2dbN26lb///e9CeuXSpUuEh4dTVlbGkiVLRJ107w9NTU188803Ym76zW9+Q3JyMpIkUVlZiY+PD1OnTkWhUNDc3Mzu3buZP39+L16YEydOcOLECVxcXHjmmWf44osvGDRoEAUFBTz00EPs3LkTT09P9u/fz8qVK4HLm4v169czatSoHnOFwWBg48aNhISEcOnSJUEempWVRWZmJvHx8ZhMJk6ePElsbCznzp1j9uzZHDt2jHvvvRc3NzeOHj2Kra0twcHBODs7s3r1asaOHUtycjLvvPMOSUlJdHZ2UlhYKLhhDh8+jJubGwMGDGDgwIE9xsvf/vY3Zs2axYEDB5g9eza5ubmEhoaycOFCNm3aRGRkJMeOHWPFihVs3bpVbDZnzZpFe3u7mBO7893Iz3R1dWXgwIHU19dTU1MjFjG33HILL730kuBWevDBB4HLMgoy+eaTTz7Jq6++yrJly1AoFEIyZ968eT1kFG4W3DQWGgA/Pz8xCQ4aNIhjx46hUqmIi4vj8ccfZ/PmzQCCe2TNmjVMmTKFCRMmYGtri6enpyCG+uqrrxg3bhxLly4lLi6OkJAQYmJixCTZ1dUl9I1mzpxJTU0NFy9eFAunKVOmkJyc3CuPknSZ4js6Opr58+czfvx4vvnmG9zc3JgyZQo2NjZCfC8sLIz77ruP06dPc+jQIezt7Zk2bRoLFizAw8OD+vp67O3txY5HZoB97bXXqKmp4dSpU4wePbpP6QFPT0/69etHUVERYWFhpKWlUVtby7BhwwgODqa+vh5nZ2f8/PwYMmQINjY2+Pn58eCDD9Le3k5jY2OvNK2trQkKChJyCE5OTixbtozg4GAyMzNZs2YNJpMJX19f8vLy+mxDd3d3oqKi0Gq1ZGZm8vXXXzNixAgWL17M0KFDCQ0NJSwsjAEDBghdGFdXV5qbm7G3txd6KklJSdfsK52dnXz99ddMnTqVCRMmYGVlhZWVFYMHD8bBwYFTp07h7++Pr68v0dHR+Pn50b9/f+rq6igqKrpm2n35wsjfjRkzhuXLl3PmzBnOnj3LhQsXcHZ2prW1VWh6dYeXlxdqtZqlS5cSExNDYmIi9vb2dHV10dzcLHhGzp07h1qt5ujRozQ2NpKQkMDatWsJDg5m+vTpqNVqTp8+jUqlEosp+XjmynxGRUUJq4yDgwP5+fk0NjYKBuXBgwfz8MMPc/78eYKCgggODiYjI4Pq6uoenDvXg+71WlpaiqurK25ubjzwwAOkpaWRnp5OSUkJEydOFPISV6K7pW3+/Pls27YNtVrdJ1W9fHQsf660dCgUCqysrDCZTDg6OooXiLy4UCgUgtBQhkxs1x3R0dH4+PiQkpKCWq3m+PHjon66Q/ZlaG5uxtbWlrFjx/aqP3mnr1QqBTdT9+dVV1dz7NgxpkyZwsiRI1EoFIJUzWw2c/78eUaPHk1rayvV1dV0dXXRv39/zp07R3Z29jUVvrdv346TkxNTpkzBwcGBr7/+mvXr1zN58mQWLVrEp59+ilqtRq1Ws23bNu655x6CgoIoKSnh0qVLYhE8duxY/vKXv3DhwgVKS0uJj4/H1dWVxMRE0tPTqaqq4pZbbuHuu+/GxsYGo9HIhg0bmDFjRq9o0qysLKFJNWjQICRJ4uuvv0ar1eLv709+fj4+Pj5IksR9991Ha2srw4cPZ+LEiZw7dw47OzsSEhJQqVScPXsWHx8furq6uOeee9Dr9RQXF7N582amTZsmuGfa2trYu3cvvr6+vdS2/fz80Gg0LFiwgMDAQFxdXbnrrrs4ffo0Bw4cwGQyMXnyZAIDA9myZQsODg4EBQXx3HPPsWnTpqvOifIz/fz88PDwYPXq1UycOJGFCxeyY8cOjEYj7e3tLF26lMcee4y1a9diNpuJjo7m17/+NZcuXcLa2hqlUklXVxcVFRUcOHCAIUOGoFKpyM3NJSsrq08JiP9W3DQLGvmFIYujBQcHiwlAfulrtVomT57MCy+8QFxcnNipdXeQUiguCw2WlZXR1dWFu7t7r52ePLhk1k2FQoG/v7+gOLeyssLa2rrXZCmjvLwcrVaLnZ0dUVFRtLS0iLP+oKAgoeEhyziYzWbBNKxUKvu0bMj5srOzw8bGhjvuuAMrKyuef/75q/oRLFy4kG+++QaVSkVFRQXnz59nwIABV30hyXUr5+n7IOff2tpavIDHjx/P008/zdSpU3tdL0kSVVVVlJaW4uHhIRhwOzo6cHZ2pn///n0+RzafNjU1kZOTg4eHx/fmr6uri+rqauD/2l+j0XDx4sVeL0JJkiguLqaurg5XV9deCr43Apl232w209DQgLe3N3PnzuUvf/mLkEHoDrkfyDotzs7OzJo1izfffBNHR0cuXrwoaNGtra155pln+Oabb/jqq69oampCo9GgUCgIDAwU/Uqml7+a06WtrS1Tp07lq6++IiEhgb179wrSwe5jTJIuR0rcfffdrF+/Ho1GI6wG14uSkhJqa2txc3MTxIByv4HLL2vZOnc1RlNZdVipVDJt2jTOnTvXQ9OmOxobGyktLRWfioqKHvWg1WpZv349/v7+JCQk8Omnn+Lh4SGOUXQ6Hf7+/tjZ2YnjAwcHhx6knUajkYsXLyJJEnZ2dqhUKp555hn27t3L559/3ut4qqOjg5SUFO64444+eUPc3d1xcHAQO3NnZ+ceu2uZ8RzoYdXJz88XY8HBwYFbb72VDz/8UOj+PPHEE2RmZvLOO+9c1dettbUVtVoNXJ6b2traqK+vx2g0Cnp+k8mESqUS1P9ms5n09HROnTolxpisji7PbxkZGcIK19jYKBTAY2JiBAlpbW0tx44d61UftbW1PfqEJEnU1tYyZMgQli9fLiwVsqhm93lLr9fT3NxMdna2qBu5z8nvC41GQ11dXY8+FxwczL333surr74qLFTwf+8dOc/yxkilUmEwGGhtbaW1tRVAjEF5nlapVL3mxO5l7f7MpKQkGhsb6ezsxM7ODjc3NwwGg3iev7+/sDLa2Nj00HOS5/NJkyYRFxfHn/70J0pKSkhKSiIxMfGG/cz+k3HTLGhk1NfXU1ZWRmVlJbfccgtarZaWlhZcXFxwcHDgzTffZNOmTbS2thIdHc2aNWtITU0VZl35PH/s2LE8//zzfP3114Jt8tKlS1RVVaHRaFCr1dx2220cP35cOK+NGDFCUNK3tbX1Om+VMX36dN544w3WrFnDmTNnmDt3LmlpaVRWVtLY2MiMGTPEuaosSDhmzBg2btzI6tWr2b9/v1gwHTp0iNraWkHNLp/dnjx5krvvvpuYmJgeukUyFAoFsbGx6HQ6hg0bxqhRo7CxscHOzg61Wk17ezs6nQ4HBwdSUlIEdb7s39I9Cqo73N3dSU9Pp6Kigvb2dlEGk8lESEgIr776Kps3b6awsLDXvZIkcfDgQcrLy6mtraWsrIzx48fz0ksv8dVXXwlrRllZGeXl5Xh6enLw4EHy8vJobGzk5MmT5OfnU15eTmVlJfX19Vc9g5cXSF9//TVpaWk0NDSQlJREeno6paWl1NXV0dHRQXt7OxcvXmT37t1UVFTQ2NhIYWEhp0+f5ttvv71qPzQYDNTX19Pc3CwE62TfE7l+Bw0axO7du/niiy/EjutKyGfoarWamJgYzpw5wyeffMK2bduoqqri8OHDlJaW0traSk5ODjk5Obzwwgt0dnaycOFCTp8+TVVVFZ2dnUyYMEG0n9yv+lrUKBQKZsyYQXZ2NosWLcLOzk4ol8t9X6PRCN+l2NhY3NzcxLHM9cJkMrF3714qKipoaGigsLCQtrY2Ojo6xDiMiooiJyeHgwcPUlxcTGVlZY886/V6Tp06JXzW3N3dmTRpkmBBvhKxsbHMmDFDfCZNmtTjmM9oNJKXlyeECYcNG8aECRN69P/x48czbNgwysvLKSwsZOLEiT0WNC0tLaIft7a2CvmC5557TvgiyZBVo0+ePMnq1atZv349ZrOZlpYW4e/g4uLChAkTSE9PJzU1lTlz5vRY/IaGhqLT6fj222/Jycmhurqaw4cPU1hYSGVlJeXl5XR1dTFnzhwqKiqIiYnBZDJx5swZnnnmGezt7XvUqew/ptFomD17thjPdXV1LFu2jDFjxnD8+HGys7OZPHmyeIE/8cQTvP3224Jx+NFHH2XQoEE4OTlRV1dHQUEBkZGRKJVKEhMTKSsro6mpCV9fX9LS0njvvff45ptv0Gg06HQ6li9fznfffcfhw4d75G/QoEHk5+eL48WqqipGjBjBqlWr2LRpk1jQdnZ2ivlYo9GI9jt9+jR5eXmUl5dTVVVFfX09nZ2dYrzb2tri6+vL+vXrSU9Pp76+nkuXLuHg4MDjjz9OQ0ODyIvslyJLRMhzv1arpbOzk4kTJ1JUVERZWRllZWXceuutqNVqWltbsbGxITQ09KpzYlFREQ4ODjz22GNoNBrmzJnDwYMHKSoqIjY2luDgYNrb2ykpKeH8+fMsXLhQzBXd52rZf+ns2bNMnDiR6dOnYzQaWb58OY888ggBAQHXPWb/03FT8dBcvHiRt99+m9/+9rcEBATg6ekphL8GDhyIWq0mJSWFkJAQBgwYQHt7OykpKXh6emJjY4OPjw+FhYX4+/vj7e1NSkoKjo6ODBo0SLzYYmJiKC0txc3NjbCwMIqKisR5vYODA5cuXcLf3x+tVovRaCQ6OrrXTtFkMpGeni4WEzY2NmJwBwYG4uXlJUT0HBwcqKurIyYmhvLycmpqaoTqcX5+Pm1tbXh4eODq6kpDQwN6vZ6oqCgRsurg4EBoaOhVmUhramrw8fFBq9WiUqlwcnKiqqqK6upqoqKiaG5upqmpCXd3d1paWkSZg4OD8fPz68V82tTURF5eHqGhoVRVVdG/f38qKipwdnbG19eX8+fPC0ruvmjEa2pqKCwsJDQ0lPr6egYNGkR6ejoqlYrBgwdjNptJTk4mNjYWg8EgzqqVSiVOTk5kZGQQERFBRUUFISEhVFVVCar3K/OqVqtJTU3F29sbGxsb/P39SUtLo1+/flRUVJCQkEBubq5wUqytrcXX1xetVou3tzcdHR3079+/z7rt7OwkJycHo9FIZGQk9vb2ZGVl4ezsLHSRYmNjxQtnyJAhfb6E5RdiYGAggYGBlJaWUlhYSHx8PF5eXqSmpuLs7ExHRwc+Pj7i2MrPzw83NzdKSkpoa2sjODgYNzc3kQd5dxgXF9cnyZfZbKa6uprAwEDq6uqEdSYnJwcrKytcXV2pqKggNjYWe3t7Nm3axIIFC4QswPVAki7LX9TU1ODr64tGoxEifV5eXpSXlxMdHU1DQwOlpaX4+vri6ekpJBnkes7Ly8PGxobIyEhUKhUajQYbG5tefmPXm6e6ujrKy8sxmUwMHToUGxsbiouLhd9KcHAwHR0d5OXloVAoiI6O7sH+ajabSUtLE0KQsoMtgK+vb4927urqIiMjQ1hIQkNDCQoKoqKigpqaGvr164eXlxddXV09nte9zWQLRVZWFv7+/jg5OeHg4EBWVhb9+/ensrKSYcOGYTQa2b59oK0KsAAAaXlJREFUu/CnKCgowGg0CgmF7j4qcjsPHDhQ+L4FBATg5+cnRDttbW2JiIigra2N0tJSwsPDqaqqwt7engEDBojjMa1WK64fMGAAVlZWJCcn4+/vT0NDAwMHDkSr1ZKXl8fgwYOxtbUlNzeXgIAAsVjoXma535SUlIj69PDwIDU1FaVSyeDBg6murqatrY2QkBCKiooIDAwUDu7+/v5kZmYSERFBeXk5QUFB1NfX069fP4qLiwkMDMTOzo4LFy7g6+uLra0t3t7eVFZWolKpCA8PF+0tSZLon/JcJ4vj1tbWEh0dTWtrKzU1Nfj5+eHr60tubi4AUVFRdHV1XXVO1Gg0lJWViWcqlUpycnJQKBRi3lm2bBnPPPMMPj4+9OvXj4KCAsxmM97e3pSWlop52NXVFUdHR3G0GR4eflNpOMm4aRY0kiSRmJjIJ598wurVq/tUW7bAAgt+WmRkZLB69WoWLFjQS0/Hgv8MSJLEt99+y4kTJ/j1r39NVFSUpZ1uAjQ0NHDvvfeyevXqPmV4/hdxUy1oysvLaW5upn///jelB7cFFvynob29nfLyciIiInocu1jwnwNJkmhsbESj0RAeHm558d0kaG5upqSkRFjOLO16Ey1oLLDAAgsssMCC/13cVDw0MmTiJGtra7FrlH0mXFxcfvKVrEyApFKpfpDoWF8wmUy0tLTg6OiInZ3dD86zHJkhR8LI38khq7KHvlxPcgjqlZ7ysjOjra2tcISTI0sUCoUg3epeB2azuRdplxxh83PsJkwmk3ieXN4fKycgpylHE8hRKCaTCYPBIMrzYyDXpxwlYW1t/bPUj0wC1tnZiaOjo3jGlbT+wM+aj6vlTa1W4+zs/KNFOeXoI4PB8B979Pxjynu1dvxvgTwndXeOlqORZALFX8K/43rnbZ1Oh8lk+o/tSzK6k1A6ODj08hnUarXY29v/6PnqPxk3VZST/OJtb2/nmWeeoaCgQPzW0dHBe++912fEz49Fe3s7K1euJDU19SdJz2w2s337do4fP87WrVt/VFo6nY6XX36Zs2fP9vi+oKCApUuXsmHDBj788ENWr16NwWCgvLycu+++m927d4sXnEaj4d577+XTTz+lq6uLTz75hB07dvDJJ5+wdu1ampqa+PWvf83777/P119/zeuvv05bWxsajYZnn32W119/nYMHD7Jp0yY2bdr0o8pzLbS3t/Pyyy/z5z//ma1bt/Lyyy9TWVl53aHVJpOpV6SRVqvltddeY+XKlSLSQpIk3n33XR577DHBIvpDIUkS+/fv58svv+Sbb77htdde+1l1wlJSUvjTn/7U4xmdnZ28+eabrFy5koMHD/LNN9/w6aef/mx56I7uvDD/+Mc/RIjwtSCP82v9vmnTJlavXv1TZvW6Ieevr34nl1ev1/Pee+/9YA6QM2fO8Mc//vG/UlNOr9eLyJ4VK1bwr3/9i7fffptt27bx6KOP/ugxdb3QarWsXLmSlJSUa/aplJQUdu3a9Yvk6Vr4vn4PcPjwYV577bU+7/300097kErejLhpFjSSJFFRUcFnn30mdpadnZ20tLSI1fWTTz4peBPUanUvWQD5e5mWWqbW12q1Yhfb0tLSg1Ze5p6R2YjlkNMr6bHlsL7W1lZBty3Tb8viiPJvLS0t7Ny5kwkTJnDnnXei1+tFOWTrk+yxL1P9y5Nkc3OzIOKSRRLt7Ox66KwoFApCQ0NpaWnhlltu4Ve/+hVr1qyhsLCQsLAwnJyc+OSTTwQN+Pnz5ykvLycgIICuri527drFrFmzWL58Oe7u7nh7e2MymRg0aBDLly/n0qVLHDx4EFdXV8E9sWDBApYtW9aLxEvOZ0tLC52dnUKeorOzU1Czy3Xf3NyMyWQSdafRaHosQJydnXF0dMTf35+7774bSZL46quvgMsvEnmRJfPHyPUl78A2b94seDRkuLi4EBoayoEDB8jNzRVRMOnp6QB4e3tjNptpbW0V4eEy9bn8tyyHIbfhleXfsGEDw4cP56677iIhIUHoZ8nhn7LMhUxZL9eJfE13mQ+ZBl2mku8uIdHe3o6bm5vgxZDh4OAgIuUWLFjA0qVLmTZtmuhrMo16935nMBhoaWkRY0jua9eS/OhrzB46dIjjx4+jVCp58skncXZ2FtwsbW1tgvNHDjWXJInMzEy2bNki6Pu750PuP25ubn3ya8jt09raKu6R/5bp52V5AJmeQC6rLC8ij1s57F3uh3JETn19PR9//LHoz3I7mM1m9u/fLwgOn3zySRwcHEQ9t7S0iLErS13IMgZy+8l/h4eHX7N83fuBTBfQvX6699XuZe+rn8nznHyNHAIt56t7P7va3HplHhcvXsy9995LV1cXo0aN4je/+Q2RkZHo9Xq6urrEXCjnSf67r7RkagOdTifmEtniKNeD3D7yWNRqtYL8U6fTkZ+fz7p163otRCVJYsSIEcybN0+MZbVa3Yu3R64LuZ7kvtp9TF7Zj69siyvzLvcBuezy+022bnWfc+S03d3d++wXCoWChx9+mMDAwB59+Frt9N+Im8r2dP78eY4fP86cOXMwm80cO3aMdevWMXbsWIYPH86qVat4++23OXbsmGCzvP3224V5U6fT8dRTT/Hoo48Kptn777+f9evXM2PGDA4fPky/fv04efIkDzzwAK+//joDBw7sQbK1a9cuAgMDGT16tDAlV1VV8cc//pGxY8dy9uxZbr/9dhQKBfv378fNzY0ZM2agVqvJyckhIiKCfv36UVlZSW5urmD0rKurw2QyMWrUKN5//3369+8vQhVDQ0OZMmUKjY2NtLe3U1FRwVNPPcXGjRsFj8fYsWP7rLOWlhZKS0vx8vLC09MTgJEjR/Ldd99x6tQpQfsvR0Y4OzsTGhrKww8/zMsvv8ycOXNEWhqNhry8PDo7OwkPDwcuD6SamhouXbpEamoqs2fP7mEKbW9vZ+3atcTGxnLo0CGWLVvGhg0biIqKYsGCBaxdu5Zbb72Vc+fO0dDQQFdXF4MGDRLkZzLXjvwshUJBR0cHJSUlVFRUsHDhQsxmM5s3b8bb25vMzExiY2MZOnQoW7duJSYmhkOHDvHggw+ybds2Fi1a1CMMV6FQ4Obmxvz581mzZg1vvfUWSUlJjB49mvT0dMxmMwcPHkSSJPbt28dzzz0ntICOHTvGY489xltvvUX//v0pKipi8uTJLFiwoAeR44QJE3j66ad54YUXmD17NtbW1uzcuRN7e3vB5xMTE8P58+dZsWIFK1eu5J133mHPnj0EBARw8uRJfv3rX/Ppp5+iUCjQ6/VMmjQJo9HId999x6OPPipCa+XJrDvkemtoaBD08bJ8x7/+9S/CwsLo168fhw4dIioqisGDB1NRUUFcXBwHDhzggQceYOfOnTQ1NWEymXjxxRd7yWL0BaPRyM6dOwkPDyckJIQ333yTv/zlL4KDqKKigtDQUFxcXKitraWyspKZM2dy7Ngx8vLymDx5MmlpaSgUCg4dOsQrr7zC8ePHcXBwIC0trc+JWqfTsX37dvz9/dmzZw/PPfccO3bsoH///iQmJgoOqI0bNzJx4kQOHTrErbfeSlJSElOnTkWv17Nr1y6ioqJITU3lrbfe4u9//zsPPvggBQUFFBQUMHToUA4dOsTs2bPJzs5GoVCQkZHBXXfdxY4dO4iLi8PPz48333yTv/3tb9TV1ZGSkoKXlxcXL17k3nvv5bnnnmPKlCmcOnWK3/3udyiVSs6ePYtCoSAiIoIBAwb0Wadms5kvvviC/Px8FAoFs2fPpr29nTNnzjBz5kxxPFxXV8e8efPYvn07oaGhnDt3jvHjx6PVarlw4QLLly/nueee4+WXX+bNN9+kX79+2NjYEBMTg0Kh4OTJkyxZskRsRPbv38/jjz9OVlYWgYGBFBUVsXTp0j6Pw2xtbRk2bBgmk0n87urqSmxsLHq9nn379pGSksLy5cvx9vYmJyeHzMxMYmJiWLRoUY80CwsLefnll4mLiyM6Olq0+enTp/nDH/7Avn37CA8Px8nJiaKiIlpaWpg1axarVq3igw8+AC4vRk6ePElSUhIzZ87swctiNpvZtm0b1dXVPPHEE6xfv57o6Gh0Oh233HJLj3JdvHiRN954g/Hjx3PixAmeeuopDh06JOp81qxZtLa2cuLECRYtWkR7ezvW1tZUVFSwYMECce2pU6d4+umnefHFFxk/fjxJSUmsWLGCyspKjh8/zuzZs8nPz8dkMrFv3z6effZZ8vLy0Gq1VFVV9UmSqNFoeP7551mxYgVlZWUcOHCAmJgYioqKeOutt34yV4l/N24aC41CoSAoKIigoCDBSzJ58mTuvPNOzp49i7+/v9hxZGZmcvbsWUaPHt1jcNjb24sXVUREBKmpqYILZdOmTYSEhDB16lR0Oh3JycmYTCZmzpzJI488AlymCY+KimL8+PE9Ij68vb3R6XQsWbKEZ555ho0bN+Ll5YVKpeKVV17h5MmTqNVqobEjc08MHz6cvXv3Ul1djZeXFw0NDYKT4o9//CPz5s2jq6uLlStX0tjYSH5+Pi4uLmg0Gk6ePEljYyPjxo0TL/wrIS/AXn31Vf7xj3/g4+Mj6uH+++/niy++ICMjg8jISFEeKysr/va3vzF9+nQeeOCBHqbYtLQ0Hn74YZ588kmGDRsmvpd3KX0JF546dYqWlhYmTJjAoEGD+Prrr5k/fz5nzpyhrq6OwYMHc/DgQSorK/H09KSxsRFPT0+MRiMvvfRSn9o+xcXFPPLII0yfPp1bb72V2tpavvvuOyZPnsytt97Kv/71LzZv3oyTkxOTJk0SGjL+/v7Ex8f34BSRceedd3LmzBmysrLQarUEBQUBl5lnt2/fDoCHhweVlZVERkYSEBBAUVGRYKxNSEhgyZIlvY7+AB544AGee+453njjDd58801qa2vZv38/U6dOJSIigvr6elxcXGhubsbd3V1YDiMiIgQ3TXNzM3Z2dkRGRvL73/9eiDT6+/tTUFDAF198weTJkxk/fnyf5ZP7g9xOZrNZWORWrlzJ7Nmz6ejo4Mknn6Srqwu9Xs+ECRMYMGAAGzduxNXVlcDAQF599dXr9jWwsrIiKCiImJgY+vXrJ3bP69atY8KECSxbtkxIO8hyGNnZ2YSEhBAeHk5XVxffffcdKpVK8Pzs37+fKVOmMGrUqD5fpqmpqajVaiZNmsT9999PcXEx6enpTJw4kalTp/Lhhx/i6+uL0Whk6dKlKJVKoqOjmTNnDklJSfj5+eHo6MjTTz/NwIEDOXXqlPAV8/T0pLm5maCgIPz9/enfvz9BQUH069ePxsZGysvLCQwMJDY2ln79+glB2E8//ZTBgwcza9Ys8vLyKC0tRafTMW3aNEaNGkVGRgYuLi4MHjxYWEyvBqVSiYeHB+7u7rzyyivs2LEDnU5HUFAQxcXF4mU2fPhwKisrBdnarFmz+OCDD3BwcBD9TKvV4ujoiNFoZPr06dx9991iHD344IM4Ozuzfv160c9KS0vJysri9OnTfUo4yLiaNAhc9rGbN28e06dP5/z586xduxatVktgYCBlZWW95g8/Pz/a29v59a9/TUBAAKdPn8bGxgalUkl1dTXfffedWAQ6OzvT0tKCp6ensP7Jzw4JCSE0NJTAwMAeeVMqlQQEBNDc3IzBYODIkSO0trb2KcPh7++PwWBgxYoVLF26lG+//RZ7e3sGDhzIypUr2bBhA3q9XrRFSUkJhw4dYtSoUeTm5pKYmChYjY1GIyaTiYkTJ4qFe3BwsNCe2rZtG3BZwiYvL48NGzYwZcoUxo0b16ePjLOzM1ZWVuj1ery9vbG2thaLJJkj6WbATbOg6QvdKd67D6IHH3yQtrY2Xn311V5HMbNmzeLYsWPU1tZiY2PD/v37iYmJEeyOAAEBAcLR1tbWVhxxySJxfenDyNTb3t7eODo69qCrr6+vJz4+nnvuuYcVK1b0uLe+vp7+/ftz22238fvf/x6lUom1tbVw2pXTaG5uxt/fn1tvvZUXX3yRpqam76WMt7a2ZtmyZQwfPpyvvvpK5FuSJGbMmEF9fT3ffPMNQ4YMEffITMmPPfYYr776Kp999hl6vR6FQsHEiRN5+OGHWb16dQ9dotDQUEaMGMGKFSuEk62M9vZ2cWYuTwgJCQl0dHRw+PBhYmNjqa+vJywsjMWLF/PHP/6xR7n7KltcXBwrV65k7dq11NTUoNfraWxsxGg04u7ujkqlQq1WC7p4+cUt40qTM1ymLZ84cSJ//vOfGTp0qLC+dXR0YDAYmDp1Ki+++CL9+/cnIyNDMLxeSed/pdm8o6ODgoIC5s2bx5o1a9i3bx+FhYW9joW6QzZXX7x4EY1Gg4uLi3CElp221Wo1EydO5A9/+ANDhw6lqqpKXHO1F01AQADDhw/n3nvvxdnZmc7Ozh71LPe7zs7OHm2m0+lQKBRCduOHOKnK49NkMlFeXk5nZ6cgfCspKaGsrAxPT88e9afRaLCysmL69OmsWrUKR0dHmpqaRDn7gkyzr1AoiIyMRKfT0djYiCRJ4vgQ6EFlr1QqBZV9d5r8wMDAHo7zV7aRJEnk5eVRV1eHu7u7uEaSJFFeSZJoaWkRxJbe3t7o9XrxTJVKhclkoqamhuzsbLy9va95RCCna2tri5WVFQ0NDYwcOZJHHnmEBQsWMGvWLMLDw/nTn/4kGJrNZrOwzvaVnjzejEYj5eXlGAwGgoKCsLa2RqPRMGnSJP7whz8wYsQIfvWrX9He3s6f//znHschV8OVv10pHVNXV0dCQgL33Xcfy5cv7zMtuV/KC7DZs2ezatUqwsLCePLJJ/noo4/Yt2/fVfPwffmR+7O9vT1/+MMf2Lx5M5s2bepx7ZUSCAEBAWJhJQdPNDc3M2bMGB5//HFmz57NggUL8PX1ZeXKlZSUlODg4CDy7u7u3uO90d1/Tz6KmjJlCi+++CJhYWHU1tZ+7/jubhWW+3BfZf5vxk21oHF2dqa6ulrsomUpAPn/8vnl8ePHeeihhwgICOi1+AgMDMTPz4+QkBDmzp1LZ2cnbm5uLF68mLNnz4rJdty4cWi1WrFy12q1zJgxg7a2NtasWYNOp+uRbmdnJ0VFRSQlJbFo0SKRF4PBwPjx43nttddYt24dSUlJaDQa8ZkwYQLvv/8+X331FYcOHepBay3/v7Ozk9GjR7NhwwY+//xzdu/ezdChQ0lPT+fAgQOUlpZSXl4uJuvuk2hLSwsrV67k4sWLrF27lurqanJzc4VGj7wTaW5uFvTv77zzDsXFxWi1WsaMGYNerxcaL3feeSeBgYG8/vrrNDU10djYSEVFBZcuXeLs2bOsXr26xwAaN24czc3N5Ofnk5+fz+23346NjQ1z586lvb1d0L5/9NFHrFmzhoMHDwpfmCvPf+WFS0NDA+PGjWPu3Lk8++yz2NjYEBUVxZkzZ7h48SILFy5k0aJFZGVlCaba+fPn4+rqyvnz53s4pup0OnJzc6mpqeG+++4jICCA/v37U1tbS0tLC87Ozuj1et544w3Wr19PTU2NoGRvbm4mNzdX+BzIchhX+tH861//Egq58fHxREVFYTQa+eyzzzh//jySJOHp6UlZWRmHDx9GrVaL55SUlIjntLa2isg4X19fXnnlFdatW0dDQwPDhw/n008/5dy5czQ1NfVYMMkyDfLR4Pnz53n//fdpbm4WZ/yyP4msh1ZVVUVRURHFxcUsWrRILBCvpl92Nbi7u3PhwgWhb9bV1cXo0aN59tln+fLLLyksLOTYsWMUFRVRU1Mj6OALCwtRKpWUlpby97//nY0bN+Li4iJYizMzM6mtre2xUAUYPnw4p0+f5p133uHbb79lwIABODs7c+HCBTIzM7nzzjuFL4xMhS/3N7ntampqKCkpobGxkcmTJxMQEMDBgwdJT0+nubkZKysrmpqahC9ZSUkJDQ0NFBQU4OLiwoULF6ioqBD1efvtt3P48GHKyspwc3MTciXdx/jZs2fJy8sTFPo1NTV0dHT0Kp/sVyW375AhQ3jppZdYv349WVlZJCYmMnPmTMaMGSPmupSUFC5evMjtt98uFpBHjhyhra2N6upqMU94enri7e3N888/z9q1a9Hr9Xh5efHKK6+wfv16cSTywAMPEBwcjFqt5o033ujTAiD74qjVampra0V0jlzfra2tYn559dVXWbduHcnJyb1e1t2lPGJiYkhOTuaDDz5gy5Ytgn396aefRq1W4+PjQ1ZWFsePH0ej0dDQ0CDy4OLiItiHu88psk+jRqMRsid/+tOfxGbxSrS1tVFSUkJ6ejoLFiwQ48LKyoqYmBheeukl1q1bR15eHidOnODWW29l6NCh9OvXj5SUFN5//322bNki/Ai7zxuyWKq8EfvLX/7C+vXrMRgM9O/fny+//FKcKlxZ5waDQWzK5TEt9+m+pGH+W3FT8dAYDAbS0tIICgqioaEBDw8PJEmitbUVPz8/Kisrhf6FTqfD1dX1qvT9zs7OYpDJu6uysjJaW1sJCgrC3t6evLw8sYssKCgQZr3GxkYiIyOF6V2v13PXXXfx5JNP4uvrK6j129raGDBgADY2NmRkZCBJEvHx8bS1tVFeXk5ISIiQb1Cr1QwePBi1Wk11dbUwWdfU1AhzakFBgTimcXZ2pqSkhKqqKry9vYWPjLwrrK2tpaKiAi8vL8LCwqivr6e2tlYoV4eHhwu9GJ1OR1FREXZ2doSEhAgzpdlsJj4+Hr1eT2FhoaA87+rqoqCgAF9fX+rr6zGbzdjY2GAwGHBxcekhFyBJl6nDKysr8fHxESbfjo4OUfdms5ns7Gza2tpISEigpaWFhoYGBgwYgJOTk0iro6ND+A4MHDgQpVJJdnY2AQEBODk5UVBQgKOjo6D9rqqqoqGhQahq19XVUVVVRUJCgjDbarVaCgoKcHd3Jzg4WIQ+5ufno9PpiIiIEGKE/fr1Izg4mIKCAjo7O3FwcBDh3a6ursJyEh0dLY7wzGYzhYWFYtEQExODu7s7NTU1FBQU0NLSQnJyMqtWrSIzMxM7OzshRlddXS0sNLKDrLW1NREREXR2dpKWloa/vz8RERF0dXWRlpaGu7s71tbWhIeHizzItPoGgwFbW1uMRiN2dnbCd6V///7odDoqKysJCwvDw8ODuro6IZsh93+z2Sz68/WitbVV+IpVVFQIZeELFy7g4uJCVFQUjY2NFBUVERISQkNDA5GRkWRkZBAbG0tHRwc5OTlERkbi7+9PU1MTWVlZ+Pr6Cof07tYaOXigpKSEhIQEXF1daWtro7i4WAjR1tbW0tDQQGhoqPAvMxqN4sW5Zs0aHnvsMQIDA/H29katVpOZmUloaKg4qsvIyCAoKAiDwUB1dbVQZPb396ewsJDw8HAqKipEGkVFRXR1dREaGoqVlRV5eXkEBATQ2tqKUqnEy8uL7Oxs4V/n6+tLU1OTWJDJMJlMFBUVodPpGDhwIJIkkZaWhqOjo7B2qtVq7O3tRX8uKirCxcWFsLAw4LIviBze6+bmRkVFhagb2ccmNDSU0NBQNBoNFy5cEP1M1o1ycXHBx8eHS5cuERsb24t0UZ5nS0tLcXR0JCIiQjjnyu4BRqORfv36kZmZiUKhID4+vlffktMIDg7G29ubiooKioqKiIuLE3OgfKRkY2NDeno6bm5uSJKEj48PZWVlODk5ERAQwIULF4iOjsbV1bUHpUFJSQnt7e2i7uGyNdPZ2bnHu6OhoYGHHnqIF154AW9vb6H8bWVlRUREhHg/ubi4EBMTw/9r77zDq6rytn2fnPTee4VAKCEEpEsRBcQCKNgwIGQcwe5YZmSwjGV0ZCzIqIgoCghIhwChhCYoSC+BNEjvvZ/knJyyvz/49npzSHDQceZ9w6z7uriAs9vaa6+99m+15ykvL6epqUmU05KSErKzs+nbty9ubm5kZmYKqxWLxUJUVBTnzp2je/fu2NjYcP78eaKiosRzOHfuHL6+vjg6OhIREWHVe61ahHh4eGBjY0NDQ4MYro6KihI9Ql2dGyqg+b9KXV0dCQkJfPzxxz/rZi2RtMdsNrNx40aOHj3Khx9+eEPrR3QF1FVZ69ev55NPPunUHV3y34miKOTm5vLss8+yYsUK0XiU/GeRAc1/AHWFib+/v5Sollw3JpOJvLw8jEYjPXr0kNYC/8uovTsNDQ1i5YxEAv/T01xWVkZoaOg13d4l/15kQCORSCQSiaTLc8P1YaviTu1VX21tbcXYq6pKqq4CUI9RxYpUTZmmpiYcHBzErPWrJ/mq57uWjL9Op0Or1V5ziezVaVbTdbVk9T/DYrGIybEajcbKskD93cnJyWpMWK/Xi9n3er0eg8EgxsxNJpPIt/bWEW1tbR0mfDo5Of2sbLter8dkMv2ilqy6gken0+Hm5ia0GhwcHIRImbpE9tfIo6t6QQaD4TeVjVcUhYaGBiFlbzQaqa2txdvbu0tqPDQ3N2Nra3td5ffXoq7YuLpN1V76XhVWU1dbAaJ8ODs7X3cZaP+eqCsf1X/b2dmJybWOjo6/ynpBTZO6glFNu16vx2KxoNVqxfum1jNXp739/u1RLQHan/taqO9PS0uL1fyy/zT/7BmpdayTk9Mv6nnsLJ+vhSo25+DgcN3v4K+pH1RxRXXuTXtxw587h8lkory8HAcHB3x9fcWxer2eqqoqPD09f7HJsqpBo+aN+rfBYECr1Ypvnio1oK7AulG4ce6E/+n2O3HiBPn5+Tz00EOsXr2apUuX8tFHH6HX6ykvL2f27NmsWbNGVKQtLS0kJiayaNEiWltb+eqrr0hKSmLlypV88cUX1NXV8dRTT7Fo0SL279/PqlWr2L59Ozk5OWRkZHQ6QzwlJaVTzZGr06t6JK1fv54lS5b87L6qz1J7zGYz3377LY899hg7duzg9ddf59ixY1gsFg4fPsykSZPIzc0V5zh//jx33HEHx44do7CwkL///e8cPHiQ+fPnk5+fz4ULF5g2bRqbN29m0aJFfPXVV5hMJi5evMh9993Hli1b2LdvHx999BH5+fk/e3+nT58mOTn5Z/fpjLS0NJ588kkMBgNJSUmcO3cORVE4ePAgqamprF69mkuXLv2ic7bP623btvHxxx//4nT9HGazmc8++4z6+nqMRiOrVq1iz549/Pjjj7/pdf5T7NixgzNnzvxbr9HU1MTrr7/Oxo0befjhh/nmm29YsGCB1XVVG5OMjAzxW3l5OY8//jjl5eXXfa22tjaWL1/O3LlzKS8vZ8eOHdx9992cP38ek8nEp59+ys6dO3/xKi2V0tJSHn/8cSuVVovFwpEjR5g8eTL79u3DbDbz3XffsXr16g6rk9Q0vvXWW2zYsIHExEQWL17Mxx9/zMqVK3nyySev2ybhxIkTzJs3r1Nl3c5QG3S/ZWd9RUWFeEbqx/Nq1qxZQ3Z29i86b3Z2Nk8//XQHmYVrpX/9+vVWZed62Lt3L++88851719RUcGXX34pbC3MZjMHDx7k7bff/tk8NZvNpKSk8Oqrr4p63WKxsGrVKpYuXdppGflnHDp0iNmzZ/PUU0/x+OOPk5qayrFjxzhx4gQ7d+4UCzq2bt1Kamoqu3bt6pL2Gdfihgpo2traWLt2LbGxsURERKDT6Rg1ahSJiYkkJSWRlpYmrNa//vprIUF99uxZSkpKCAwMRFEUtmzZwtixY5k1axahoaGiN6BXr15MmjSJGTNmMHDgQHr06MGJEycoLS21SoeiKNxxxx0MHTpUtBIaGho6FBydTsfHH39MbW0tAQEBNDQ0UFdXZyVlX1NTI3o6li1bJlaTqNjZ2REWFoaNjQ3Tpk1j6NCh/P3vf0dRFHr06EFbWxsrV64ULTdVyC46OpqffvoJvV7P5MmTSUhIoK2tjaioKJqamhg/fjyzZs3iiy++oKSkhG7dutHY2Mhtt93G5MmTefTRR/Hy8upw36pkvclkYvDgwdx5553id71eL/Rg2t9nS0sLLS0tVFdXY7FYCAsLE63pe++9l7i4OHQ6HWvWrCEqKoqEhATCwsJERVZdXS3sKNRln1d/mFpbW/n444+prq4mICBALGFsL3dfU1Mj/t/+ntrbS9TX12MymYQdgWqdoNVqeeaZZ/Dw8CA7O5tjx44xZcoURo8ejU6ns0qjKtXe2NgoNCXU7epy1c50TdqXB7UlWVNTI/ZX5d3VZcYGg4GamhosFgsmk0mkV71Hk8lk9f/26ZoyZQoDBgywkmtXezWampqEKrDaYtbpdNTX1/+ij6JGo2H27NliqfQtt9zCM888I0QE1Za4vb09LS0tVFVVYTKZ8Pf3F5o1as+Yem112fLVz9HBwYHw8HAAwsPDueeeexgwYAB79+4lNzeXiIgI7rnnHquWvBoE19TUiA+mag2g5qv6m62tbQfDRa1Wy9ixY7nrrruEQKZGo+GRRx65ppLyAw88wIwZMzCZTAwZMoRnnnmG+Ph4sYS8trZWaLvodDqRjvYEBwfT2NhIQ0ODqOPU1Z6qBYJqeaAu3120aBF1dXVWWjlqT09DQwNGo5HW1laqq6tFvreXnlD3UZcE+/n5iWe0f/9+UlJSOryTM2fOpFu3bmL1nyp8enW5b25upqamBpPJREBAADqdjqamJpHeCxcusHr1amHJUVNTI9SwH3roIWJiYjCbzVbL4NX7q62ttbJxaGhowNPT00pHCxB1lmpH0dzcjMlkoqmpiYCAAObMmWOl0qzW5+3tVq7G3t6ekJAQUZ8oiiJWOo0bNw4/P7/OX5xroCgKHh4evPvuu8ybN49bb72V0NBQNm/ezNChQ+nduzdr1qwhJSUFJycnhgwZIiQEbhRuqCGnixcviqEKtQKqqakRSquBgYHAFeG18vJyUlJSuOeee8jKyiI2NhaNRoOTkxP9+/fnscce4/XXX+fOO+8UXYYFBQWcP3+e48eP89BDD6HVagkLC2Pr1q08+eSTYj+j0cjixYsJCAhg9OjRHDp0CC8vL2JiYoiOjhbpLS8vZ+fOnYwdOxZFUcjOzmb16tVkZmby97//nc8//5xhw4axd+9enn76abZu3Sp0UNp3E6py90VFRZw5c4ZBgwYJgSXVgHLu3LkYjUa8vb3FENBNN93EwoUL+etf/8oTTzyBt7e38H+pqqoiLy+PiIgIsczRZDKRnp5OYWEhGRkZzJo1yyr/a2tr2bZtG+Hh4Xh5eZGRkUFJSQlz5szhhRdeYPTo0aSmphIcHIy3tzfnz5/npZde4uWXX2b48OGcPn2aW2+9lUmTJgFXukk/+eQT4uLi6NOnD7m5uaSlpXH48GH69evH+PHjOXjwIE1NTaLnbe/evQQGBhIWFmalkFxdXU1ycjKjRo1CURTy8vJYu3YtZ86cYeHChRw4cAA7Ozu2bdvGm2++KVSTAXJycliyZIkwqXz22Wf5/vvvKSgowMvLi7a2NubMmcO8efP4y1/+QnZ2NmVlZRQUFKDVajlx4gTBwcEcO3ZMWEb06NEDk8lEcXEx8fHxQlgrMjKSI0eO8NFHH1kp7ra0tLBp0ybCw8PZvn07r7zyCuvWraNfv37s27ePyZMnU1ZWxrZt27j55pvZu3cv99xzDz/88AN333031dXV7N+/nx49enD27Fk+/vhjNm3aJOwLHn/8cd566y0hqW80GunVqxcDBgzgxIkTODs7ExcXR2FhoegZKS0t5c477+S1117j7rvvJiUlhb/+9a9ERERc1/vq6upK3759rYZzfXx82LBhg0jXyy+/jMViYefOnVRUVBAdHc1zzz0n3rXjx49TW1vLwYMHuffee8UwRlFREffddx8ODg7iHWnf9W9nZ8fLL79MQkICzc3NvPbaax1WkakGkiNGjODQoUM88cQTvPTSS4wYMYJTp06RkJBAREQEKSkpdO/enaqqqg73qNVqmTNnDomJibz99tu89957Ik1XY29vT3x8vJWCrZubG3369EGn05GcnMzhw4d56aWXsLe3JzMzk4sXL9KzZ0/uu+8+cY+q3ciyZcs4duwYb731lrAN2bVrF7///e/Jzs4mKCiInJwc4uPj2bVrF+PHj7dqpPzwww8sW7aMyMhIRowYgU6no7S0FK1Wy+23387Jkydxdnamd+/efP7558ycOZPi4mLOnz/PvHnzRECze/dufHx8GDlyJJ6enqI8v/HGG0yaNAl3d3cKCwuFnYCvry9w5QN96tQpsrKycHd3Jy0tTYiifvvttxw7doynn36azMxMzp07x4QJE1i3bh1Dhw4lJSWFF154gffff59bbrkFJycnli5dypgxYzh69CgLFy7kxx9/xMbGhqSkJObPn096ejptbW1CPLA9TU1NPPXUU7z++uusW7eOnj17Mn78eJKSkujbty/Lly/nnXfeISkpCX9/fyIjIykuLmbDhg0cPXqUf/zjHx2Gj9ra2nB0dMTV1VUEmqpujWo1o6LX68nOzu4QGDk5OdGtWzfxPVBFP48cOcLo0aNpbW2loaEBW1tb3N3dycrKQqfTMXHiRODKkFRBQYFYst/VuaF6aNR19ioWi4UDBw4wf/58PvjgA0JCQoArFcejjz7KihUrSE1NJTIyUswVsLGx4Y033uDhhx/m2WefZdWqVaIQqYaA7QtVQEAAaWlpVr/Z2dnh7+9PY2MjOp2OHTt24OPjIwIqFT8/P/z8/IQ/SlRUFI899pgQfBo8eDDu7u6kpqZiY2NDQEAA/fr163TMuba2ltdffx2j0ciLL74oCnhUVBQ33XQTGzdu5MSJE1bmkNHR0axevZqioiKmT58uPlRGo5GNGzfyySef8I9//KNDnl6dByqtra3s3LkTrVZL9+7dCQkJob6+Xhw/ZMgQ7rrrLqqrq0lISKCiokJ8RKZMmcL8+fPZuHGj6J52cHDA29sbnU5HcHAwPj4+jBgxgu7du9Pa2ipEAz09PYWBXXJyMs7OzoSFhVmlzdvbGz8/P2JjY7GxsSE8PJxHH30UnU5HWloaycnJKMoVtdiKigpxnEajwcvLi7q6OhFgqIrD/v7+JCYmcvHiRdzc3ETexMTE4O/vT2xsLF9//TVxcXFMmDCBsrIyYXA5YcIE/vjHP2JnZ0dcXBz33XcfpaWlPPTQQ6K3qj1HjhwBYPTo0SQmJpKenk5ubi4333wzY8aM4YsvviAgIABbW1umT5+OxWJhwIABjB8/ntOnTxMYGIiXlxcvvfQSYWFh/Pjjj/Tv3x9fX18uX74sWs8TJkzg+eefx8/Pj+bmZpqamti2bRtBQUF4enry5ZdfcvPNNzN58mS+//57oZQ8efJkIiMjycnJ6VAurkVnEviAVbpqamqwsbHhzjvv5K9//St79uwRKsUmk4kVK1bQ1tZGeHg4RUVFFBQUsGvXLoYMGfKzy9zVYGHgwIEcP368Q4scrtQFQ4cOxdXVldOnT2NnZ4eiKIwaNYpx48Zx+vRpVq1aRWxsLCNGjLhmi9rBwUFYGFz9XP9ZfrRXqZ06dSqDBg3iwoULrFq1iqamJkJDQykrK+vwPvr5+fHcc89x5513snz5clatWoXRaBT5dPnyZfbv38/w4cMJCAjAz8+P3r17W10/ODgYRVF45ZVXyMjIoKqqCj8/P+rq6kTjJSgoSHgfqUJ7qvAbXAnoQkNDiYmJEcEMXDFEdXV1Ra/XU1NTQ1JSEjExMVaaOhaLhS+//FL09KamppKRkYGrqytz5sxh7ty5rFy5kpCQECIiIggKCmLQoEF4eHhw4cIFzGYz7u7utLS0iLp3xowZtLW1cenSJTZv3ozFYhHaQOvWrWPs2LGMGDGiQ9nx9PSkf//+ZGRk0K1bN06ePElZWRmDBg0iKirKStAyLi4OOzs7goKCSExMRKPRWNUpKjU1NUIfrLKykvz8fKFtc7Vys9qrps5xVP+079FSlYUbGhpITU0lKChI9KKq241Go1D2VvlnDt5diRsqoLG3t7ca1tFqtdx7771MnDiRzz//XLQEFUVh5MiRWCwWVq5cyeDBg8UxFRUVVFRU8Mgjj7Bw4UJWrFghxjKjoqIYOHAgiYmJYlhDnVjYvoC0r5jCw8P5/e9/z3vvvcfx48c7pPnqQEjtWTEYDBw9ehQHBwcxUbmzY1SCgoJ44403OHHiBKdOnRL72NjYMHv2bFauXElzc7NVz8OZM2eIjIzks88+IywsjJSUFOBKBfy73/2OmJgYVq5cKfLU1taW2NhYhg0bxgMPPEBxcbFVWnx9ffnDH/7A4sWLhdFa+zxpL+euToBTrQG0Wi1eXl5WolbqcVej/lZdXU1ISAj33HMPL7/8MsHBwTz11FN8/PHHfP/99z87/KFKf9vY2NDc3IzZbGbcuHG89tprHYKh9qjnbC9z3v63q2lqaqK+vh6NRoO/vz8mk0nYCKjlpr3MfvvJ2+2pqamhoqICjUYjAjp1iM7X19fqebeXNtdqtcKaQj1/UFAQNjY2HD9+HLPZLByuVVsN1coDoHv37sycOZO3336bM2fOiCExOzs7vLy8MJvNIh/bS7T/WiwWS4d0wZV3QxVrU9NmNpupra1l2LBhPPnkk4wbN4577rmHsLAw5s2bJwKfzjCZTGzdupUZM2YwYsQI3nvvvQ6tcr1ez5EjR3BxcbHq6VHLsNrDdvVEzKvvJyUlhaFDh/L73/+e119/XQy1Xqth0Blq/qoTmlW7lJkzZzJjxowO51HrkoiICDQaDY2NjYwaNYo//OEPDB06lBkzZuDg4MCrr75qpSrbvixrNBpRHmpqaoiOjuaBBx7gqaeeIjo6WpSLs2fPWglldoY6pNMe9ZhBgwYxceJE5s+f3yEgVpVttVqt8HBTbURCQkKsejHb2to4evQo9vb2ODk5iYUSKmq9Y2NjQ2trKwaDgVtvvZX58+cTGhpKRUWFqI86CywnT55MUlKScLo+evQoPXv2vGZgrr7XNjY2nQ6llZeXExgYSEhICKdPn8bLy4u8vDyio6M7BFSqoWh5ebnVH9W2oz3qvDCNRoOLiwu2trZieN7T05OgoCDxTTMajR2mDnRlbqiApn///qIFVFdXh06no6qqiueee46qqiq++OILysvLuXjxIgaDgdmzZxMbG4tWq6Wmpoby8nJaW1v56KOPyMrKEhL6JpOJ+vp6cnJyOH/+PAcPHmTFihVoNBpKS0sZNGiQVTrUirahoYH09HTs7Oz4/e9/36EVqLb4Tp48SU1NjRhvbW5uprS0lIMHDwpJ+7y8PFxdXTl27JiV7LYqxV5fX4+vry+vvPIKf/nLX0hPTyc/P5+MjAx69erFiBEjuPnmm6mqqqKlpYWKigqOHDnChg0bKCoqwsXFhbi4OKqqqsQ8jNdff50ffviBDRs2UFZWhk6nIzU1lXPnzrFmzRpOnDhhdT+qRsfTTz9NU1OTkM5X5bXVMff2lg06nY62tjaysrL46aefmDJlipCdV+e5qB9R9Td1bPrmm29m+fLlfPHFF+zcuZO8vDwMBgNPPPFEhwmUaiV47NgxqqurxZh9c3MzHh4eQqb922+/7fCcnJ2daWlpITk5merqasrKykTvm/q3WvHW19dTW1sr5sc8+OCDHDhwgLy8PJycnIiLixN5YzQaxXyA9ulRJc/bV1QjRoxg7969fPDBB2zfvl20As+fP8/Fixd56KGHrPJUp9NRV1dHY2MjLS0tmM1miouLyc7ORqfTMXDgQA4cOEBWVhZ1dXVcvHhRpEudL9DQ0MCFCxfw8PBg1qxZGAwGpk2bxu7du7l06RK9evXC398fvV4vnu215gtcC0VRRDlR5fzbp0s1XC0sLOTs2bOMHz8eR0dHca/9+vXj1VdfZcWKFVy+fJmDBw9y5513ctNNN1m1PE0mE2VlZTQ0NFBTU8P69eupr6+nb9++JCYmkpKSwvr1662CmpqaGn744QcyMjKoq6vj8uXLYs5JQ0MDLS0tjBs3juXLl3Po0CFqamqs5tOZzWYOHDjAqVOnGDZsGFOnTqWwsJDFixfT3NzMggULOgxTqfM4VNsBdV6UWtbUeTGjRo3izTff5JtvvuHkyZMdhtPMZjP5+fnk5uaSmJhIeHg4r732GitWrKCoqIj9+/fz4IMPCkVto9HIqVOnrDyD2svk33LLLXz00UcsW7aM77//nvT0dFEuVLPW3bt3c+bMGerq6sQ729jYiLe3N+fOnbPqpTCZTGLu0/Hjx+nZsyf33HOP1XtrY2PDQw89xJ49e8jOziY4OJjY2Fjs7OzIy8vj/PnzJCQk4OXlRXZ2NqmpqVZ15qVLl8Q11PdRtZtQV36+/fbbYo5hdHQ0y5Yt49SpU53aB0RHR+Po6Ej//v0ZPXq0aGyq99ra2oq7uzs//fQTJSUl4lqqjU37+Unl5eUcOnQIo9FIVFQUJpMJe3t79u/fL3pWrq6DBg4cyKBBg6z+9OnTp0MgnZ2dLTzVvLy86N27N8XFxWRkZHD77bczfvx4YWhrZ2fXqcFvV+WG0qExm82sX7+eW2+9FRsbG4qLi/Hw8CAyMpK6ujpKS0vx8PCgrq6O0NBQXF1dxYTJvLw8bG1tCQ8Pp6ysTEz6io2NxWg0kpeXB1ypLNS5KP7+/qxevZp77rkHb29vkQ6TyURubi6KoghJfa1WS0REhNVwkaIoXLp0SbSm1K5RVRK9rKwMJycn2tra8Pb2xmw2U1dXR+/evcVSSJPJZCWb7u7uzuXLl7G3txdLVVVZcRcXF9ED5ePjg729PY2NjdTU1BAYGEhERATl5eVUVFTg6+tLaGgo5eXl1NbW4uLiQm1trXhRjEYj3bp1sxoXbm1tpbCwEBsbG3FsS0sLoaGhFBYW4uvrS1tbGzqdTlhRBAQE8PzzzzN37lyCgoLo1q0b9fX1lJSUEBwcTE1NDVqtFl9fX+FWrHZpR0VFCc8kdSipqKhIDCm1n6ugKIpo/dnb29PU1ERwcDBFRUWix0LtTg4KCupgh6HmsZubG56enmKSompToD5nX19f8ZwiIyNxdXWloKAAnU5HWFgYdnZ25OTk4OHhgb+/v/g3XGmNhoSEiHxRJ1aqaSgqKqK4uJh+/fqJcfeCggK8vb0JCQmhtLRUSJqrthYGgwG9Xk9JSQlbtmxhzpw5hIaG4uXlxeXLlzEajaKstLW1CTuQvLw8IblfWVkp3g2NRkNOTg4Wi4WIiAiMRiMFBQWEhIRQWVmJi4sLYWFh171cWFEUKisrKS0txd3dnYiICHJzc4W0gp2dHZ6enuTm5grbCr1eT15eHv7+/nh4eJCamoqHhwc9evQQfmPOzs5WzsltbW3k5uZiMBiEbYCjoyORkZHCakS1jVDLjdlsFsOJOp1ODD36+voKp+zIyEixikaVsFcVhNV3U1EUunfvLtIN0K1bNwoKCqyup+ZHbW0tRUVFODs7ExUVRUtLC/n5+YSEhFBbWyvm7qWlpaHRaDpYC1gsFvLz82loaCA0NBRfX19aWlo4f/48/v7+dOvWjZKSElpbW3FzcyMgIICsrCycnZ3FM1YUhbKyMqqrq4mKisLZ2ZnMzExaWlro168fbW1tlJWViXJhMBhIS0sjNDQUi8WCu7s7BQUF+Pv74+LiQmZmJv369RND+6qdipOTE25ubtTU1GBvb094eLjVMm+z2Uxubi5tbW1ERETg4uJCSUkJlZWVBAQEEBwcTFtbG6mpqfTs2ZOioiJh3+Hi4oJOp8Pe3h5nZ2eqq6sJDQ0VFht2dnakpaURERFBaGgozc3NXLhwAV9fXxwcHAgNDbVKi6IoVFdX4+XlJXo4XFxchG1MVFSUmMCslpXQ0FCKi4vx8/MjICBA5G1lZSWVlZUEBQWJCeI6nY6Kigrs7e2Jior61crg+fn5aDQaIiIixOTxnJwc7O3t6dGjB1qtlry8PBobGwkJCRFLxm8EbqiABq5MNktLS2PAgAH/Vql4NRhxc3Pr8AGUXD+NjY088sgjvPnmm8TFxcl8/DegKApbt25l7969fPTRR/9WbRmJRCL53+KGC2ja386/8+P4n7rOjU5zczOFhYVibFfm5W+P2rvT0tJCeHi41bwDiUQiuVG44QIaiUQikUgk/33cUJOCO0MdQ1Qn2F2N2WymoaHhulU1/xXaS/f/WlQhr5KSkg6rMlRhq86E5X4t6oz+3+p8EsmvQX2PVcG09lgsFkpKSigqKhJL51UdoF+jtqqKDLa3ZVAFBFVROVV0T12xdL3nbWtrE2J3/wl+62sajUYrAb7fAlV/5dc8q/+rqLYG1ztJXq3X1QnEqngm/I+1grrAQi2D6rarvwP/zdzwAU1dXR3r1q1j06ZNVjLbFosFg8FAcXExTzzxRIcZ7f8OGhoa+MMf/kBWVpaoNH8pzc3NrFmzhu3bt3Px4kXxu6IorF69mieeeIJ33nmng/dUe9RrX8+LVllZyeOPP96pjoJE8p+ira2Nd999l4MHD3a6PSsri8cee4yCggLgin3GwYMHf3EgrihX7EFOnDjBxo0bSUlJwWg0kpSURGZmJps2baK1tZWSkhJ27drF6dOnOXz48HV/4Pfu3ct77733i9L0a1CU/7FKSUlJYcGCBb/JeQsKCli9evVvcq72bNu2zao+6wqoS6E7W5K9ZcsWTpw4wdq1a6/7fGfPnmXq1Kns3LmT1atX8/e//10IDr733nu8+uqrbNmyhddee42ioiIUReHw4cOsX7/+PxYg/1/nhgpoFOWKTH1lZaUweTt+/DgFBQXMmjWLnj17iv0OHDjA9u3b8fT0xGAw0NzcLHQ9lP8v715VVWW1fE7Vjmhubqa+vh69Xo9er6eiokLI2jc3N1NXVyfkwFU5eVVS28nJCZPJxLFjx1i/fj0Gg4HW1lZqa2s79NyorSt1CaHFYuHcuXNkZmYyffp0+vfvL/atqqoiLS2NF154gTfffLPDPAnVLFFdJvzhhx9SU1MjWkeqrLyqCKwu3/b19RXaF+qSUVVe/j/VsyW5cVElDlTrAoPBIOQW1KBb7Xl0cXHpNFDXaDS4urpy9913s2XLFjQaDb6+vtx9992/yBhV5ezZs7i6utKzZ0+2bt1Keno6RUVFDBw4EKPRyNGjR/nuu++IiYlh6NChbN++3Wq5sWpJ0draKpbOG41GGhsbCQgIoLm5WSztV+uUmpoakQfq0vu6ujohz9/e2qH9/9UVdVcv8zeZTCxZskTI8P+za6r5Xltb2+Ea7YmMjGTGjBnAFZ0e1fqjPRaLRUgQtLdqaGhoEPXM1UydOpW+ffta9Uao0gZq/Wo2m0X9qvYSqelWLSDa12WqRYXRaBSyEWazmerqaqHMq0pEqN+MxsZGkWY1H9RrqXV5VVWVqLc/+ugjysvLrZZk19fXs2XLFuLj45k1a5aow9XePPU86nXUMtytWzeam5u5/fbbSUxMJDY2lieffFKsqvXz8yMhIQE7Ozu++eYbzGYzR48eZfXq1dft83Wjc0NZHxQXF7Nt2zbi4uLYsWMHzz77LDk5OZSVlVFVVSVUKFUFYQcHB0aMGEFLSwu7du3i8OHDPP/88zg4OHDp0iXS09OJiIhg+vTpYrJqamoq7777LnFxccTGxopu58LCQh599FFSUlIICQkhODiYw4cPC4XJhQsX8vnnn6PRaISnUkVFBWPHjiU5OZkBAwYIvQf1Wo2NjaxYsYLBgwezfft2Hn74YXJzc4WoUvsl06oQ2QsvvMDUqVN5+umnrZb8rl27lr59+3Lp0iXhYXPnnXdy+fJlcnNzcXd35/z58zzzzDMkJSXRrVs3NmzYwGuvvQZcWT22e/duRo8eTVNTExUVFeh0OiZMmHBDCTNJ/nMoisLGjRtxc3PjxIkTzJo1i507d1JYWIiPjw8tLS28/PLLrFy5kri4ODIzM+ndu3en59LpdNx3330888wzJCYm0tjYaKVMC1d6awsKCjp8pL29vcVyZY1Gw6xZs6ioqGDt2rXMnDmT7OxsHB0d0Wg0QoL/0qVLuLq6YmtrK7yt1OCppaWFZ555hldffVXYlUyZMoWNGzcSHx9PYWEhmzdv5siRI3zyySfCxDYpKYnExERWrFhBcHAwGo2GgQMHArB9+3Yee+wxMjMzcXFx4eTJk8yePZsLFy7g7e1NQUEB9913n/CiUnWToqKihORBZ9fcunUrc+bMYfPmzSI/hg0bhq2tLdu2bWPu3LnCFkZRFA4dOsS2bdtYsGABy5cvZ9CgQdTW1jJ+/HhR3xQWFvLSSy8xZcoU9uzZw1tvvUVRURHl5eXY2dmRnZ3N888/L5aaGwwGFixYwKBBgwgNDWXBggVMnTqVpKQk7r77bi5fvkxISAg9e/bkk08+YcyYMezbt4+//OUvJCUlCRG5sWPHUlpaipubGxcuXOD+++9n3rx5fPbZZ+Tk5NDW1kZrayt2dnZs3ryZ559/nvXr1wsn9Pz8fMaMGUNycjJ//etfycrKwmg0smPHDhITE9m/fz+NjY3CKf3hhx9m9+7djB07lsDAQHH/ZWVllJeXk5OTg62tLRs3bmTQoEEkJSUxe/Zsli1bJgRE582b1+lqXBsbG2677TY+/PBDUlNThb1NbW0tpaWljB49mry8PG666SZycnL44YcfmDhx4n/9ooobpodGURQ2bNhAUFAQI0eOxN3dnZSUFKKjowkLC6Nbt27iYasqk9HR0Xh6euLo6MjUqVMZOnQoFy9eZM2aNUJW/OooOjg4GJPJxAsvvEBjYyP5+fl4eHiI1sHOnTuxt7cXmgnNzc34+vpajaWqOi3dunXD09OTw4cPU1JSQr9+/azu6fDhw7S2tjJs2DAGDhzI6tWriY6OJjg4WChUqgQGBvLaa6+xatUqDhw4YCU/rigKWVlZHDx4kOHDhwvLhZ49e7Js2TKGDBnCxIkTxb03NjZy88038+STT+Lk5ITRaGTZsmXcfffdxMTEUF9fz+bNm4mJibmmyZ5Ecj306tWLsLAwqqqqhHZOQEAAs2fPJiMjgzNnzlBeXs6wYcOIjY3t9Bytra3Y29sLa5A9e/YIZd32tFfbbf/namEyjUaDg4MDISEhpKamCqVldZvBYMBoNFqpOrfvdXBzc2Pw4MGkpqbSvXt3Tp48SUlJifBYCwkJITExUejFrFu3ThjDqh9M9f1rv628vJxevXoRHh5OdXU1+fn5FBUVkZyczODBg63u19XVVdhv2NnZWV2zoKBAnLdbt27U1dXh5uZG3759efHFF9m0aZPYVlZWZpU34eHhwmj36NGjwguqPaoW0+TJk4mOjiYrK4tly5YxfPhw7rrrLn766ScrRWAHBwcRwAYHB2M2m5kyZQohISEEBQUxbdo0Tp06RXBwMA4ODsydO5e77rqLXbt24eHhQa9evXjllVdYtWoVQ4cO5Y477uD8+fO0tLTQq1cvUlNTqa+vp3v37mzevBmj0UhkZCS1tbV4enoSERFBQkIC5eXlTJo0iaCgINLS0lizZo3Q26qqqsLb25uwsDAeeeQRITipWteoZUij0RAZGYmPjw8DBgzg+++/x9bWluHDh4seP2dnZ2666SbmzZvXoYy2x9bWVuiQAVy+fJldu3Zx//3388ADD3DgwAG0Wi2xsbEsX75cznPkBgpoANHFCODv73/dwyFXS7dXV1fTt29fEhISmD59eqey4qpnhtr6euGFFwgMDOT5559n8eLF7N+/H8CqK7IzHB0dmT9/Prt372bFihVW2wwGg1AS9fPz+1lZcZWgoCB69erVwbxy7ty5GAwG3nzzTav5M+okYlW8rq6ujpKSEhRFEe7Eqnz+9u3bURSFwYMHM2nSJF577TWys7N/kYS7RNKekydPCrE69X1VJeoBoeoMnVsLAMJjCOC+++5j5cqVuLu7d2it6nQ6CgsLKSgosPqj1hlwpadz6dKlGI1Ghg0bxtq1a3F2drZy2w4KCsLHx0c4JNvY2Fh5EAFMmjSJ5ORk9Ho9BoOBQ4cOCUVW1YLCxsYGg8FAU1MTY8aM4fnnnxdaTHZ2dmLoZtSoUTz33HMMGDCAkydPUltbi4+PD4qiMG3aNMLCwvjzn/9MfX39NfO5vey/Xq+3uma/fv3QaDTY29uL4Xb1mmoPEfyPHYK675///Gd++OEHli5d2uH9V+tUGxsboQrc3jLj5+oL1b6jvXWA+rFWfwsICBBqv/b29mi1WqFCbmtrK7yQ7r//ftasWSPsRlpaWhg7diwvvfSSaBS2v177PKqvr2fkyJE8++yz3HTTTVbXvzr917ofvV4v6nD1m6RajFxtmXP1+dT5M2ogr36TJkyYIIYOAwMDGT16NJWVlVy4cOEXW2rcaNwwQ04ajYZ7772Xzz77jOzsbBoaGnjggQc4fPgwdXV1tLW1WSly+vr6curUKfr27UtLS4uQyHZ2duaWW27hzTffZOLEiURFRTF+/Hjgf2TJVTnrUaNG8ac//Yna2lr8/f3p1asXDQ0NzJkzh5qaGsLCwli1ahUeHh5CBVKVx/fx8eHHH38kNTWVrKwsnnnmGfbt22d1T6NGjSIlJYWLFy9y6dIlHnjgAWEvoNfrhSIpwK5du2hpaaFbt24MHTq0g6Hkvn37SEhI4JtvvrGyXHjggQfYs2cPrq6uhIWF8dBDD/Hkk0/y9ttvEx0dzbhx4zAYDEyfPp1FixYJ9dWoqCgmTZqEwWBg8eLFjB07lr59+/6bn7LkRsJisXDkyBGMRiOVlZWkp6djY2Mj5pypLezly5eLSf2qX5vasq2vr2fXrl3CdLVnz57Ex8d3MIKFK+/88OHDO/x+tfu1+o4WFBQwePBghg8fTk5ODrW1tZSXl3PPPfcQFBREZmYmADExMVZK4Wor3dPTU6jjqnL76jw01f4DrphAzp8/n379+jFkyBAxD2TQoEFERkby5z//mfj4eIYMGcKRI0fEnIyMjAyKi4uZOHEidXV1VvP91OGxH3/8EXd3dyt7DdXP65VXXiE2Npbhw4fT0NBAdXU1zs7OREREiGuOGDFCBAftbSqqqqo4deoUTz31FNu3b7fKv5aWFlpbW8V96vV6HnzwQXbt2oWiKPTu3Ztu3bqJ/dV5QA4ODmJuomobUFdXh729Pa2trbS1tVFVVUVmZiaXLl1i8uTJrFmzRgyzTZ8+nd27dwvX+piYGLRaLc7OzsTExODr64u7uzuvvfYavXr1YvTo0TQ2NuLg4CAsTNT63WKx0L17d+bPn098fDxDhw4VeVdfXy/sRGxtbTl69Ci33nqrCAhVCxt1KO7tt98mMzOTgoICpkyZwrJly8ScHzVIV+0Q9Ho9Z8+epa2tjYMHD/Lqq6/i6ekp5uDo9Xrs7OzYuHEj0dHRxMbGYjabGTx4MF999RXvvPMOixYt4vnnn7f6Bvy3cEPp0CiKQkVFBTU1NQQHB+Pu7k5RUREGg4Hw8HCrAKClpYVLly4REhJCVVUVAQEB1NfXi+GorKws4EplpY71KopCTU0NlZWVBAcH4+HhITwxVKfakpISMaSk0WhIT08XrU9PT09KSkpwd3fH29ubjIwMunfvLgzGAgMDcXZ2tpr7otoA+Pn54evrS0lJSQeBNFVKOysrS8h4q60OdXtxcTEGgwEvLy+8vb3JycnBwcGB4OBg0RJQ86iyspLCwkJiYmKEnYCPj4+oVDw9PWloaMDe3p7Q0FDy8/Px8/Pr0EqVSH4OtRKvqqrC19eX5uZmEbB4eHhQWVlJREQEjY2NFBcX4+vri6enJ56enlbzzFTbhKCgIODKSkAXF5dr9uj8szS1tLSQl5eHXq+nd+/eODs7U1lZSUVFhbCZsFgs5OTkYDAYiIqKwsXFpYNdRl1dHe7u7rS1tWGxWHBxcaG8vJyGhgYCAwMpKyvD19cXZ2dn0tPTCQgIwN/fn4KCApycnAgNDUWv15OWloafnx/h4eFUVFSI/NLpdLi6uoq/VWn99nnb0NCAm5sbTU1N4pp+fn44OTmJawYFBZGfn49WqyU8PJy2tjara7b/6FZUVAhJ/+rqasxmM4GBgVb339TUJOwFampqcHJyEtcwmUyEh4eLOUmAsM/QarW4urpSXV1NUFAQlZWVODk5YWdnJyYov/XWW7zyyiv4+/vj7e1Nfn6+sDqxsbEhPz9fuK+r9b06tKQa0WZkZBAcHExgYCAFBQXY2toKC4OgoCCqqqpwcXHB09OT9PR0/Pz8CAsLo7CwEECkMTIykoqKChRFITIyUswzqq+vp7S0VAztqz6BAQEB4pthb29PWFiYCMwVRaG0tFTYy6hDqOqQU35+vug1t7e3F95+gYGBmM1mCgoKMBgMREREUFRURPfu3a3sMP5buKECGolEIpHceCiKwpkzZ/jb3/7GV1991WHCt0QCMqCRSCQSyf9xFEWhqqqK+vp6AgMDrVZ4SiQqMqCRSCQSiUTS5bmhVjlJJBKJRCL570QGNBKJRCKRSLo8MqCRSCQSiUTS5ZEBjUQikUgkki6PDGgkEolEIpF0eWRAI5FIJBKJpMsjAxqJRPKboKqdqr5H9fX15OfnC+8wVWm2paVFeAbl5eWJ/0skEsm/ggxoJBLJv4zFYuHo0aNMnDiRnJwc8vLymDp1KomJiTzzzDMYDAbS0tKYNGkSx44do76+nhkzZvDEE08wd+5c9Hr9//YtSCSSLo4MaCQSyb9MdnY2CxYsoLGxEbPZjF6vJzExkc8++4zjx4+TnZ3Nm2++SU1NjXBfrqys5A9/+IPopZFIJJJ/BRnQSCSSf5mIiAiWLl0qXK579+7Nvffey+LFixk5ciTdu3dnyZIlxMTEAODt7Y2HhwdvvfUWQUFBVsaxEolE8muQAY1EIvmXcXBwwNXVVfxfr9czb948amtreeONN3B0dMTDw0Nsv3jxIq2traxfv57CwkIuX778v5FsiURyAyEDGolE8pthZ2eHjY0N+/btY9WqVeTl5fHss89SU1NjtT0kJASj0cgrr7yCk5MTQUFB/8spl0gkXR1pTimRSH4TzGYz2dnZhIeH09jYSElJCQC2trb06tULOzs7cnJy8Pf3x83NjZKSEnJycujRowdBQUFoNJr/5TuQSCRdGRnQSCQSiUQi6fLIISeJRCKRSCRdHhnQSCQSiUQi6fLIgEYikUgkEkmXRwY0EolEIpFIujwyoJFIJBKJRNLlkQGNRCKRSCSSLo8MaCQSiUQikXR5ZEAjkUgkEomkyyMDGolEIpFIJF0eGdBIJBKJRCLp8siARiKRSCQSSZdHBjQSiUQikUi6PDKgkUgkEolE0uWRAY1EIpFIJJIujwxoJBKJRCKRdHlkQCORSCQSiaTLIwMaiUQikUgkXR4Z0EgkEolEIunyyIBGIpFIJBJJl0cGNBKJRCKRSLo8MqCRSCQSiUTS5ZEBjUQikUgkki6PDGgkEolEIpF0eWRAI5FIJBKJpMsjAxqJRCKRSCRdHhnQSCQSiUQi6fLIgEYikUgkEkmXRwY0EolEIpFIujwyoJFIJBKJRNLlkQGNRCKRSCSSLo8MaCQSiUQikXR5ZEAjkUgkEomkyyMDGolEIpFIJF0eGdBIJBKJRCLp8siARiKRSCQSSZdHBjQSiUQikUi6PLb/2wm4kdDr9eh0Ory9vdFoNFbbFEVBURSamppwd3fvsL39fo2Njbi5uWFj8/PxpsFgwGw24+zsjNFoFP+/nmOvF4vFQnNzM25ubtdM87VQ78XOzg6NRoOTk1OH7U1NTbi4uIj0dnaNpqYmnJycsLW1RVGUa+7X/rz19fXU1NQQGBiIi4tLp/vr9XoURbFK1/WcX6W5uRk7OzscHBxobm7G3t4ee3t7sb21tRWNRoOjo+M/PVf76zc2NuLu7v6z6VAUhZqaGlpaWnB0dMTHxwetVvuLrtPS0oK9vT12dnbXlabOypX6DF1dXa+7zDU0NKDVanF1dcVsNlNTU4Ovr+8/Pd5sNtPQ0ICzszMODg7AlTx2cnIS+XT1Pq2trRiNRtzd3bFYLDQ0NODk5CSeydXHX31v8D/PQD3e2dkZe3t7NBoNBoMBjUYjnvvV+7S2ttLW1oa7u7vV/RkMBmxtbdFoNDQ3N4tyaG9vj6IoNDQ0iHqivr4euPIeqOkJCAjAaDRSVlaGj48Pnp6eaDQa8TxsbW1xdna+ruehYjQaaW1txcbGhtraWgDs7e3x9vamsrISi8Uirl1bW0tbWxsAXl5euLm5YTKZaGxsxMvLC5PJRHNzs0gXQFtbG0ajERcXl1+ULhWLxUJ5eTkmkwkHBwd8fX2vWebVZweIfFHfKzU96j6tra1otVpRpn5r2tfTkn8vsofmN0JRFDZv3syzzz6LwWAQAYyiKFgsFlJSUigvL+fLL7/EYDBgsVhQFAWTyYTJZBL7mc1mjh07hsFg6LCtra1NnNNoNJKamsqBAwcAyMnJYfv27fz000/iuM6OVa9hNBpRFMXq34qi0NbWZpW2uro6vvnmG3GP6jb1b6PR2Ol1VL766iuysrLIz8/HYrFgNBqt9jt9+jQlJSXs2rULwCrdanrWrFlDdXU1iqJw4sQJcnNzxTaz2WyVHkVRyMrK4v3332f//v28/fbbNDQ0iGuq+5nNZk6ePMmPP/4ozqUoCpWVlRw4cKDTe1PTrrJlyxYyMjIA2LRpE+np6Vb75eTkUFxc3CHPrk6H+sdoNGI2mzl69ChpaWmcO3euw7Htef/999m4cSNLly5lxYoVHZ7t1cdaLBaRDwDbt28nJyen07J19X0vW7aMlpaWDuXKaDTy9ddfi4+tenz753J1vi1dupRPP/0URVHIzMxk9uzZtLS0/GzaFUVhw4YNLFmyhHfffZfm5mbOnTvHggULxLkVRWHFihV88cUXvPvuuxQWFrJgwQI++OADzpw5w+rVq1myZAnvvPMO9fX1XLhwgb/97W+YzWar91h9PtXV1ezbt09cf/fu3Xz66ad8+OGHGAwGjEYjf/vb3zh79qxI6549e/j000/54IMPKCoq4r333mPRokUcOnRI5H9FRQXvvvsuJSUlVFZW8vLLL/Pxxx9z4cIF4EpQtmjRIvR6PRaLhYULF3L+/HnWrVvHiy++yO7du8nNzeWdd95h//79vPvuu+Tl5YljP/zwQ8rKyqzKVvv3v339oD5LgMLCQlauXMm2bdtYsmQJycnJ/PDDD2RnZ/PGG2+QnJzMnj17qKmp4dVXX2X79u0kJydTUFAg3t2FCxdSVFTEli1bxLulloecnBx27NjR4R1Q09RZ/dGepqYm5s+fT3JyMosXL+a7777rtLyp5Xvnzp3i3TcYDBw9etRq/8bGRnbt2kVmZiYVFRU/W092lrb2dUpn+6vXSUtLY+/evT/7Lkt+G2QPzW+E0WjkwoUL9OjRg6ysLFFwfX19SU1NZcuWLTQ2NpKens6nn36Ks7Mzt99+OytXrkSj0fC73/2O5ORkjEYjnp6eZGdns2/fPvLy8njyySc5ePAglZWVDBs2jMDAQNavX09dXR1jx44FrlQmer2e2tpaSkpK+PbbbwGYOXMm27Zto76+nltuuYWWlhYyMjKoq6vjkUceYevWreh0OqZNm0ZBQQFnzpwhKCiIyZMn8/nnn4tKD670aCQlJXH//fezbt06Bg4cyPr169FqtcyZM4e1a9dSW1vLyJEjufXWWwFoaWmhpaWFtrY2Fi5ciMlkoqmpCWdnZ3r06IGtrS3nzp1j3bp1REREsHnzZhRF4ZFHHuHHH38kOzubrKwsJk2ahMViYceOHQQGBtKnTx8OHjyIu7s7c+fOZfv27UybNg17e3s2b97M7Nmz6dGjB2lpaRgMBr799luKi4sJDw8nOjqalJQU/Pz8CA8PZ/PmzaSlpREeHo63tzdJSUm4u7uzfft2bG1tmT17NmvXrsXGxgZPT09+97vfYWNjg16vt8qbpqYm3n//fbRaLV5eXsTGxlJcXMzFixe57bbb2LJlC8HBwRw5cgRvb29uu+02NmzYwJAhQ7hw4QI6nY7777+fmpoaMjMzqa6u5uzZszz88MNs2LCBadOmWbXybGxsmD17Nm5ubrzyyivk5uaydu1a2trauP3228nLyyM2NpYjR44wfPhwioqKSE9PByAmJga9Xk9LSwv/+Mc/qKur4+abb8bHx4ddu3ahKApPPvkka9eupaamhqysLIqLi9mwYQOKopCQkMDevXupqKggKyuL2bNnoygKaWlprF+/Hjs7O2bPns13332HjY0NPj4+zJo1CxsbG2xsbMjJyaG5uZljx47h5eVFbW0ta9asQa/XM3HiRC5dusSDDz7IunXrePDBB7G3tycwMJCxY8fyySef0NDQQFVVFTU1NSI/zGYznp6eJCQk8NFHH7Fr1y7i4+MZMGAA69ato0ePHrzwwgt8+umnlJaWUlVVRVVVlfi4KIrCrl27OHv2LC4uLkRHR7Np0yaGDRuGk5MTR44c4eWXX2b58uUUFBRQWFhIeXk5RqOR7OxsioqK+PHHH8U+O3bsoF+/fowePZpFixaJHi4XFxfKy8sxm81UVFTg7e3NiBEj6NWrl7gXnU4n0qXT6Rg8eDC9evWiubmZRx99lO+++46xY8cybtw4ioqKMBqNAJSVlXHu3DlGjRrFypUrGTp0KOnp6TQ2NjJu3DhaW1tJTU2lsrKS4OBgGhsb+eMf/4izszMWi0X0KA0dOpRBgwbh4OBAZWUlERERTJ48GRsbGzw8PHB0dOTuu+/G1tYWb29vABwcHHjwwQdZsGABvr6+zJs3j08++YSmpia0Wi233norer2e7du3c/r0aQICApg0aRLffPMN8fHxlJWVUVFRwaBBg7jjjjs69Jqp9en06dPJzc1lz549ZGRksGHDBuzs7EhMTGTDhg3U1dUxatQo1q5di8lkIiMjA39/f+zt7cnMzGT9+vU4ODgQHx/Pt99+S0JCAo6OjixdupSysjIGDBiAm5sbZ8+epaGhgVmzZrF//34qKioYOXIkY8eORVEUUaeEhYUREBDA8ePHMRgMJCQksHPnTvR6PY6OjgwdOpTa2lrWrl3Lvffey4YNG0hISPhFPaqS60P20PxG5ObmUlVVhbu7O3v27KG6uprq6mp0Oh3V1dX079+f+Ph4vLy8mDt3LhUVFWzfvp2pU6cyceJE9u3bR05ODg899JD48I4ZM4aIiAgMBgPp6enExcWRlJREUlISc+bMYfz48R0i/ezsbPbs2cMdd9xBQkICqampFBQU0KdPH7Zs2UJubi4jR45k4MCBXLhwgeLiYuLj49FqtSQlJdG/f39OnTrFxo0bGTJkCImJieLFM5lMZGdnoygKly9fpqCgAEVRGDBgANnZ2WRnZxMbG0tSUpJVq7empobKykoqKyt59NFHMRgMPProo2RlZZGXl0d4eDg33XQTP/zwA56ennh4eLBp0yYyMzP505/+RHh4OHDlAx4XF8fAgQPZv38/zz//vAj+pkyZIrr9Gxsb8fHxQaPREBUVhZOTE0FBQfTr14+ffvqJ0tJS+vTpQ+/evWlqahIfviNHjuDv78/gwYP5/vvvmT17NkOGDOHQoUNUVVXx+OOPU1hYKO7t6q5rs9lMVVUVc+fOpaCggLKyMjQaDadPn+bEiRPY2dmxefNm+vXrx6VLlzh//jzu7u7079+fkpISBg4ciJOTEzk5OfTr14/hw4fT2trKiRMnKCsru+bQla2tLU5OThw8eJD4+HgeffRRDh48SHZ2NkeOHCE1NZVDhw7h6uqK0Whkzpw5pKWlAVd69uzs7HjxxRc5dOgQWq2WgQMHUllZyenTp2lqauLFF1/Ex8eH5ORkPDw88PLyYuXKlVRXV/PHP/4RPz8/kZbk5GR+97vfMXDgQA4fPkx1dTWPP/44eXl5It+0Wi0xMTGcOXOGyspKwsPDURSFvn374uPjQ2pqKtXV1Zw4cYK6ujrxXEePHs2FCxeoqqrC3t6eUaNGiaE59bzjxo1j1apVeHl5ibS6urqi1+sZP348a9euxdHRkejoaEaOHImXl5dVXnp4eIh3IywsjEGDBuHq6ip6ERwdHXF1dSUzM5Njx44RFxdHQ0MDERERDB48WOzj5uZGWFgYZ8+e5csvv6S8vJzY2FhiY2OJiYkRwYuTkxNxcXFUV1ezZs2a6265V1dXExgYKIaAfHx8UBSFgIAA4uPj8fX1xdnZGR8fH7y9vXnuuefYu3cvRUVFjBkzhpCQEG677TZcXFzEkJaKxWLh8OHDbNmyhSNHjqAoChkZGWzZsoXk5GRaW1uprq5m69atbN26ldLSUvE+hIeHU1paSnh4OHZ2duTk5DB37lyCgoK4ePEier2eY8eO8ac//YmWlhYuXbqEoij07NmT8+fPExcXx7Zt20Qvy9VkZ2ezZMkSlixZQt++fdm4cSMRERGYTCa+//57cnNziYuLIzg4mPj4eHr37k1lZSXTpk0jJyeH5ORkZs2axaRJk4iMjGTQoEG0traSkZFBY2MjL7/8MidOnCAnJ4eBAwdy8803k56ezqVLl+jTp48o64qiWNUpeXl5jBgxghkzZrBz507y8vJ46qmncHFx4dKlSzg5OZGbm8uxY8cwm82/2ZQAiTUyV38DFEXh4MGDjBo1igEDBpCTk0NraysGg4Hq6mrMZrP4+Hl6euLg4IBWq0Wj0YguYY1Gg6urq6igS0tL2bJlCwkJCdja2uLp6UloaChjxozBzs7OquekMywWC3q9ntbWVry9vQkLC2P06NHY2dnh5eWFvb09tra23HXXXdTV1bFnzx6cnZ0JCQkRFb3JZBJduXClwjIajeh0OhoaGggKCmL06NH89NNPXLp0CS8vL0JDQxk9enSnL6yTkxMODg64urqKOQjqeeFKwBIQEEBMTAz9+/e3GiZR92vfamufd+27gyMjIzl16hR6vZ6vv/6aEydOsHfvXnx8fLC3txfXUc/l4uJCSEgIo0ePxs3NDUVROjyb9s9NvY6LiwulpaXiOTs7O1vtp95zWFgYmzZtYuTIkTg6OhIcHMywYcPw9fUlMDBQtHarqqrYt2+fVZ4MGzaMxYsXM2LECAwGAzqdzuoZNzU1kZeXJ4I4Nc12dnZERUVx/vx5+vXrR0ZGhuiBsrOzs5qzpB6jKAp79uzBYrGID6TalW6xWNBqtfj7+9OzZ08GDBgghszaDyfZ2Nh0yDd7e/sOrdHBgwezfv16IiIisLe3Jz09ncuXLxMYGAjA0KFD+eyzzxg5cqR41itWrGDo0KHExsZSVlZmdb7m5mZaW1v57LPPCA4O5v777yckJISSkhIqKirw8vLi888/x8fHh+nTp2Nr27Fz2mQysWPHDlxdXXF1dbUKLrRaLU5OTtTV1VFdXS3e1bNnz3Lu3DlR/trvExwczF133cXEiRPp3bu3CIquvmZ8fDyDBg0S81Y0Gg02NjbU1NTQ0NCAyWTqkH+9evXi6NGj6PV69uzZw969e8U29T0JDAzE1tbW6h3SarU4Ozvj5OSEi4uL1by09s9w4sSJPPLII6KnNT4+nkceeYQHHngAR0dHAgMDmTFjBjNnziQkJESUlTVr1jB79mxycnLIzs4W5ad9I6B9uVIDMrWOCwkJ4ZZbbqGtrY3W1tYOz6hXr1786U9/Yt68eWRkZGBraysCi5iYGO68807a2trYsWOHOMbHx0fMj1GvrdPpRK+W+r6raVIURfSyqnOaJk6ciK2tLZs2bRLzlPbu3Yuvry8ODg5iiF69p/bvjZrv/fv358svv2T06NHodDoMBkOH+5P8a8ghp98Ai8WCp6cnU6ZMwcnJifr6eoKCgkhOTsbT05MBAwZQX19PZmYmcXFx2NjY0L9/f/r168eqVauwtbUlMTGR48ePo9Vq6devH/n5+Wi1WtavX8/tt99OaGgoKSkpjBw5kn79+rF8+XKcnJwYP348cCVQio6OxtvbmwEDBrB69Wo0Gg0JCQlUVVWxf/9+xo0bh7e3N25uboSGhmI0Gtm/fz8mk4kJEyYQExPD7t27iYiIYMKECXzxxRdcvHiR+Ph4AFGRff311/Tu3RsbGxsOHz6Mq6sr48aNIykpiX379nHLLbeID3JcXBxhYWHAlcrbzs6O/v37Y29vT1xcHHZ2dvj4+Iihhi1btqDVapk5cyZ1dXV88MEH+Pv7i4m7YWFhZGVlMXHiRD799FP8/PyIjo5m165dopfmwQcf5KuvvuLo0aMEBwczZMgQLl68yJkzZ3B1dcXFxYWAgAAURcHDwwMnJyd27dpFdHQ0fn5+NDQ0cPvtt7NmzRqcnJyYOXMmJ0+eFM9NDQYmTJjA119/zblz5wgPD6dXr16Ul5eL/cLCwnB1dSUkJAStVktISAiTJ09m7969BAUFERkZiZubG7a2tpw6dQqLxcLtt99OcXExQUFBHDhwgJtvvhlHR0f69+9Pbm4uLS0tDB48GIAePXqwatUq7O3teeyxxwgJCWHJkiWcP3+eBx54ADs7O7RaLfHx8fj5+eHt7U2vXr2ws7MjLi5OBMmVlZX84x//YOLEieh0Oi5cuCAqcn9/fxYtWkSPHj2YNGkS3333HVqtloSEBOrq6li4cCHdu3cXvSiTJk3i22+/xcXFhRkzZnDq1KkO+da7d2+io6Pp27cvI0eOxMnJiaioKE6fPi3mfvTs2RMnJyf69OmDRqNBq9USFBTEwoUL8fHxoWfPnqI3SaPRiN49i8XCqVOnuHDhAtOnT+f8+fNs376d++67jw0bNnDu3DnS09OZPXs23t7eDBw4UKRL7Tk6deoUHh4e6HQ6mpqaaGpqws3NjcmTJ/P1118TEhLCmDFjuO222zhx4gSenp5UVFRQVlbGlClT+PrrrwkODiY6OpqvvvoKg8HAww8/zKVLl4ArwUGfPn1wdXXF0dGRb775BkVRePDBB4ErAcWDDz7I8uXLURSFqVOn4uDggJOTE/Hx8Wg0Gm655RZWrFjB3/72N1xcXJgzZ44IhPr374+3tzfdu3enT58+/PjjjyxevJhJkybR1taGp6cnPXv2xM3NjT59+ohhTDc3N/r27YuLiwsHDhzg9OnTuLu7M2HCBBobG1m8eDEajYZp06YRFBTE0qVL0Wg0DBs2jDFjxojJ0HfeeSexsbGcPXsWo9HIF198gaurK5MnT6aiooLw8HA+/PBDIiMj6dWrFxqNhoiICLp160ZKSgrDhg3j8uXL2NraEhcXJ+pZe3t7brrpJjQaDT4+PoSGhjJgwAA2bdqEs7MzI0aM4NChQ7S0tDBy5EgqKirIzs4mLi5OBBR9+vRhzZo1uLq6Mn36dAwGA15eXkRGRlJfX89HH33ELbfcgru7O97e3jg6OuLp6cmxY8eoq6tj2LBhYoFDSEgIp0+fxtXVlbq6OrKzs3FxceHhhx/m/fffZ/HixXh4eDBixAiampoICAhg165dREREcPToUUJDQ+nWrdu/4Yv034tGkbOT/mU6m1EPiFUB6m+KonTouVAj+OvpgjSbzaKlpnZbtj//tc6rtko6G7O9elv7a7RPc/t7aH8fV9+jeq7O0nQ993Z1frRPT/trqD0BatpU2qfTbDaLlrj6W/tz/1zeqmm5uleos+tcK2+vdUz7a6l0dh6z2cyWLVtobW1l5syZVudRj7maa5Wzn6OzMqCWGzWf1W3X+3z+Wb51hnqcxWLhu+++w8HBgfvvv9/qfjvLO7gyodXd3R1PT89flR9XX6N9+W5fnju736u51jt0rTxon6ft03H1753VM2rvzbXqgfbn+rky+kvf1+s53mKx8OGHH/L000/j6OjYoXe1s/Rc/Xv7Y9r3FF+df53VQT9X56r7X329a6Wrs7qtfZ2yfv164uPj6dmzp5jI/eyzz4qeaLPZzNKlS4mOjhaN0KvvT/KvIwMaieT/KIpyZamps7PzP11afSOh3rc6JHK9x4D8QPxfQh2a+TWSD10NnU4nJBA6u29FubIU383NTU4G/jciAxqJRCKRSCRdHjkpWCKRSCQSSZdHBjQSiUQikUi6PDKgkUgkEolE0uWRAY1EIpFIJJIujwxoJBKJRCKRdHlkQCORSCQSiaTLIwMaiUQikUgkXR4Z0EgkEolEIunyyIBGIpFIJBJJl0cGNBKJRCKRSLo8MqCRSCQSiUTS5ZEBjUQikUgkki6PDGgkEolEIpF0ef4fWx9WzQA+ulwAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 1300x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import fitz \n",
    "\n",
    "# open pdf and load target page\n",
    "pdf_path = \"maiti2020.pdf\"\n",
    "doc = fitz.open(pdf_path)\n",
    "page = doc.load_page(3)\n",
    "\n",
    "# get the image of the page\n",
    "img = page.get_pixmap(dpi=300)\n",
    "\n",
    "# optional: save the image\n",
    "img.save(\"lstm_result.png\")\n",
    "doc.close()\n",
    "\n",
    "# convert the pixmap to a numpy array\n",
    "img_array = np.frombuffer(img.samples_mv, dtype=np.uint8).reshape((img.h, img.w, img.n))\n",
    "\n",
    "# display the image using Matplotlib\n",
    "import matplotlib.pyplot as plt\n",
    "plt.figure(figsize=(13,10))\n",
    "plt.imshow(img_array)\n",
    "plt.title(f\"Query: '{query}' | Most Relevant Page:\")\n",
    "plt.axis('off') # turn off axis\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the workflow is the same \n",
    "\n",
    "`ingest documents -> split into chunks -> embed chunks -> make a query -> embed the query -> compare query embedding to chunk embeddings`\n",
    "\n",
    "that is the `retrieval` part of retrieval augmented generation (rag)."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Similarity measure: dot product and cosine similarity"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "the main difference between dot product and cosine similarity is that, cosine similarity has a normalization step.\n",
    "\n",
    "| similarity measure | description | code |\n",
    "| :------------------|:------------|:----:|\n",
    "| dot product | -measure of magnitude and direction between two vectors.<br>  -vectors that are aligned in direction and magnitude hav a higher positive value <br>-vectors that are opposite in direction and magnitude have a higher negative value. | torch.dot, np.dot, sentence_transformers.util.dot_score |\n",
    "| cosine similarity | - vectors get normalized by magnitude/ euclidean norm /l2 norm so that they have unit length and are compared more so on direction. <br>-vectors that are aligned in direction have a value close to +1.<br> -vectors that are opposite in direction have a value close to -1. | torch.nn.functional.cosine_similarity, 1-scipy.spatial.distance.cosine (subtract the distance from 1 for similarity measure), sentence_transformers.util.cos_sim |"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For text similarity, you generally want to use cosine similarity as we are after the semantic measurements (direction) rather than magnitude.\n",
    "\n",
    "in our case, our embedding model all-mpnet-base-v2 outputs normalized outputs so dot product and cosine similarity return the same results, However, dot product is faster due to not need to perform a normalize step.\n",
    "\n",
    "to make things bit more concrete, lets make a simple dot product and cosine similarity functions and view their results on different vectors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 67,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Dot product between vector1 and vector2: tensor(14.)\n",
      "Dot product between vector1 and vector3: tensor(32.)\n",
      "Dot product between vector1 and vector4: tensor(-14.)\n",
      "cosine similarity between vector1 and vector2: tensor(1.0000)\n",
      "cosine similarity between vector1 and vector3: tensor(0.9746)\n",
      "cosine similarity between vector1 and vector4: tensor(-1.0000)\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "\n",
    "def dot_product(vector1, vector2):\n",
    "    return torch.dot(vector1, vector2)\n",
    "\n",
    "def cosine_similarity(vector1, vector2):\n",
    "    dot_product = torch.dot(vector1, vector2)\n",
    "\n",
    "    # get euclidean/l2 norm of each vector (removes the magnitude, keeps direction)\n",
    "    norm_vector1 = torch.sqrt(torch.sum(vector1**2))\n",
    "    norm_vector2 = torch.sqrt(torch.sum(vector2**2))\n",
    "\n",
    "    return dot_product / (norm_vector1 * norm_vector2)\n",
    "\n",
    "# example tensors\n",
    "vector1 = torch.tensor([1,2,3], dtype=torch.float32)\n",
    "vector2 = torch.tensor([1,2,3], dtype=torch.float32)\n",
    "vector3 = torch.tensor([4,5,6], dtype=torch.float32)\n",
    "vector4 = torch.tensor([-1,-2,-3], dtype=torch.float32)\n",
    "\n",
    "# calculate dot product\n",
    "print(\"Dot product between vector1 and vector2:\", dot_product(vector1, vector2))\n",
    "print(\"Dot product between vector1 and vector3:\", dot_product(vector1, vector3))\n",
    "print(\"Dot product between vector1 and vector4:\", dot_product(vector1, vector4))\n",
    "\n",
    "# calculate cosine similarity\n",
    "print(\"cosine similarity between vector1 and vector2:\", cosine_similarity(vector1, vector2))\n",
    "print(\"cosine similarity between vector1 and vector3:\", cosine_similarity(vector1, vector3))\n",
    "print(\"cosine similarity between vector1 and vector4:\", cosine_similarity(vector1, vector4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Notice for both dot product and cosine similarity the comparisons of `vector1` and `vector2` are opposite of `vector1` and `vector4`.\n",
    "\n",
    "comparing `vector1` and `vector2` both equations turn positive values and for `vector1` and `vector4`, the result is in negative direction.\n",
    "\n",
    "This makes sense because `vector4` is the negative version of `vector1`.\n",
    "\n",
    "whereas, comparing `vector1` and `vector3` shows a different outcome.\n",
    "\n",
    "For dot product, the value is positive and larger than the other two comparisons.\n",
    "\n",
    "however, for cosine similarity, thanks to normalization step, comparing `vector1` and `vector3` results in a positive value close to 1 but not exactly 1.\n",
    "\n",
    "It is because of this that when comparing text embeddings, cosine similarity is generally favoured as it measures the difference in directiion of a pair of vectors rather than the difference in magnitude.\n",
    "\n",
    "And it is this difference in direction that is more generally considered to capture the semantic meaning/vibe of the text.\n",
    "\n",
    "The good news is that, as mentioned before, the outputs of our embedding model `all-mpnet-base-v2` are already normalized.\n",
    "\n",
    "So we can continue using the dot product (cosine similarity is dot product + normalization).\n",
    "\n",
    "With similarity measures explained, let's functionize our semantic search steps from above so we can repeat them."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Functionizing our semantic search pipeline"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets put all of the steps from above for semantic search into a function or two so we can repeat the workflow."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [],
   "source": [
    "def retrieve_relevant_resources(query: str, embeddings: torch.tensor, model: SentenceTransformer=embedding_model, n_resources_to_return: int=5, print_time: bool=True):\n",
    "\n",
    "    # embed the query\n",
    "    query_embedding = model.encode(query, convert_to_tensor=True)\n",
    "\n",
    "    # get dot product scores on embeddings\n",
    "    start_time = timer()\n",
    "    dot_score = util.dot_score(query_embedding, embeddings)[0]\n",
    "    end_time = timer()\n",
    "\n",
    "    if print_time:\n",
    "        print(f\"[INFO] Time taken to get scores on {len(embeddings)} embeddings: {end_time-start_time:.5f} seconds.\")\n",
    "\n",
    "    scores, indices = torch.topk(input=dot_scores, k=n_resources_to_return)\n",
    "    return scores, indices\n",
    "\n",
    "def print_top_results_and_scores(query: str, embeddings: torch.tensor, pages_and_chunks: list[dict]=pages_and_chunks, n_resources_to_return: int=5):\n",
    "    print(f\"Query: {query}\")\n",
    "    print(\"Results:\")\n",
    "    scores, indices = retrieve_relevant_resources(query=query, embeddings=embeddings, n_resources_to_return=n_resources_to_return)\n",
    "\n",
    "    for score, index in zip(scores, indices):\n",
    "        print(f\"Score: {score:.4f}\")\n",
    "        print_wrapped(pages_and_chunks[index][\"sentence_chunk\"])\n",
    "        print(f\"Page number: {pages_and_chunks[index]['page_number']}\")\n",
    "        print(\"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] Time taken to get scores on 21 embeddings: 0.00020 seconds.\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(tensor([0.5522, 0.5183, 0.5035, 0.5023, 0.4935], device='cuda:0'),\n",
       " tensor([15, 12,  4, 14, 11], device='cuda:0'))"
      ]
     },
     "execution_count": 69,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "query = \"long short term memory\"\n",
    "\n",
    "# get just the scores and indices of top related results.\n",
    "\n",
    "scores, indices = retrieve_relevant_resources(query=query, embeddings=embeddings)\n",
    "scores, indices"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 70,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Query: long short term memory\n",
      "Results:\n",
      "[INFO] Time taken to get scores on 21 embeddings: 0.00022 seconds.\n",
      "Score: 0.5522\n",
      "C. 210 days Look-back We could test the model LSTM only on M ∈{400, 700} and N =\n",
      "300 as 210 days look-back demands more sophis- ticated computational resources.\n",
      "The result is presented in Tables IX and X. The best result is obtained with M =\n",
      "700 and N = 300 whereas there is no improvement for DPA.1218 Authorized licensed\n",
      "use limited to: Carleton University. Downloaded on May 31,2021 at 04:09:15 UTC\n",
      "from IEEE Xplore. Restrictions apply.\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.5183\n",
      "TABLE II HYPER-PARAMETER VALUES FOR LSTM MODEL Internal Units Learning rate\n",
      "Dropout Rate Batch Size Epochs 495 0.00385 0.35302 60 600 Nevertheless,\n",
      "minimizing adversarial loss solely cannot assure satisfying predictions. For\n",
      "example, G might produce samples to “confuse” D, without being similar to ¯ YT\n",
      "+1, and then D will train to distinguish these samples, directing G to generate\n",
      "other “confusing” samples, and this futile process will continue. To ﬁx this\n",
      "issue, the generative model G must decrease the prediction error loss; that is,\n",
      "Lp loss LP ( ¯ Y , ˆ Y ) =| ˆ Y −¯ Y | (8) Besides, direction of stock price\n",
      "prediction is also critical for trading, so we deﬁne direction prediction loss\n",
      "function Ldpl as: Ldpl( ¯ Y , ˆ Y ) =| sgn( ˆ YT +1 −YT ) −sgn(YT +1 −YT ) | (9)\n",
      "where sgn represents sign function. We combined all these losses deﬁned, and\n",
      "achieve the ﬁnal loss on G: LG(X, Y ) = LG adv( ˆ Y ) + LP ( ¯ Y , ˆ Y ) + Ldpl(\n",
      "¯ Y , ˆ Y ) (10) We perform one SGD step on G to minimize LG(X, Y) while keeping\n",
      "the weights of D ﬁxed. The values of the hyper- parameters for G are mentioned\n",
      "in Table III.b) Discriminator: The Discriminator is a simple dense layer of\n",
      "leakyRelu activation functions followed by a sigmoid function layer. The only\n",
      "job of D is to determine if the input is ¯ Y (real) or ˆ Y (generated), our\n",
      "target loss here is equal to the adversarial loss for D. We keep the weights of\n",
      "G ﬁxed, and perform one SGD iteration on D to reduce the target loss: LD(X, Y )\n",
      "= LD adv( ¯ Y , ˆ Y ) = Lsce(D( ˆ Y ), 0) + Lsce(D( ¯ Y ), 1) (11) The values of\n",
      "the hyper-parameters for D are mentioned in Table IV. F. Evaluation Metrics\n",
      "Based on a speciﬁc model, for every stock at each time t, a prediction is\n",
      "obtained for the next time point t+1. Assuming the total number of time points\n",
      "to be tested is T0; we employed the following measure to gauge the performance\n",
      "of the different models.1) Root Mean Squared Relative Error (RMSRE): RMSRE = v u\n",
      "u t 1 T0 T0 X t=1 ˆ Yt+1 −Yt+1 Yt+1 !\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.5035\n",
      "LSTM and GAN models to predict the stock prices on the Indian market and compare\n",
      "their performance. A. LSTM LSTM can be trained to learn dependencies ranging\n",
      "over very long time intervals of time. It beats the vanishing gradients problem\n",
      "faced by a general RNN by substituting the ordinary neuron by a complex\n",
      "architecture called the LSTM unit or block [3]. The working of LSTM can be\n",
      "summarised by the following set of equations.zt = tanh(W zxt + Rzht−1 + bz) it =\n",
      "σ(W ixt + Riht−1 + bi) ft = σ(W fxt + Rfht−1 + bf) ot = σ(W oxt + Roht−1 + bo)\n",
      "st = zt · it + st−1 · ft ht = tanh(st) · ot (1) where it denotes the input gate\n",
      "and ot denotes the output gate. The forget gate, memory cell, and hidden state\n",
      "are denoted by ft, st, and ht, respectively. The σ and tanh functions are deﬁned\n",
      "in 2 and 3 respectively.σ(z) = 1 1 + e−z (2) tanh(z) = ez −e−z ez + e−z (3) B.\n",
      "GAN Generative adversarial networks was introduced by Good- fellow et al. [4],\n",
      "where image patches are generated from random noise using two networks trained\n",
      "simultaneously. It has shown very good results when applied to various tasks\n",
      "particularly image synthesizing.\n",
      "Page number: 1\n",
      "\n",
      "\n",
      "Score: 0.5023\n",
      "Tables V and VI show the average values of RMSRE and DPA with different (M, N).\n",
      "The numbers clearly indicate that LSTM has performed better than GAN model in\n",
      "terms of RMSRE metric. Although we have used speciﬁc loss function component for\n",
      "direction in GAN there has been no signiﬁcant improvement in DPA. There is\n",
      "strong possibility that the models have not achieved the optimum set of weights\n",
      "to solve the optimization prob- lem and produce good results. The best RMSRE has\n",
      "been observed for (M, N) = (700, 300), in LSTM model. B. 90 days Look-back We\n",
      "have also tested model LSTM on past 90 days data. For this purpose we have taken\n",
      "M ∈{400, 700, 1200} for N = 300 and M ∈{800, 1400, 2400} for N = 600. The result\n",
      "is shown in Tables VII and VIII. We can observe that the best RMSRE was again\n",
      "obtained for M = 1200 and N = 300. The model performance for DPA is very poor.\n",
      "Page number: 3\n",
      "\n",
      "\n",
      "Score: 0.4935\n",
      "The green bar shows the entire dataset, the blue bar shows the training set for\n",
      "one round experiment, and the yellow bar is the corresponding testing set. Image\n",
      "Source- [5] D. LSTM Model Architecture We have used a single layer of LSTM\n",
      "internal units followed by a Dropout layer. The activation function in LSTM is\n",
      "tanh and the loss function used is Mean Squared Error. The values for the hyper-\n",
      "parameters are mentioned in Table II. E. GAN Model Architecture As discussed\n",
      "earlier a GAN is a network of 2 models- generator G, and discriminator D. Here,\n",
      "we used the loss functions deﬁned and used by Zhou et al. [5].a) Generator: We\n",
      "have used a layer of LSTM followed by a dropout layer as our generator. To make\n",
      "the discrimina- tive model D as “confused” as possible, so that D does not\n",
      "distinguish between actual and closing price, the generative model G should have\n",
      "an adversarial loss which should be minimized such that D will not discriminate\n",
      "the prediction correctly. Consequently, we label actual closing price ¯ Y as\n",
      "class 1 and predicted closing price ˆ Y as class 0, and deﬁne the adversarial\n",
      "loss for G as 6 LG adv( ˆ Y ) = Lsce(D( ˆ Y ), 1) (6) where, Lsce is the sigmoid\n",
      "cross-entropy loss, deﬁned as 7 Lsce(A, B) = − X i Bi log(sigmoid(Ai)) + (1 −Bi)\n",
      "log(1 −sigmoid(Ai)) (7) 1217 Authorized licensed use limited to: Carleton\n",
      "University. Downloaded on May 31,2021 at 04:09:15 UTC from IEEE Xplore.\n",
      "Page number: 2\n",
      "\n",
      "\n"
     ]
    }
   ],
   "source": [
    "print_top_results_and_scores(query=query, embeddings=embeddings)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Semantic search/vector search extensions\n",
    "\n",
    "for larger datasets, we should implement an index.<br>\n",
    "index is for sorting embeddings before you search through them.<br>\n",
    "so, it narrows down the search space.<br>\n",
    "indexing libraries like `Faiss`<br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### getting an llm for local generation "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### checking local gpu memory availability"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 71,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Availability GPU memory: 4 GB\n"
     ]
    }
   ],
   "source": [
    "import torch \n",
    "gpu_memory_bytes = torch.cuda.get_device_properties(0).total_memory\n",
    "gpu_memory_gb = round(gpu_memory_bytes / (2**30))\n",
    "print(f\"Availability GPU memory: {gpu_memory_gb} GB\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "a gemma model will be used and following if else statements will help to determine which variant is most suitable for the local system"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Your available GPU memory is 4 GB, you may not have enough memory to run a Gemma LLM locally without quantization.\n"
     ]
    }
   ],
   "source": [
    "if gpu_memory_gb < 5.1:\n",
    "    print(f\"Your available GPU memory is {gpu_memory_gb} GB, you may not have enough memory to run a Gemma LLM locally without quantization.\")   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 73,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: bitsandbytes in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (0.43.3)\n",
      "Requirement already satisfied: accelerate in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (0.33.0)\n",
      "Requirement already satisfied: torch in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from bitsandbytes) (2.4.0+cu118)\n",
      "Requirement already satisfied: numpy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from bitsandbytes) (1.26.4)\n",
      "Requirement already satisfied: packaging>=20.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from accelerate) (24.1)\n",
      "Requirement already satisfied: psutil in c:\\users\\incha\\appdata\\roaming\\python\\python310\\site-packages (from accelerate) (6.0.0)\n",
      "Requirement already satisfied: pyyaml in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from accelerate) (6.0.1)\n",
      "Requirement already satisfied: huggingface-hub>=0.21.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from accelerate) (0.24.5)\n",
      "Requirement already satisfied: safetensors>=0.3.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from accelerate) (0.4.4)\n",
      "Requirement already satisfied: filelock in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.21.0->accelerate) (3.15.4)\n",
      "Requirement already satisfied: fsspec>=2023.5.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.21.0->accelerate) (2024.5.0)\n",
      "Requirement already satisfied: requests in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.21.0->accelerate) (2.32.3)\n",
      "Requirement already satisfied: tqdm>=4.42.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.21.0->accelerate) (4.66.4)\n",
      "Requirement already satisfied: typing-extensions>=3.7.4.3 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from huggingface-hub>=0.21.0->accelerate) (4.12.2)\n",
      "Requirement already satisfied: sympy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->bitsandbytes) (1.13.1)\n",
      "Requirement already satisfied: networkx in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->bitsandbytes) (3.3)\n",
      "Requirement already satisfied: jinja2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->bitsandbytes) (3.1.4)\n",
      "Requirement already satisfied: colorama in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from tqdm>=4.42.1->huggingface-hub>=0.21.0->accelerate) (0.4.3)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from jinja2->torch->bitsandbytes) (2.1.5)\n",
      "Requirement already satisfied: charset-normalizer<4,>=2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.21.0->accelerate) (3.3.2)\n",
      "Requirement already satisfied: idna<4,>=2.5 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.21.0->accelerate) (3.7)\n",
      "Requirement already satisfied: urllib3<3,>=1.21.1 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.21.0->accelerate) (1.26.20)\n",
      "Requirement already satisfied: certifi>=2017.4.17 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from requests->huggingface-hub>=0.21.0->accelerate) (2024.6.2)\n",
      "Requirement already satisfied: mpmath<1.4,>=1.1.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sympy->torch->bitsandbytes) (1.3.0)\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "%pip install bitsandbytes accelerate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 74,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Collecting flash-attn\n",
      "  Using cached flash_attn-2.6.3.tar.gz (2.6 MB)\n",
      "  Preparing metadata (setup.py): started\n",
      "  Preparing metadata (setup.py): finished with status 'done'\n",
      "Requirement already satisfied: torch in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from flash-attn) (2.4.0+cu118)\n",
      "Collecting einops (from flash-attn)\n",
      "  Using cached einops-0.8.0-py3-none-any.whl.metadata (12 kB)\n",
      "Requirement already satisfied: filelock in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (3.15.4)\n",
      "Requirement already satisfied: typing-extensions>=4.8.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (4.12.2)\n",
      "Requirement already satisfied: sympy in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (1.13.1)\n",
      "Requirement already satisfied: networkx in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (3.3)\n",
      "Requirement already satisfied: jinja2 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (3.1.4)\n",
      "Requirement already satisfied: fsspec in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from torch->flash-attn) (2024.5.0)\n",
      "Requirement already satisfied: MarkupSafe>=2.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from jinja2->torch->flash-attn) (2.1.5)\n",
      "Requirement already satisfied: mpmath<1.4,>=1.1.0 in c:\\users\\incha\\appdata\\local\\programs\\python\\python310\\lib\\site-packages (from sympy->torch->flash-attn) (1.3.0)\n",
      "Using cached einops-0.8.0-py3-none-any.whl (43 kB)\n",
      "Building wheels for collected packages: flash-attn\n",
      "  Building wheel for flash-attn (setup.py): started\n",
      "  Building wheel for flash-attn (setup.py): finished with status 'error'\n",
      "  Running setup.py clean for flash-attn\n",
      "Failed to build flash-attn\n",
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  error: subprocess-exited-with-error\n",
      "  \n",
      "  × python setup.py bdist_wheel did not run successfully.\n",
      "  │ exit code: 1\n",
      "  ╰─> [110 lines of output]\n",
      "      fatal: not a git repository (or any of the parent directories): .git\n",
      "      \n",
      "      \n",
      "      torch.__version__  = 2.4.0+cu118\n",
      "      \n",
      "      \n",
      "      running bdist_wheel\n",
      "      Guessing wheel URL:  https://github.com/Dao-AILab/flash-attention/releases/download/v2.6.3/flash_attn-2.6.3+cu118torch2.4cxx11abiFALSE-cp310-cp310-win_amd64.whl\n",
      "      Precompiled wheel not found. Building from source...\n",
      "      c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\utils\\cpp_extension.py:495: UserWarning: Attempted to use ninja as the BuildExtension backend but we could not find ninja.. Falling back to using the slow distutils backend.\n",
      "        warnings.warn(msg.format('we could not find ninja.'))\n",
      "      running build\n",
      "      running build_py\n",
      "      creating build\n",
      "      creating build\\lib.win-amd64-3.10\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\bert_padding.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\flash_attn_interface.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\flash_attn_triton.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\flash_attn_triton_og.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\flash_blocksparse_attention.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\flash_blocksparse_attn_interface.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\fused_softmax.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      copying flash_attn\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\n",
      "      creating build\\lib.win-amd64-3.10\\hopper\n",
      "      copying hopper\\benchmark_attn.py -> build\\lib.win-amd64-3.10\\hopper\n",
      "      copying hopper\\flash_attn_interface.py -> build\\lib.win-amd64-3.10\\hopper\n",
      "      copying hopper\\setup.py -> build\\lib.win-amd64-3.10\\hopper\n",
      "      copying hopper\\test_flash_attn.py -> build\\lib.win-amd64-3.10\\hopper\n",
      "      copying hopper\\__init__.py -> build\\lib.win-amd64-3.10\\hopper\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\layers\n",
      "      copying flash_attn\\layers\\patch_embed.py -> build\\lib.win-amd64-3.10\\flash_attn\\layers\n",
      "      copying flash_attn\\layers\\rotary.py -> build\\lib.win-amd64-3.10\\flash_attn\\layers\n",
      "      copying flash_attn\\layers\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\layers\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\losses\n",
      "      copying flash_attn\\losses\\cross_entropy.py -> build\\lib.win-amd64-3.10\\flash_attn\\losses\n",
      "      copying flash_attn\\losses\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\losses\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\baichuan.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\bert.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\bigcode.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\btlm.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\falcon.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\gpt.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\gptj.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\gpt_neox.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\llama.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\opt.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\vit.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      copying flash_attn\\models\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\models\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      copying flash_attn\\modules\\block.py -> build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      copying flash_attn\\modules\\embedding.py -> build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      copying flash_attn\\modules\\mha.py -> build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      copying flash_attn\\modules\\mlp.py -> build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      copying flash_attn\\modules\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\modules\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      copying flash_attn\\ops\\activations.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      copying flash_attn\\ops\\fused_dense.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      copying flash_attn\\ops\\layer_norm.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      copying flash_attn\\ops\\rms_norm.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      copying flash_attn\\ops\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      copying flash_attn\\utils\\benchmark.py -> build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      copying flash_attn\\utils\\distributed.py -> build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      copying flash_attn\\utils\\generation.py -> build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      copying flash_attn\\utils\\pretrained.py -> build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      copying flash_attn\\utils\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\utils\n",
      "      creating build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\cross_entropy.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\k_activations.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\layer_norm.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\linear.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\mlp.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\rotary.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      copying flash_attn\\ops\\triton\\__init__.py -> build\\lib.win-amd64-3.10\\flash_attn\\ops\\triton\n",
      "      running build_ext\n",
      "      c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\utils\\cpp_extension.py:380: UserWarning: Error checking compiler version for cl: [WinError 2] The system cannot find the file specified\n",
      "        warnings.warn(f'Error checking compiler version for {compiler}: {error}')\n",
      "      building 'flash_attn_2_cuda' extension\n",
      "      creating build\\temp.win-amd64-3.10\n",
      "      creating build\\temp.win-amd64-3.10\\Release\n",
      "      creating build\\temp.win-amd64-3.10\\Release\\csrc\n",
      "      creating build\\temp.win-amd64-3.10\\Release\\csrc\\flash_attn\n",
      "      creating build\\temp.win-amd64-3.10\\Release\\csrc\\flash_attn\\src\n",
      "      C:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\MSVC\\14.40.33807\\bin\\HostX86\\x64\\cl.exe /c /nologo /Ox /W3 /GL /DNDEBUG /MD -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\flash_attn -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\flash_attn\\src -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\cutlass\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\torch\\csrc\\api\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\TH -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\THC -IC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\Include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\MSVC\\14.40.33807\\include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\MSVC\\14.40.33807\\ATLMFC\\include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Auxiliary\\VS\\include -IC:\\Program Files (x86)\\Windows Kits\\10\\include\\10.0.22621.0\\ucrt -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\um -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\shared -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\winrt -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\cppwinrt -IC:\\Program Files (x86)\\Windows Kits\\NETFXSDK\\4.8\\include\\um /EHsc /Tpcsrc/flash_attn/flash_api.cpp /Fobuild\\temp.win-amd64-3.10\\Release\\csrc/flash_attn/flash_api.obj /MD /wd4819 /wd4251 /wd4244 /wd4267 /wd4275 /wd4018 /wd4190 /wd4624 /wd4067 /wd4068 /EHsc -O3 -std=c++17 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=flash_attn_2_cuda -D_GLIBCXX_USE_CXX11_ABI=0 /std:c++17\n",
      "      cl : Command line warning D9002 : ignoring unknown option '-O3'\n",
      "      cl : Command line warning D9002 : ignoring unknown option '-std=c++17'\n",
      "      flash_api.cpp\n",
      "      C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\bin\\nvcc -c csrc/flash_attn/src/flash_bwd_hdim128_bf16_causal_sm80.cu -o build\\temp.win-amd64-3.10\\Release\\csrc/flash_attn/src/flash_bwd_hdim128_bf16_causal_sm80.obj -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\flash_attn -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\flash_attn\\src -IC:\\Users\\incha\\AppData\\Local\\Temp\\pip-install-ashd95og\\flash-attn_a4433c530ff041369f91d712a2a31d29\\csrc\\cutlass\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\torch\\csrc\\api\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\TH -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\torch\\include\\THC -IC:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\include -Ic:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\Include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\MSVC\\14.40.33807\\include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Tools\\MSVC\\14.40.33807\\ATLMFC\\include -IC:\\Program Files\\Microsoft Visual Studio\\2022\\Community\\VC\\Auxiliary\\VS\\include -IC:\\Program Files (x86)\\Windows Kits\\10\\include\\10.0.22621.0\\ucrt -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\um -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\shared -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\winrt -IC:\\Program Files (x86)\\Windows Kits\\10\\\\include\\10.0.22621.0\\\\cppwinrt -IC:\\Program Files (x86)\\Windows Kits\\NETFXSDK\\4.8\\include\\um -Xcudafe --diag_suppress=dll_interface_conflict_dllexport_assumed -Xcudafe --diag_suppress=dll_interface_conflict_none_assumed -Xcudafe --diag_suppress=field_without_dll_interface -Xcudafe --diag_suppress=base_class_has_different_dll_interface -Xcompiler /EHsc -Xcompiler /wd4068 -Xcompiler /wd4067 -Xcompiler /wd4624 -Xcompiler /wd4190 -Xcompiler /wd4018 -Xcompiler /wd4275 -Xcompiler /wd4267 -Xcompiler /wd4244 -Xcompiler /wd4251 -Xcompiler /wd4819 -Xcompiler /MD -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ --expt-relaxed-constexpr -O3 -std=c++17 -U__CUDA_NO_HALF_OPERATORS__ -U__CUDA_NO_HALF_CONVERSIONS__ -U__CUDA_NO_HALF2_OPERATORS__ -U__CUDA_NO_BFLOAT16_CONVERSIONS__ --expt-relaxed-constexpr --expt-extended-lambda --use_fast_math -gencode arch=compute_80,code=sm_80 -gencode arch=compute_90,code=sm_90 --threads 4 -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=flash_attn_2_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++17 --use-local-env\n",
      "      flash_bwd_hdim128_bf16_causal_sm80.cu\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_OPERATORS__' with '/U__CUDA_NO_HALF_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_CONVERSIONS__' with '/U__CUDA_NO_HALF_CONVERSIONS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF2_OPERATORS__' with '/U__CUDA_NO_HALF2_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_BFLOAT16_CONVERSIONS__' with '/U__CUDA_NO_BFLOAT16_CONVERSIONS__'\n",
      "      flash_bwd_hdim128_bf16_causal_sm80.cu\n",
      "      C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\include\\crt/host_config.h(153): fatal error C1189: #error:  -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_OPERATORS__' with '/U__CUDA_NO_HALF_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_CONVERSIONS__' with '/U__CUDA_NO_HALF_CONVERSIONS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF2_OPERATORS__' with '/U__CUDA_NO_HALF2_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_BFLOAT16_CONVERSIONS__' with '/U__CUDA_NO_BFLOAT16_CONVERSIONS__'\n",
      "      flash_bwd_hdim128_bf16_causal_sm80.cu\n",
      "      C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\include\\crt/host_config.h(153): fatal error C1189: #error:  -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_OPERATORS__' with '/U__CUDA_NO_HALF_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF_CONVERSIONS__' with '/U__CUDA_NO_HALF_CONVERSIONS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_HALF2_OPERATORS__' with '/U__CUDA_NO_HALF2_OPERATORS__'\n",
      "      cl : Command line warning D9025 : overriding '/D__CUDA_NO_BFLOAT16_CONVERSIONS__' with '/U__CUDA_NO_BFLOAT16_CONVERSIONS__'\n",
      "      flash_bwd_hdim128_bf16_causal_sm80.cu\n",
      "      C:\\Program Files\\NVIDIA GPU Computing Toolkit\\CUDA\\v11.8\\include\\crt/host_config.h(153): fatal error C1189: #error:  -- unsupported Microsoft Visual Studio version! Only the versions between 2017 and 2022 (inclusive) are supported! The nvcc flag '-allow-unsupported-compiler' can be used to override this version check; however, using an unsupported host compiler may cause compilation failure or incorrect run time execution. Use at your own risk.\n",
      "      error: command 'C:\\\\Program Files\\\\NVIDIA GPU Computing Toolkit\\\\CUDA\\\\v11.8\\\\bin\\\\nvcc.exe' failed with exit code 4294967295\n",
      "      [end of output]\n",
      "  \n",
      "  note: This error originates from a subprocess, and is likely not a problem with pip.\n",
      "  ERROR: Failed building wheel for flash-attn\n",
      "ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (flash-attn)\n",
      "\n",
      "[notice] A new release of pip is available: 24.1.2 -> 24.2\n",
      "[notice] To update, run: python.exe -m pip install --upgrade pip\n"
     ]
    }
   ],
   "source": [
    "import torch\n",
    "from transformers import AutoTokenizer, AutoModelForCausalLM\n",
    "from transformers.utils import is_flash_attn_2_available\n",
    "\n",
    "# 1. create quantization config\n",
    "# for models that require 4-bit quantization \n",
    "from transformers import BitsAndBytesConfig\n",
    "quantization_config = BitsAndBytesConfig(load_in_4bit=True, bnb_4bit_compute_dtype=torch.float16)\n",
    "\n",
    "# set up flash attention 2 for faster inference, default to \"sdpa\" or \"scaled dot producy attention\"\n",
    "# flash attention 2 requires nvidia gpu compute capability of 8.0 or above.\n",
    "%pip install flash-attn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 75,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[INFO] Using attention implementation: sdpa\n"
     ]
    }
   ],
   "source": [
    "if (is_flash_attn_2_available()) and (torch.cuda.get_device_capability(0)[0] >= 8):\n",
    "  attn_implementation = \"flash_attention_2\"\n",
    "else:\n",
    "  attn_implementation = \"sdpa\"\n",
    "print(f\"[INFO] Using attention implementation: {attn_implementation}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 77,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\incha\\AppData\\Local\\Programs\\Python\\Python310\\lib\\site-packages\\transformers\\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884\n",
      "  warnings.warn(\n"
     ]
    }
   ],
   "source": [
    "from transformers import AutoTokenizer, AutoModelForCausalLM\n",
    "\n",
    "model = AutoModelForCausalLM.from_pretrained(\"EleutherAI/gpt-neo-1.3B\",\n",
    "                                             torch_dtype=torch.float16,\n",
    "                                            quantization_config=quantization_config,\n",
    "                                            low_cpu_mem_usage=True)\n",
    "tokenizer = AutoTokenizer.from_pretrained(\"EleutherAI/gpt-neo-1.3B\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.10.0"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}