Recommendation System on Amazon SageMaker - Advanced (Factorization Machine).ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Recommendation System on Amazon SageMaker - Advanced (Factorization Machine)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this notebook, we are going utilize Apache Airflow with Amazaon SageMaker to build recommendation model "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The model would give recommendataion based on the Amazon customer ratings on over 160K digital videos. Details on this dataset can be found at its [AWS Public Datasets page](https://s3.amazonaws.com/amazon-reviews-pds/readme.html)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## High-Level ML workflow with Apache Airflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 1,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from IPython.display import Image\n",
    "\n",
    "file_name1 = './img/airflow.png'\n",
    "Image(file_name1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### SageMaker workflow with Airflow"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The workflow performs the following tasks:\n",
    "\n",
    "1. Data Pre-processing: Extract and pre-process data from S3 to prepare the training data. \\\n",
    "\n",
    "2. Prepare Training Data: To build the recommender system, we will use SageMaker's built-in algorithm - Factorization machines. The algorithm expects training data only in RecordIO Protobuf format with Float32 tensors. In this task, pre-processed data will be transformed to RecordIO Protobuf format. \\\n",
    "\n",
    "3. Training the Model: Train the SageMaker's built-in Factorization Machine model with the training data and generate model artifacts. The training job will be launched by the Airflow SageMaker operator SageMakerTrainingOperator. \\\n",
    "\n",
    "4. Tune the Model Hyper-parameters: A conditional/optional task to tune the hyper-parameters of Factorization Machine to find the best model. The hyper-parameter tuning job will be launched by the SageMaker Airflow operator SageMakerTuningOperator. \\\n",
    "\n",
    "5. Batch inference: Using the trained model, get inferences on the test dataset stored in Amazon S3 using Airflow SageMaker operator SageMakerTransformOperator."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Airflow DAG for ML Workflow"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Airflow DAG integrates all the ML tasks in a ML workflow. Airflow DAG is a python script where you express individual tasks as Airflow operators, set task dependencies and associate the tasks to the DAG to run either on demand or scheduled interval. The Airflow DAG script is divided into following sections\n",
    "\n",
    "Set DAG with parameters such as schedule_interval to run the workflow at scheduled time\n",
    "Set up training, tuning and inference configurations for each operators using Sagemaker Python SDK for Airflow operators.\n",
    "Create individual tasks as Airflow operators defining trigger rules and associating them with the DAG object. Refer previous section for defining the individual tasks\n",
    "Specify task dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "execution_count": 3,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "file_name2 = './img/airflow-sagemaker-dag.png'\n",
    "Image(file_name2)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "[Background](https://github.com/awslabs/amazon-sagemaker-examples/blob/master/introduction_to_applying_machine_learning/gluon_recommender_system/gluon_recommender_system.ipynb)\n",
    "(from Glouon recommender system github)\n",
    "\n",
    "In many ways, recommender systems were a catalyst for the current popularity of machine learning. One of Amazon's earliest successes was the \"Customers who bought this, also bought...\" feature, while the million dollar Netflix Prize spurred research, raised public awareness, and inspired numerous other data science competitions.\n",
    "\n",
    "Recommender systems can utilize a multitude of data sources and ML algorithms, and most combine various unsupervised, supervised, and reinforcement learning techniques into a holistic framework. However, the core component is almost always a model which predicts a user's rating (or purchase) for a certain item based on that user's historical ratings of similar items as well as the behavior of other similar users. The minimal required dataset for this is a history of user item ratings. In our case, we'll use 1 to 5 star ratings from over 2M Amazon customers on over 160K digital videos. More details on this dataset can be found at its AWS Public Datasets page.\n",
    "\n",
    "Matrix factorization has been the cornerstone of most user-item prediction models. This method starts with the large, sparse, user-item ratings in a single matrix, where users index the rows, and items index the columns. It then seeks to find two lower-dimensional, dense matrices which, when multiplied together, preserve the information and relationships in the larger matrix.\n",
    "\n",
    "Matrix factorization has been extended and generalized with deep learning and embeddings. These techniques allows us to introduce non-linearities for enhanced performance and flexibility. This notebook will fit a neural network-based model to generate recommendations for the Amazon video dataset. It will start by exploring our data in the notebook, training a model on the data and fit our model using a SageMaker managed training cluster. We'll then deploy to an endpoint and check our method.\n",
    "\n",
    "We will also see how the tasks in the machine learning pipeline can be orchestrated and automated through Apache Airflow integration with Sagemaker."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np                                # For matrix operations and numerical processing\n",
    "import pandas as pd                               # For munging tabular data\n",
    "import matplotlib.pyplot as plt                   # For charts and visualizations\n",
    "%matplotlib inline\n",
    "from IPython.display import Image                 # For displaying images in the notebook\n",
    "from IPython.display import display               # For displaying outputs in the notebook\n",
    "from time import gmtime, strftime                 # For labeling SageMaker models, endpoints, etc.\n",
    "import sys                                        # For writing outputs to notebook\n",
    "import math                                       # For ceiling function\n",
    "import json                                       # For parsing hosting outputs\n",
    "import os                                         # For manipulating filepath names\n",
    "import jsonlines\n",
    "import boto3 \n",
    "import csv\n",
    "import io\n",
    "import re\n",
    "import s3fs\n",
    "import copy\n",
    "import random\n",
    "import time\n",
    "from scipy.sparse import lil_matrix\n",
    "\n",
    "# plot aesthetics\n",
    "import seaborn as sns\n",
    "sns.set(color_codes=True)\n",
    "sns.set_context('paper')\n",
    "five_thirty_eight = [\"#30a2da\", \"#fc4f30\", \"#e5ae38\", \"#6d904f\", \"#8b8b8b\",]\n",
    "sns.set_palette(five_thirty_eight)\n",
    "\n",
    "%matplotlib inline\n",
    "\n",
    "from urllib.request import urlretrieve\n",
    "import zipfile\n",
    "from dateutil.parser import parse\n",
    "from random import shuffle\n",
    "import random\n",
    "import datetime\n",
    "\n",
    "import sagemaker                                 \n",
    "from sagemaker.predictor import csv_serializer \n",
    "from sagemaker.predictor import json_deserializer\n",
    "import sagemaker.amazon.common as smac\n",
    "from sagemaker import get_execution_role\n",
    "from sagemaker.tuner import HyperparameterTuner, ContinuousParameter\n",
    "from sagemaker.analytics import HyperparameterTuningJobAnalytics, TrainingJobAnalytics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sagemaker session : <sagemaker.session.Session object at 0x0000023C7C5E0CC8>\n",
      "S3 bucket : sagemaker-us-west-2-570447867175\n",
      "Prefix : sagemaker/airflow-recsys\n",
      "Region selected : us-west-2\n",
      "IAM role : arn:aws:iam::570447867175:role/SageMakerNotebookRole\n"
     ]
    }
   ],
   "source": [
    "sess   = sagemaker.Session()\n",
    "bucket = sess.default_bucket()                     \n",
    "prefix = 'sagemaker/airflow-recsys'\n",
    "region = boto3.Session().region_name\n",
    "role = 'arn:aws:iam::570447867175:role/SageMakerNotebookRole' # pass your IAM role name\n",
    "smclient = boto3.Session().client('sagemaker')\n",
    "\n",
    "print('Sagemaker session :', sess)\n",
    "print('S3 bucket :', bucket)\n",
    "print('Prefix :', prefix)\n",
    "print('Region selected :', region)\n",
    "print('IAM role :', role)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Load Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's start by loading data locally from an S3 public bucket. \n",
    "As mentioned at the top, this contains 1 to 5 star ratings from over 2M Amazon customers on over 160K digital videos. \n",
    "\n",
    "Note, because this dataset is over a half gigabyte, the load from S3 may take ~10 minutes. Also, since Amazon SageMaker Notebooks start with a 5GB persistent volume by default, and we don't need to keep this data on our instance for long, we'll bring it to the temporary volume (which has up to 20GB of storage)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "## Uncomment this line to download data to your local directory\n",
    "## !aws s3 cp s3://amazon-reviews-pds/tsv/amazon_reviews_us_Digital_Video_Download_v1_00.tsv.gz ./data/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "b'Skipping line 92523: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 343254: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 524626: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 623024: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 977412: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 1496867: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 1711638: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 1787213: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 2395306: expected 15 fields, saw 22\\n'\n",
      "b'Skipping line 2527690: expected 15 fields, saw 22\\n'\n"
     ]
    },
    {
     "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>marketplace</th>\n",
       "      <th>customer_id</th>\n",
       "      <th>review_id</th>\n",
       "      <th>product_id</th>\n",
       "      <th>product_parent</th>\n",
       "      <th>product_title</th>\n",
       "      <th>product_category</th>\n",
       "      <th>star_rating</th>\n",
       "      <th>helpful_votes</th>\n",
       "      <th>total_votes</th>\n",
       "      <th>vine</th>\n",
       "      <th>verified_purchase</th>\n",
       "      <th>review_headline</th>\n",
       "      <th>review_body</th>\n",
       "      <th>review_date</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>US</td>\n",
       "      <td>12190288</td>\n",
       "      <td>R3FU16928EP5TC</td>\n",
       "      <td>B00AYB1482</td>\n",
       "      <td>668895143</td>\n",
       "      <td>Enlightened: Season 1</td>\n",
       "      <td>Digital_Video_Download</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>Y</td>\n",
       "      <td>I loved it and I wish there was a season 3</td>\n",
       "      <td>I loved it and I wish there was a season 3... ...</td>\n",
       "      <td>2015-08-31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>US</td>\n",
       "      <td>30549954</td>\n",
       "      <td>R1IZHHS1MH3AQ4</td>\n",
       "      <td>B00KQD28OM</td>\n",
       "      <td>246219280</td>\n",
       "      <td>Vicious</td>\n",
       "      <td>Digital_Video_Download</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>Y</td>\n",
       "      <td>As always it seems that the best shows come fr...</td>\n",
       "      <td>As always it seems that the best shows come fr...</td>\n",
       "      <td>2015-08-31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>US</td>\n",
       "      <td>52895410</td>\n",
       "      <td>R52R85WC6TIAH</td>\n",
       "      <td>B01489L5LQ</td>\n",
       "      <td>534732318</td>\n",
       "      <td>After Words</td>\n",
       "      <td>Digital_Video_Download</td>\n",
       "      <td>4</td>\n",
       "      <td>17</td>\n",
       "      <td>18</td>\n",
       "      <td>N</td>\n",
       "      <td>Y</td>\n",
       "      <td>Charming movie</td>\n",
       "      <td>This movie isn't perfect, but it gets a lot of...</td>\n",
       "      <td>2015-08-31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>US</td>\n",
       "      <td>27072354</td>\n",
       "      <td>R7HOOYTVIB0DS</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>239012694</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>Digital_Video_Download</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>Y</td>\n",
       "      <td>Five Stars</td>\n",
       "      <td>excellant this is what tv should be</td>\n",
       "      <td>2015-08-31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>US</td>\n",
       "      <td>26939022</td>\n",
       "      <td>R1XQ2N5CDOZGNX</td>\n",
       "      <td>B0094LZMT0</td>\n",
       "      <td>535858974</td>\n",
       "      <td>On The Waterfront</td>\n",
       "      <td>Digital_Video_Download</td>\n",
       "      <td>5</td>\n",
       "      <td>0</td>\n",
       "      <td>0</td>\n",
       "      <td>N</td>\n",
       "      <td>Y</td>\n",
       "      <td>Brilliant film from beginning to end</td>\n",
       "      <td>Brilliant film from beginning to end. All of t...</td>\n",
       "      <td>2015-08-31</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "  marketplace  customer_id       review_id  product_id  product_parent  \\\n",
       "0          US     12190288  R3FU16928EP5TC  B00AYB1482       668895143   \n",
       "1          US     30549954  R1IZHHS1MH3AQ4  B00KQD28OM       246219280   \n",
       "2          US     52895410   R52R85WC6TIAH  B01489L5LQ       534732318   \n",
       "3          US     27072354   R7HOOYTVIB0DS  B008LOVIIK       239012694   \n",
       "4          US     26939022  R1XQ2N5CDOZGNX  B0094LZMT0       535858974   \n",
       "\n",
       "                           product_title        product_category  star_rating  \\\n",
       "0                  Enlightened: Season 1  Digital_Video_Download            5   \n",
       "1                                Vicious  Digital_Video_Download            5   \n",
       "2                            After Words  Digital_Video_Download            4   \n",
       "3  Masterpiece: Inspector Lewis Season 5  Digital_Video_Download            5   \n",
       "4                      On The Waterfront  Digital_Video_Download            5   \n",
       "\n",
       "   helpful_votes  total_votes vine verified_purchase  \\\n",
       "0              0            0    N                 Y   \n",
       "1              0            0    N                 Y   \n",
       "2             17           18    N                 Y   \n",
       "3              0            0    N                 Y   \n",
       "4              0            0    N                 Y   \n",
       "\n",
       "                                     review_headline  \\\n",
       "0         I loved it and I wish there was a season 3   \n",
       "1  As always it seems that the best shows come fr...   \n",
       "2                                     Charming movie   \n",
       "3                                         Five Stars   \n",
       "4               Brilliant film from beginning to end   \n",
       "\n",
       "                                         review_body review_date  \n",
       "0  I loved it and I wish there was a season 3... ...  2015-08-31  \n",
       "1  As always it seems that the best shows come fr...  2015-08-31  \n",
       "2  This movie isn't perfect, but it gets a lot of...  2015-08-31  \n",
       "3                excellant this is what tv should be  2015-08-31  \n",
       "4  Brilliant film from beginning to end. All of t...  2015-08-31  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Note, we'll set 'error_bad_lines=False' when reading the file\n",
    "# as there appear to be a very small number of records which would create a problem otherwise.\n",
    "path = 'C:\\\\Users\\\\bokhy\\\\Desktop\\\\Python\\\\github\\\\amazon-sagemaker\\\\data'  # Set Your PATH\n",
    "\n",
    "reviews = pd.read_csv(os.path.join(path, 'amazon_reviews_us_Digital_Video_Download_v1_00.tsv.gz'), delimiter='\\t', error_bad_lines=False)\n",
    "reviews.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this example, let's limit ourselves to customer_id, product_id, and star_rating. Including additional features in our recommendation system could be beneficial, but would require substantial processing (particularly the text data) which would take us beyond the scope of this notebook.\n",
    "\n",
    "Note: we'll keep product_title on the dataset to help verify our recommendations later in the notebook, but it will not be used in algorithm training."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "reviews = reviews[['customer_id', 'product_id', 'star_rating', 'product_title']]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Because most people haven't seen most videos, and people rate fewer videos than we actually watch, we'd expect our data to be sparse. Our algorithm should work well with this sparse problem in general, but we may still want to clean out some of the long tail. Let's look at some basic percentiles to confirm."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "customers = reviews['customer_id'].value_counts()\n",
    "products = reviews['product_id'].value_counts()\n",
    "\n",
    "quantiles = [0, 0.01, 0.02, 0.03, 0.04, 0.05, 0.1, 0.25, 0.5, 0.75, 0.9, 0.95, 0.96, 0.97, 0.98, 0.99, 1]\n",
    "product_q = pd.DataFrame(zip(quantiles, products.quantile(quantiles)), columns=[\"quantile\", \"products\"])\n",
    "customer_q = pd.DataFrame(zip(quantiles, customers.quantile(quantiles)), columns=[\"quantile\", \"customers\"])\n",
    "# product_q.tail(10)\n",
    "# customer_q.tail(10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "axp = sns.barplot(x=\"quantile\", y=\"products\", data=product_q, palette=five_thirty_eight)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "axc = sns.barplot(x=\"quantile\", y=\"customers\", data=customer_q, palette=five_thirty_eight)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "As we can see, only about 5% of customers have rated 5 or more videos, and only 25% of videos have been rated by 9+ customers."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Data Pre-processing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of records before removing the long tail =    3998345\n",
      "# of records after  removing the long tail =    1168955\n",
      "# of records after  removing duplicates    =    1168903\n"
     ]
    }
   ],
   "source": [
    "customers = customers[customers >= 5]\n",
    "products = products[products >= 10]\n",
    "\n",
    "print(\"# of records before removing the long tail = {:10d}\".format(reviews.shape[0]))\n",
    "reduced_df = reviews.merge(pd.DataFrame({'customer_id': customers.index})).merge(pd.DataFrame({'product_id': products.index}))\n",
    "print(\"# of records after  removing the long tail = {:10d}\".format(reduced_df.shape[0]))\n",
    "reduced_df = reduced_df.drop_duplicates(['customer_id', 'product_id'])\n",
    "print(\"# of records after  removing duplicates    = {:10d}\".format(reduced_df.shape[0]))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now, we'll recreate our customer and product lists since there are customers with more than 5 reviews, but all of their reviews are on products with less than 5 reviews (and vice versa)."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of customer : 43430756    467\n",
      "39122522    463\n",
      "20052283    373\n",
      "5291529     367\n",
      "44167709    348\n",
      "           ... \n",
      "25353442      1\n",
      "28902161      1\n",
      "58223         1\n",
      "34755484      1\n",
      "52662373      1\n",
      "Name: customer_id, Length: 140344, dtype: int64\n",
      "# of customer : B00I3MQNWG    5461\n",
      "B003AZCYCE    3847\n",
      "B007SPQZMC    3069\n",
      "B009ZQC7MY    2966\n",
      "B004MWZLYC    2371\n",
      "              ... \n",
      "B001IWWV74       1\n",
      "B00O192LP8       1\n",
      "B00BMX53AQ       1\n",
      "B00B8BURS0       1\n",
      "B010GYUAH2       1\n",
      "Name: product_id, Length: 38385, dtype: int64\n"
     ]
    }
   ],
   "source": [
    "customers = reduced_df['customer_id'].value_counts()\n",
    "products = reduced_df['product_id'].value_counts()\n",
    "print('# of customer :',customers)\n",
    "print('# of customer :',products)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Next, we'll number each user and item, giving them their own sequential index. This will allow us to hold the information in a sparse format where the sequential indices indicate the row and column in our ratings matrix."
   ]
  },
  {
   "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>customer_id</th>\n",
       "      <th>product_id</th>\n",
       "      <th>star_rating</th>\n",
       "      <th>product_title</th>\n",
       "      <th>customer</th>\n",
       "      <th>product</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>27072354</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>5</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>10974</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>16030865</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>5</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>481</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>44025160</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>5</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>31093</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>18602179</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>5</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>2096</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>14424972</td>\n",
       "      <td>B008LOVIIK</td>\n",
       "      <td>5</td>\n",
       "      <td>Masterpiece: Inspector Lewis Season 5</td>\n",
       "      <td>34200</td>\n",
       "      <td>106</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customer_id  product_id  star_rating  \\\n",
       "0     27072354  B008LOVIIK            5   \n",
       "1     16030865  B008LOVIIK            5   \n",
       "2     44025160  B008LOVIIK            5   \n",
       "3     18602179  B008LOVIIK            5   \n",
       "4     14424972  B008LOVIIK            5   \n",
       "\n",
       "                           product_title  customer  product  \n",
       "0  Masterpiece: Inspector Lewis Season 5     10974      106  \n",
       "1  Masterpiece: Inspector Lewis Season 5       481      106  \n",
       "2  Masterpiece: Inspector Lewis Season 5     31093      106  \n",
       "3  Masterpiece: Inspector Lewis Season 5      2096      106  \n",
       "4  Masterpiece: Inspector Lewis Season 5     34200      106  "
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "customer_index = pd.DataFrame({'customer_id': customers.index, 'customer': np.arange(customers.shape[0])})\n",
    "product_index = pd.DataFrame({'product_id': products.index, \n",
    "                              'product': np.arange(products.shape[0])})\n",
    "\n",
    "reduced_df = reduced_df.merge(customer_index).merge(product_index)\n",
    "reduced_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's look at the feature dimension size whch will required for preparing the training and test data sets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "140344 38385 178729\n"
     ]
    }
   ],
   "source": [
    "nb_customer = reduced_df['customer'].max() + 1\n",
    "nb_products = reduced_df['product'].max() + 1\n",
    "feature_dim = nb_customer + nb_products\n",
    "print(nb_customer, nb_products, feature_dim)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Trim down the data set to include only customer, product, star_rating which is all we need for the training algorithm to build the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "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>customer</th>\n",
       "      <th>product</th>\n",
       "      <th>star_rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>10974</td>\n",
       "      <td>106</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>481</td>\n",
       "      <td>106</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>31093</td>\n",
       "      <td>106</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>2096</td>\n",
       "      <td>106</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>34200</td>\n",
       "      <td>106</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   customer  product  star_rating\n",
       "0     10974      106            5\n",
       "1       481      106            5\n",
       "2     31093      106            5\n",
       "3      2096      106            5\n",
       "4     34200      106            5"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "product_df = reduced_df[['customer', 'product', 'star_rating']]\n",
    "product_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Train/Test Split"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will be using SageMaker's implementation of Factorization Machines (FM) for building a recommender system. The algorithm expects float32 tensors in protobuf whereas the data sets are pandas dataframe on disk. Most of the conversion effort is handled by the Amazon SageMaker Python SDK.\n",
    "\n",
    "The FM algorithm will utilize sparse input and since the data sets are dense matrix, it has to be converted a sparse matrix with one-hot encoded feature vectors with customers and products. Thus, each sample in the data set will be a wide boolean vector with 178729 feature space (140344 customer + 38385 products) with only two values set to 1 with respect to the customer and product."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df, validate_df, test_df = np.split(\n",
    "    product_df.sample(frac=1), \n",
    "    [int(.6*len(product_df)), int(.8*len(product_df))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of rows in the training data set   =     701341\n",
      "# of rows in the validation data set =     233781\n",
      "# of rows in the test data set       =     233781\n"
     ]
    }
   ],
   "source": [
    "print(\"# of rows in the training data set   = {:10d}\".format(train_df.shape[0]))\n",
    "print(\"# of rows in the validation data set = {:10d}\".format(validate_df.shape[0]))\n",
    "print(\"# of rows in the test data set       = {:10d}\".format(test_df.shape[0]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>customer</th>\n",
       "      <th>product</th>\n",
       "      <th>star_rating</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>280998</th>\n",
       "      <td>91013</td>\n",
       "      <td>177</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>313718</th>\n",
       "      <td>68640</td>\n",
       "      <td>743</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>186136</th>\n",
       "      <td>69063</td>\n",
       "      <td>127</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>277722</th>\n",
       "      <td>103771</td>\n",
       "      <td>273</td>\n",
       "      <td>5</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>568106</th>\n",
       "      <td>4494</td>\n",
       "      <td>21971</td>\n",
       "      <td>1</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        customer  product  star_rating\n",
       "280998     91013      177            5\n",
       "313718     68640      743            5\n",
       "186136     69063      127            5\n",
       "277722    103771      273            5\n",
       "568106      4494    21971            1"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# look at training data\n",
    "train_df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's get the feature dimensions by adding total number of (unique) customers and products"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "# of customers =     140344\n",
      "# of products  =      38385\n",
      "# of features  =     178729\n"
     ]
    }
   ],
   "source": [
    "# get feature dimension\n",
    "all_df = pd.concat([train_df, validate_df, test_df])\n",
    "nb_customer = np.unique(all_df['customer'].values).shape[0]\n",
    "nb_products = np.unique(all_df['product'].values).shape[0]\n",
    "feature_dim = nb_customer + nb_products\n",
    "print(\"# of customers = {:10d}\".format(nb_customer))\n",
    "print(\"# of products  = {:10d}\".format(nb_products))\n",
    "print(\"# of features  = {:10d}\".format(feature_dim))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Create One-Hot Encoded Sparse Matrix"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our training matrix is now even sparser: Of all 183,833,321,511 values (1028559 rows 178729 columns), only 2,057,118 are non-zero (1,028,5592). In other words, the matrix is 99.99% sparse. Storing this as a dense matrix would be a massive waste of both storage and computing power. To avoid this, use a scipy.lil_matrix sparse matrix for features and a numpy array for ratings.\n",
    "\n",
    "Let's define a function that takes the data set and returns a sparse feature matrix and numpy array with ratings."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def convert_sparse_matrix(df, nb_rows, nb_customer, nb_products):\n",
    "    # dataframe to array\n",
    "    df_val = df.values\n",
    "\n",
    "    # determine feature size\n",
    "    nb_cols = nb_customer + nb_products\n",
    "    print(\"# of rows = {}\".format(str(nb_rows)))\n",
    "    print(\"# of cols = {}\".format(str(nb_cols)))\n",
    "\n",
    "    # extract customers and ratings\n",
    "    df_X = df_val[:, 0:2]\n",
    "    # Features are one-hot encoded in a sparse matrix\n",
    "    X = lil_matrix((nb_rows, nb_cols)).astype('float32')\n",
    "    df_X[:, 1] = nb_customer + df_X[:, 1]\n",
    "    coords = df_X[:, 0:2]\n",
    "    X[np.arange(nb_rows), coords[:, 0]] = 1\n",
    "    X[np.arange(nb_rows), coords[:, 1]] = 1\n",
    "\n",
    "    # create label with ratings\n",
    "    Y = df_val[:, 2].astype('float32')\n",
    "\n",
    "    # validate size and shape\n",
    "    print(X.shape)\n",
    "    print(Y.shape)\n",
    "    assert X.shape == (nb_rows, nb_cols)\n",
    "    assert Y.shape == (nb_rows, )\n",
    "\n",
    "    return X, Y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Convert training data set to one-hot encoded sparse matrix\n",
      "# of rows = 701341\n",
      "# of cols = 178729\n",
      "(701341, 178729)\n",
      "(701341,)\n",
      "Convert validation data set to one-hot encoded sparse matrix\n",
      "# of rows = 233781\n",
      "# of cols = 178729\n",
      "(233781, 178729)\n",
      "(233781,)\n",
      "Convert test data set to one-hot encoded sparse matrix\n",
      "# of rows = 233781\n",
      "# of cols = 178729\n",
      "(233781, 178729)\n",
      "(233781,)\n"
     ]
    }
   ],
   "source": [
    "print(\"Convert training data set to one-hot encoded sparse matrix\")\n",
    "train_X, train_Y = convert_sparse_matrix(train_df, train_df.shape[0], nb_customer, nb_products)\n",
    "print(\"Convert validation data set to one-hot encoded sparse matrix\")\n",
    "validate_X, validate_Y = convert_sparse_matrix(validate_df, validate_df.shape[0], nb_customer, nb_products)\n",
    "print(\"Convert test data set to one-hot encoded sparse matrix\")\n",
    "test_X, test_Y = convert_sparse_matrix(test_df, test_df.shape[0], nb_customer, nb_products)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Convert to Protobuf format and Upload to S3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will use Sagemaker's utility function write_spmatrix_to_sparse_tensor to convert scipy sparse matrix to protobuf format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "def save_as_protobuf(X, Y, bucket, key):\n",
    "    \"\"\"Converts features and predictions matrices to recordio protobuf and\n",
    "       writes to S3\n",
    "\n",
    "    Args:\n",
    "        X:\n",
    "          2D numpy matrix with features\n",
    "        Y:\n",
    "          1D numpy matrix with predictions\n",
    "        bucket:\n",
    "          s3 bucket where recordio protobuf file will be staged\n",
    "        prefix:\n",
    "          s3 url prefix to stage prepared data to use for training the model\n",
    "        key:\n",
    "          protobuf file name to be staged\n",
    "\n",
    "    Returns:\n",
    "        s3 url with key to the protobuf data\n",
    "    \"\"\"\n",
    "    buf = io.BytesIO()\n",
    "    smac.write_spmatrix_to_sparse_tensor(buf, X, Y)\n",
    "    buf.seek(0)\n",
    "    obj = '{}'.format(key)\n",
    "    boto3.resource('s3').Bucket(bucket).Object(obj).upload_fileobj(buf)\n",
    "    return 's3://{}/{}'.format(bucket, obj)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training data set in protobuf format uploaded at s3://sagemaker-us-west-2-570447867175/prepare/train/train.protobuf\n",
      "Validation data set in protobuf format uploaded at s3://sagemaker-us-west-2-570447867175/prepare/validate/validate.protobuf\n"
     ]
    }
   ],
   "source": [
    "s3_train_path = save_as_protobuf(train_X, train_Y, bucket, 'prepare/train/train.protobuf')\n",
    "print(\"Training data set in protobuf format uploaded at {}\".format(s3_train_path))\n",
    "s3_val_path = save_as_protobuf(validate_X, validate_Y, bucket, 'prepare/validate/validate.protobuf')\n",
    "print(\"Validation data set in protobuf format uploaded at {}\".format(s3_val_path))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We will chunk the test data to avoid the payload size issues when performing batch predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "def chunk(x, batch_size):\n",
    "    \"\"\"split array into chunks of batch_size\n",
    "    \"\"\"\n",
    "    chunk_range = range(0, x.shape[0], batch_size)\n",
    "    chunks = [x[p: p + batch_size] for p in chunk_range]\n",
    "    return chunks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_0.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_1.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_2.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_3.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_4.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_5.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_6.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_7.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_8.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_9.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_10.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_11.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_12.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_13.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_14.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_15.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_16.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_17.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_18.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_19.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_20.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_21.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_22.protobuf\n",
      "s3://sagemaker-us-west-2-570447867175/prepare/test/test_23.protobuf\n"
     ]
    }
   ],
   "source": [
    "test_x_chunks = chunk(test_X, 10000)\n",
    "test_y_chunks = chunk(test_Y, 10000)\n",
    "N = len(test_x_chunks)\n",
    "for i in range(N):\n",
    "    test_data = save_as_protobuf(\n",
    "        test_x_chunks[i],\n",
    "        test_y_chunks[i],\n",
    "        bucket,\n",
    "        \"prepare/test/test_\" + str(i) + \".protobuf\")\n",
    "    print(test_data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7. Start Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "'get_image_uri' method will be deprecated in favor of 'ImageURIProvider' class in SageMaker Python SDK v2.\n"
     ]
    }
   ],
   "source": [
    "from sagemaker.amazon.amazon_estimator import get_image_uri\n",
    "container = get_image_uri(region, 'factorization-machines', 'latest')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Note that:\n",
    "\n",
    "- feature_dim is set to 178729, which is the number of customers + products in the training data set. \\\n",
    "\n",
    "- predictor_type is set to 'regressor' since we are trying to predict the rating \\\n",
    "\n",
    "- mini_batch_size is set to 200. This value can be tuned for relatively minor improvements in fit and speed, but selecting a reasonable value relative to the dataset is appropriate in most cases. \\\n",
    "\n",
    "- num_factors is set to 64. Factorization machines find a lower dimensional representation of the interactions for all features. Making this value smaller provides a more parsimonious model, closer to a linear model, but may sacrifice information about interactions. Making it larger provides a higher-dimensional representation of feature interactions, but adds computational complexity and can lead to overfitting. In a practical application, time should be invested to tune this parameter to the appropriate value."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Parameter image_name will be renamed to image_uri in SageMaker Python SDK v2.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Wall time: 0 ns\n"
     ]
    }
   ],
   "source": [
    "%time\n",
    "\n",
    "output_location = 's3://{}/train/'.format(bucket)\n",
    "s3_train_path = 's3://{}/prepare/train/train.protobuf'.format(bucket)\n",
    "s3_val_path = 's3://{}/prepare/validate/validate.protobuf'.format(bucket)\n",
    "\n",
    "fm = sagemaker.estimator.Estimator(container,\n",
    "                                   role, \n",
    "                                   train_instance_count=1, \n",
    "                                   train_instance_type='ml.m4.xlarge',\n",
    "                                   output_path=output_location,\n",
    "                                   sagemaker_session=sess)\n",
    "\n",
    "fm.set_hyperparameters(feature_dim=feature_dim,\n",
    "                      predictor_type='regressor',\n",
    "                      mini_batch_size=200,\n",
    "                      num_factors=512,\n",
    "                      bias_lr=0.02,\n",
    "                      epochs=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:sagemaker:'s3_input' class will be renamed to 'TrainingInput' in SageMaker Python SDK v2.\n",
      "WARNING:sagemaker:'s3_input' class will be renamed to 'TrainingInput' in SageMaker Python SDK v2.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-07-01 21:25:06 Starting - Starting the training job...\n",
      "2020-07-01 21:25:10 Starting - Launching requested ML instances......\n",
      "2020-07-01 21:26:20 Starting - Preparing the instances for training......\n",
      "2020-07-01 21:27:15 Downloading - Downloading input data...\n",
      "2020-07-01 21:27:56 Training - Downloading the training image..\u001b[34mDocker entrypoint called with argument(s): train\u001b[0m\n",
      "\u001b[34mRunning default environment configuration script\u001b[0m\n",
      "\u001b[34m/opt/amazon/lib/python2.7/site-packages/pandas/util/nosetester.py:13: DeprecationWarning: Importing from numpy.testing.nosetester is deprecated, import from numpy.testing instead.\n",
      "  from numpy.testing import nosetester\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Reading default configuration from /opt/amazon/lib/python2.7/site-packages/algorithm/resources/default-conf.json: {u'factors_lr': u'0.0001', u'linear_init_sigma': u'0.01', u'epochs': 1, u'_wd': u'1.0', u'_num_kv_servers': u'auto', u'use_bias': u'true', u'factors_init_sigma': u'0.001', u'_log_level': u'info', u'bias_init_method': u'normal', u'linear_init_method': u'normal', u'linear_lr': u'0.001', u'factors_init_method': u'normal', u'_tuning_objective_metric': u'', u'bias_wd': u'0.01', u'use_linear': u'true', u'bias_lr': u'0.1', u'mini_batch_size': u'1000', u'_use_full_symbolic': u'true', u'batch_metrics_publish_interval': u'500', u'bias_init_sigma': u'0.01', u'_num_gpus': u'auto', u'_data_format': u'record', u'factors_wd': u'0.00001', u'linear_wd': u'0.001', u'_kvstore': u'auto', u'_learning_rate': u'1.0', u'_optimizer': u'adam'}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Reading provided configuration from /opt/ml/input/config/hyperparameters.json: {u'predictor_type': u'regressor', u'epochs': u'10', u'feature_dim': u'178729', u'bias_lr': u'0.02', u'num_factors': u'512', u'mini_batch_size': u'200'}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Final configuration: {u'factors_lr': u'0.0001', u'linear_init_sigma': u'0.01', u'epochs': u'10', u'feature_dim': u'178729', u'num_factors': u'512', u'_wd': u'1.0', u'_num_kv_servers': u'auto', u'use_bias': u'true', u'factors_init_sigma': u'0.001', u'_log_level': u'info', u'bias_init_method': u'normal', u'linear_init_method': u'normal', u'linear_lr': u'0.001', u'factors_init_method': u'normal', u'_tuning_objective_metric': u'', u'bias_wd': u'0.01', u'use_linear': u'true', u'bias_lr': u'0.02', u'mini_batch_size': u'200', u'_use_full_symbolic': u'true', u'batch_metrics_publish_interval': u'500', u'predictor_type': u'regressor', u'bias_init_sigma': u'0.01', u'_num_gpus': u'auto', u'_data_format': u'record', u'factors_wd': u'0.00001', u'linear_wd': u'0.001', u'_kvstore': u'auto', u'_learning_rate': u'1.0', u'_optimizer': u'adam'}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 WARNING 139853468579648] Loggers have already been setup.\u001b[0m\n",
      "\u001b[34mProcess 1 is a worker.\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Using default worker.\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:28:19.272] [tensorio] [warning] TensorIO is already initialized; ignoring the initialization routine.\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:28:19.275] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 0, \"duration\": 5, \"num_examples\": 1, \"num_bytes\": 13344}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] nvidia-smi took: 0.0252158641815 secs to identify 0 gpus\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Number of GPUs being used: 0\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] [Sparse network] Building a sparse network.\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:19 INFO 139853468579648] Create Store: local\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"initialize.time\": {\"count\": 1, \"max\": 38.780927658081055, \"sum\": 38.780927658081055, \"min\": 38.780927658081055}}, \"EndTime\": 1593638899.313662, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593638899.269243}\n",
      "\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 0, \"sum\": 0.0, \"min\": 0}, \"Total Batches Seen\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}, \"Total Records Seen\": {\"count\": 1, \"max\": 200, \"sum\": 200.0, \"min\": 200}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 200, \"sum\": 200.0, \"min\": 200}, \"Reset Count\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}}, \"EndTime\": 1593638899.313908, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"init_train_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593638899.313855}\n",
      "\u001b[0m\n",
      "\u001b[34m[21:28:19] /opt/brazil-pkg-cache/packages/AIAlgorithmsMXNet/AIAlgorithmsMXNet-1.1.x.202835.0/AL2012/generic-flavor/src/src/kvstore/./kvstore_local.h:280: Warning: non-default weights detected during kvstore pull. This call has been ignored. Please make sure to use row_sparse_pull with row_ids.\u001b[0m\n",
      "\u001b[34m[21:28:19] /opt/brazil-pkg-cache/packages/AIAlgorithmsMXNet/AIAlgorithmsMXNet-1.1.x.202835.0/AL2012/generic-flavor/src/src/kvstore/./kvstore_local.h:280: Warning: non-default weights detected during kvstore pull. This call has been ignored. Please make sure to use row_sparse_pull with row_ids.\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=0 train rmse <loss>=4.29982994985\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=0 train mse <loss>=18.4885375977\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=0 train absolute_loss <loss>=4.13939147949\u001b[0m\n",
      "\n",
      "2020-07-01 21:28:16 Training - Training image download completed. Training in progress.\u001b[34m[07/01/2020 21:28:28 INFO 139853468579648] Iter[0] Batch [500]#011Speed: 17860.79 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=500 train rmse <loss>=2.03843828642\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=500 train mse <loss>=4.15523064756\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=500 train absolute_loss <loss>=1.66225406274\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:33 INFO 139853468579648] Iter[0] Batch [1000]#011Speed: 19949.36 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1000 train rmse <loss>=1.67034172802\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1000 train mse <loss>=2.79004148836\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1000 train absolute_loss <loss>=1.30557930267\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:38 INFO 139853468579648] Iter[0] Batch [1500]#011Speed: 20141.91 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1500 train rmse <loss>=1.52615376761\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1500 train mse <loss>=2.32914532239\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=1500 train absolute_loss <loss>=1.18373035799\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:44 INFO 139853468579648] Iter[0] Batch [2000]#011Speed: 19242.24 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:44 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2000 train rmse <loss>=1.44651889707\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:44 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2000 train mse <loss>=2.09241691959\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:44 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2000 train absolute_loss <loss>=1.12032067886\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:49 INFO 139853468579648] Iter[0] Batch [2500]#011Speed: 19612.09 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:49 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2500 train rmse <loss>=1.39508388042\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:49 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2500 train mse <loss>=1.9462590334\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:49 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=2500 train absolute_loss <loss>=1.08089371926\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:54 INFO 139853468579648] Iter[0] Batch [3000]#011Speed: 20154.97 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:54 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3000 train rmse <loss>=1.35946884949\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:54 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3000 train mse <loss>=1.84815555274\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:54 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3000 train absolute_loss <loss>=1.05460130627\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] Iter[0] Batch [3500]#011Speed: 20024.25 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3500 train rmse <loss>=1.33256139722\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3500 train mse <loss>=1.77571987737\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, batch=3500 train absolute_loss <loss>=1.03493631094\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:28:59.177] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 2, \"duration\": 37118, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, train rmse <loss>=1.33274029034\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, train mse <loss>=1.7761966815\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=0, train absolute_loss <loss>=1.03498966125\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"epochs\": {\"count\": 1, \"max\": 10, \"sum\": 10.0, \"min\": 10}, \"update.time\": {\"count\": 1, \"max\": 39864.004135131836, \"sum\": 39864.004135131836, \"min\": 39864.004135131836}}, \"EndTime\": 1593638939.178155, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593638899.313752}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #progress_metric: host=algo-1, completed 10 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 3508, \"sum\": 3508.0, \"min\": 3508}, \"Total Records Seen\": {\"count\": 1, \"max\": 701541, \"sum\": 701541.0, \"min\": 701541}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 2, \"sum\": 2.0, \"min\": 2}}, \"EndTime\": 1593638939.178495, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 0}, \"StartTime\": 1593638899.314115}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=17593.1097897 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=0 train rmse <loss>=1.1175042233\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=0 train mse <loss>=1.24881568909\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:28:59 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=0 train absolute_loss <loss>=0.877727127075\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:04 INFO 139853468579648] Iter[1] Batch [500]#011Speed: 20039.21 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:04 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=500 train rmse <loss>=1.1488301572\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:04 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=500 train mse <loss>=1.3198107301\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:04 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=500 train absolute_loss <loss>=0.910167439779\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:09 INFO 139853468579648] Iter[1] Batch [1000]#011Speed: 20503.07 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:09 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1000 train rmse <loss>=1.14321873135\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:09 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1000 train mse <loss>=1.3069490677\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:09 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1000 train absolute_loss <loss>=0.906293207754\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:14 INFO 139853468579648] Iter[1] Batch [1500]#011Speed: 19837.39 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:14 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1500 train rmse <loss>=1.14000665868\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:14 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1500 train mse <loss>=1.29961518183\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:14 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=1500 train absolute_loss <loss>=0.903135591694\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:18 INFO 139853468579648] Iter[1] Batch [2000]#011Speed: 20536.00 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2000 train rmse <loss>=1.13688631301\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2000 train mse <loss>=1.29251048872\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2000 train absolute_loss <loss>=0.900177649632\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:23 INFO 139853468579648] Iter[1] Batch [2500]#011Speed: 20363.70 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2500 train rmse <loss>=1.13402118143\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2500 train mse <loss>=1.28600403994\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=2500 train absolute_loss <loss>=0.897650443119\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:28 INFO 139853468579648] Iter[1] Batch [3000]#011Speed: 20209.37 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3000 train rmse <loss>=1.13213540636\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3000 train mse <loss>=1.28173057833\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3000 train absolute_loss <loss>=0.896272173497\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] Iter[1] Batch [3500]#011Speed: 20484.41 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3500 train rmse <loss>=1.13032031124\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3500 train mse <loss>=1.27762400601\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, batch=3500 train absolute_loss <loss>=0.8948325084\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:29:33.788] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 4, \"duration\": 34605, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, train rmse <loss>=1.1308390065\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, train mse <loss>=1.27879685862\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=1, train absolute_loss <loss>=0.89507395074\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34607.905864715576, \"sum\": 34607.905864715576, \"min\": 34607.905864715576}}, \"EndTime\": 1593638973.788857, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593638939.178265}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #progress_metric: host=algo-1, completed 20 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 7015, \"sum\": 7015.0, \"min\": 7015}, \"Total Records Seen\": {\"count\": 1, \"max\": 1402882, \"sum\": 1402882.0, \"min\": 1402882}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 3, \"sum\": 3.0, \"min\": 3}}, \"EndTime\": 1593638973.789139, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 1}, \"StartTime\": 1593638939.180913}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20265.0718587 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=0 train rmse <loss>=1.08386630128\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=0 train mse <loss>=1.17476615906\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=0 train absolute_loss <loss>=0.854978637695\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:38 INFO 139853468579648] Iter[2] Batch [500]#011Speed: 20149.43 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=500 train rmse <loss>=1.11003944753\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=500 train mse <loss>=1.23218757508\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:38 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=500 train absolute_loss <loss>=0.879555651728\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:43 INFO 139853468579648] Iter[2] Batch [1000]#011Speed: 20046.52 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1000 train rmse <loss>=1.10441066578\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1000 train mse <loss>=1.21972291868\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1000 train absolute_loss <loss>=0.875000363406\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:48 INFO 139853468579648] Iter[2] Batch [1500]#011Speed: 19949.27 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1500 train rmse <loss>=1.10134920622\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1500 train mse <loss>=1.21297007404\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=1500 train absolute_loss <loss>=0.871673963014\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:53 INFO 139853468579648] Iter[2] Batch [2000]#011Speed: 20877.03 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2000 train rmse <loss>=1.09858019506\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2000 train mse <loss>=1.20687844497\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2000 train absolute_loss <loss>=0.868713406435\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:58 INFO 139853468579648] Iter[2] Batch [2500]#011Speed: 20086.54 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:58 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2500 train rmse <loss>=1.09616464311\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:58 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2500 train mse <loss>=1.2015769248\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:29:58 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=2500 train absolute_loss <loss>=0.866299890688\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:03 INFO 139853468579648] Iter[2] Batch [3000]#011Speed: 19743.82 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:03 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3000 train rmse <loss>=1.09459602852\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:03 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3000 train mse <loss>=1.19814046565\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:03 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3000 train absolute_loss <loss>=0.864968445221\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] Iter[2] Batch [3500]#011Speed: 20295.57 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3500 train rmse <loss>=1.09324053098\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3500 train mse <loss>=1.19517485857\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, batch=3500 train absolute_loss <loss>=0.863737686754\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:30:08.608] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 6, \"duration\": 34814, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, train rmse <loss>=1.09377983364\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, train mse <loss>=1.19635432448\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=2, train absolute_loss <loss>=0.863983856802\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34817.24405288696, \"sum\": 34817.24405288696, \"min\": 34817.24405288696}}, \"EndTime\": 1593639008.608852, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593638973.788952}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #progress_metric: host=algo-1, completed 30 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 10522, \"sum\": 10522.0, \"min\": 10522}, \"Total Records Seen\": {\"count\": 1, \"max\": 2104223, \"sum\": 2104223.0, \"min\": 2104223}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 4, \"sum\": 4.0, \"min\": 4}}, \"EndTime\": 1593639008.609073, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 2}, \"StartTime\": 1593638973.791576}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20143.2769039 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=0 train rmse <loss>=1.05500284638\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=0 train mse <loss>=1.11303100586\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:08 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=0 train absolute_loss <loss>=0.832950744629\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:13 INFO 139853468579648] Iter[3] Batch [500]#011Speed: 20529.91 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:13 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=500 train rmse <loss>=1.07423990528\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:13 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=500 train mse <loss>=1.15399137409\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:13 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=500 train absolute_loss <loss>=0.848336917169\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:18 INFO 139853468579648] Iter[3] Batch [1000]#011Speed: 20714.64 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1000 train rmse <loss>=1.0679795239\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1000 train mse <loss>=1.14058026347\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:18 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1000 train absolute_loss <loss>=0.842576617718\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:23 INFO 139853468579648] Iter[3] Batch [1500]#011Speed: 20603.65 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1500 train rmse <loss>=1.06459937813\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1500 train mse <loss>=1.13337183592\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=1500 train absolute_loss <loss>=0.838669258082\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:28 INFO 139853468579648] Iter[3] Batch [2000]#011Speed: 20467.74 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2000 train rmse <loss>=1.06174503816\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2000 train mse <loss>=1.12730252607\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:28 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2000 train absolute_loss <loss>=0.835412839692\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:33 INFO 139853468579648] Iter[3] Batch [2500]#011Speed: 20182.97 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2500 train rmse <loss>=1.05940158545\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2500 train mse <loss>=1.12233171926\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:33 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=2500 train absolute_loss <loss>=0.832861987651\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:37 INFO 139853468579648] Iter[3] Batch [3000]#011Speed: 20138.09 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3000 train rmse <loss>=1.05784938585\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3000 train mse <loss>=1.11904532313\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3000 train absolute_loss <loss>=0.83136128696\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] Iter[3] Batch [3500]#011Speed: 19938.23 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3500 train rmse <loss>=1.05665129504\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3500 train mse <loss>=1.1165119593\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, batch=3500 train absolute_loss <loss>=0.830136821325\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:30:43.065] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 8, \"duration\": 34451, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, train rmse <loss>=1.05721008306\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, train mse <loss>=1.11769315971\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=3, train absolute_loss <loss>=0.830385432813\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34454.05983924866, \"sum\": 34454.05983924866, \"min\": 34454.05983924866}}, \"EndTime\": 1593639043.065668, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639008.608924}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #progress_metric: host=algo-1, completed 40 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 14029, \"sum\": 14029.0, \"min\": 14029}, \"Total Records Seen\": {\"count\": 1, \"max\": 2805564, \"sum\": 2805564.0, \"min\": 2805564}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 5, \"sum\": 5.0, \"min\": 5}}, \"EndTime\": 1593639043.065982, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 3}, \"StartTime\": 1593639008.611573}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20355.5408337 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=0 train rmse <loss>=1.02558640244\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=0 train mse <loss>=1.05182746887\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:43 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=0 train absolute_loss <loss>=0.807324142456\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:48 INFO 139853468579648] Iter[4] Batch [500]#011Speed: 20025.12 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=500 train rmse <loss>=1.03688880232\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=500 train mse <loss>=1.07513838838\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:48 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=500 train absolute_loss <loss>=0.813193287954\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:53 INFO 139853468579648] Iter[4] Batch [1000]#011Speed: 20033.23 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1000 train rmse <loss>=1.03002493771\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1000 train mse <loss>=1.06095137231\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:53 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1000 train absolute_loss <loss>=0.806271210022\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:57 INFO 139853468579648] Iter[4] Batch [1500]#011Speed: 20469.58 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1500 train rmse <loss>=1.02633639013\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1500 train mse <loss>=1.05336638571\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:30:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=1500 train absolute_loss <loss>=0.801842948715\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:02 INFO 139853468579648] Iter[4] Batch [2000]#011Speed: 20333.84 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2000 train rmse <loss>=1.02338826687\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2000 train mse <loss>=1.04732354476\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2000 train absolute_loss <loss>=0.798372898016\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:07 INFO 139853468579648] Iter[4] Batch [2500]#011Speed: 20262.97 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2500 train rmse <loss>=1.02110156797\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2500 train mse <loss>=1.0426484121\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=2500 train absolute_loss <loss>=0.795758115241\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:12 INFO 139853468579648] Iter[4] Batch [3000]#011Speed: 20170.50 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3000 train rmse <loss>=1.01956248544\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3000 train mse <loss>=1.03950766172\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3000 train absolute_loss <loss>=0.794179240388\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] Iter[4] Batch [3500]#011Speed: 19805.29 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3500 train rmse <loss>=1.018490373\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3500 train mse <loss>=1.03732263989\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, batch=3500 train absolute_loss <loss>=0.793013074843\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:31:17.873] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 10, \"duration\": 34803, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, train rmse <loss>=1.01907111053\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, train mse <loss>=1.03850592832\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=4, train absolute_loss <loss>=0.793265354949\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34805.606842041016, \"sum\": 34805.606842041016, \"min\": 34805.606842041016}}, \"EndTime\": 1593639077.874112, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639043.065761}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #progress_metric: host=algo-1, completed 50 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 17536, \"sum\": 17536.0, \"min\": 17536}, \"Total Records Seen\": {\"count\": 1, \"max\": 3506905, \"sum\": 3506905.0, \"min\": 3506905}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 6, \"sum\": 6.0, \"min\": 6}}, \"EndTime\": 1593639077.874342, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 4}, \"StartTime\": 1593639043.068472}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20150.0028764 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=0 train rmse <loss>=0.994879084769\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=0 train mse <loss>=0.989784393311\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=0 train absolute_loss <loss>=0.778330230713\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:23 INFO 139853468579648] Iter[5] Batch [500]#011Speed: 19586.21 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=500 train rmse <loss>=0.997978063023\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=500 train mse <loss>=0.995960214276\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:23 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=500 train absolute_loss <loss>=0.775100279939\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:27 INFO 139853468579648] Iter[5] Batch [1000]#011Speed: 20348.36 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1000 train rmse <loss>=0.99073218389\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1000 train mse <loss>=0.981550260195\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1000 train absolute_loss <loss>=0.767417660083\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:32 INFO 139853468579648] Iter[5] Batch [1500]#011Speed: 20710.54 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1500 train rmse <loss>=0.986860672128\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1500 train mse <loss>=0.973893986193\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=1500 train absolute_loss <loss>=0.76268183165\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:37 INFO 139853468579648] Iter[5] Batch [2000]#011Speed: 20645.34 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2000 train rmse <loss>=0.98391357778\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2000 train mse <loss>=0.96808592854\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:37 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2000 train absolute_loss <loss>=0.759146777186\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:42 INFO 139853468579648] Iter[5] Batch [2500]#011Speed: 20398.48 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:42 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2500 train rmse <loss>=0.981745449232\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:42 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2500 train mse <loss>=0.963824127087\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:42 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=2500 train absolute_loss <loss>=0.756577895316\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:47 INFO 139853468579648] Iter[5] Batch [3000]#011Speed: 20198.60 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:47 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3000 train rmse <loss>=0.980272346399\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:47 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3000 train mse <loss>=0.960933873114\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:47 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3000 train absolute_loss <loss>=0.755012275088\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] Iter[5] Batch [3500]#011Speed: 20427.23 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3500 train rmse <loss>=0.979353513636\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3500 train mse <loss>=0.959133304671\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, batch=3500 train absolute_loss <loss>=0.753957637372\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:31:52.392] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 12, \"duration\": 34514, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, train rmse <loss>=0.979958392209\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, train mse <loss>=0.96031845046\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=5, train absolute_loss <loss>=0.754216007867\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34516.39699935913, \"sum\": 34516.39699935913, \"min\": 34516.39699935913}}, \"EndTime\": 1593639112.393318, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639077.874197}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #progress_metric: host=algo-1, completed 60 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 21043, \"sum\": 21043.0, \"min\": 21043}, \"Total Records Seen\": {\"count\": 1, \"max\": 4208246, \"sum\": 4208246.0, \"min\": 4208246}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 7, \"sum\": 7.0, \"min\": 7}}, \"EndTime\": 1593639112.393546, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 5}, \"StartTime\": 1593639077.876889}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20318.8426446 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=0 train rmse <loss>=0.963233331508\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=0 train mse <loss>=0.927818450928\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:52 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=0 train absolute_loss <loss>=0.748736801147\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:57 INFO 139853468579648] Iter[6] Batch [500]#011Speed: 20267.14 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=500 train rmse <loss>=0.958525950064\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=500 train mse <loss>=0.918771996945\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:31:57 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=500 train absolute_loss <loss>=0.735659285905\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:02 INFO 139853468579648] Iter[6] Batch [1000]#011Speed: 20276.04 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1000 train rmse <loss>=0.951120992935\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1000 train mse <loss>=0.904631143202\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:02 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1000 train absolute_loss <loss>=0.72749628652\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:07 INFO 139853468579648] Iter[6] Batch [1500]#011Speed: 20675.89 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1500 train rmse <loss>=0.947189959697\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1500 train mse <loss>=0.89716881975\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:07 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=1500 train absolute_loss <loss>=0.722638812437\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:12 INFO 139853468579648] Iter[6] Batch [2000]#011Speed: 19928.30 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2000 train rmse <loss>=0.944341653585\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2000 train mse <loss>=0.891781158695\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:12 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2000 train absolute_loss <loss>=0.719164477686\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:17 INFO 139853468579648] Iter[6] Batch [2500]#011Speed: 20022.83 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2500 train rmse <loss>=0.94235536194\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2500 train mse <loss>=0.888033628178\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:17 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=2500 train absolute_loss <loss>=0.716740390422\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:22 INFO 139853468579648] Iter[6] Batch [3000]#011Speed: 20130.36 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:22 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3000 train rmse <loss>=0.940999670359\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:22 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3000 train mse <loss>=0.885480379615\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:22 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3000 train absolute_loss <loss>=0.715284676328\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] Iter[6] Batch [3500]#011Speed: 20173.78 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3500 train rmse <loss>=0.940265273175\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3500 train mse <loss>=0.884098783938\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, batch=3500 train absolute_loss <loss>=0.714409229395\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:32:27.109] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 14, \"duration\": 34712, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, train rmse <loss>=0.940895465868\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, train mse <loss>=0.885284277691\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=6, train absolute_loss <loss>=0.714672194373\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34714.37692642212, \"sum\": 34714.37692642212, \"min\": 34714.37692642212}}, \"EndTime\": 1593639147.110551, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639112.393399}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #progress_metric: host=algo-1, completed 70 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 24550, \"sum\": 24550.0, \"min\": 24550}, \"Total Records Seen\": {\"count\": 1, \"max\": 4909587, \"sum\": 4909587.0, \"min\": 4909587}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 8, \"sum\": 8.0, \"min\": 8}}, \"EndTime\": 1593639147.110788, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 6}, \"StartTime\": 1593639112.396141}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20202.9608175 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=0 train rmse <loss>=0.931119645999\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=0 train mse <loss>=0.866983795166\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:27 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=0 train absolute_loss <loss>=0.718440322876\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:32 INFO 139853468579648] Iter[7] Batch [500]#011Speed: 20143.55 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=500 train rmse <loss>=0.919583817953\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=500 train mse <loss>=0.845634398241\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:32 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=500 train absolute_loss <loss>=0.696509056244\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:36 INFO 139853468579648] Iter[7] Batch [1000]#011Speed: 20440.19 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1000 train rmse <loss>=0.912202748867\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1000 train mse <loss>=0.83211385504\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1000 train absolute_loss <loss>=0.68824149571\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:41 INFO 139853468579648] Iter[7] Batch [1500]#011Speed: 20573.71 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1500 train rmse <loss>=0.908304161813\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1500 train mse <loss>=0.825016450367\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=1500 train absolute_loss <loss>=0.68346881287\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:46 INFO 139853468579648] Iter[7] Batch [2000]#011Speed: 20230.93 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2000 train rmse <loss>=0.90562472715\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2000 train mse <loss>=0.820156146426\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2000 train absolute_loss <loss>=0.68018610705\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:51 INFO 139853468579648] Iter[7] Batch [2500]#011Speed: 20322.15 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2500 train rmse <loss>=0.903861342035\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2500 train mse <loss>=0.816965325625\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=2500 train absolute_loss <loss>=0.678000840115\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:56 INFO 139853468579648] Iter[7] Batch [3000]#011Speed: 20582.32 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3000 train rmse <loss>=0.902654679262\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3000 train mse <loss>=0.814785469993\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:32:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3000 train absolute_loss <loss>=0.67671151335\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] Iter[7] Batch [3500]#011Speed: 20427.61 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3500 train rmse <loss>=0.902119015542\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3500 train mse <loss>=0.813818718203\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, batch=3500 train absolute_loss <loss>=0.676043475962\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:33:01.536] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 16, \"duration\": 34421, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, train rmse <loss>=0.902775043864\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, train mse <loss>=0.815002779823\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=7, train absolute_loss <loss>=0.676310574314\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34423.45309257507, \"sum\": 34423.45309257507, \"min\": 34423.45309257507}}, \"EndTime\": 1593639181.53685, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639147.110638}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #progress_metric: host=algo-1, completed 80 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 28057, \"sum\": 28057.0, \"min\": 28057}, \"Total Records Seen\": {\"count\": 1, \"max\": 5610928, \"sum\": 5610928.0, \"min\": 5610928}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 9, \"sum\": 9.0, \"min\": 9}}, \"EndTime\": 1593639181.53708, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 7}, \"StartTime\": 1593639147.113357}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20373.6914023 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=0 train rmse <loss>=0.899064234525\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=0 train mse <loss>=0.808316497803\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:01 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=0 train absolute_loss <loss>=0.68767074585\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:06 INFO 139853468579648] Iter[8] Batch [500]#011Speed: 20678.66 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:06 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=500 train rmse <loss>=0.881905607789\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:06 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=500 train mse <loss>=0.77775750105\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:06 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=500 train absolute_loss <loss>=0.658960270986\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:11 INFO 139853468579648] Iter[8] Batch [1000]#011Speed: 19893.51 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:11 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1000 train rmse <loss>=0.874663310875\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:11 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1000 train mse <loss>=0.765035907391\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:11 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1000 train absolute_loss <loss>=0.650788496782\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:16 INFO 139853468579648] Iter[8] Batch [1500]#011Speed: 19506.30 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:16 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1500 train rmse <loss>=0.870844316663\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:16 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1500 train mse <loss>=0.758369823865\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:16 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=1500 train absolute_loss <loss>=0.646167129857\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:21 INFO 139853468579648] Iter[8] Batch [2000]#011Speed: 20162.63 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:21 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2000 train rmse <loss>=0.868364910341\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:21 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2000 train mse <loss>=0.754057617512\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:21 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2000 train absolute_loss <loss>=0.643135473222\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:26 INFO 139853468579648] Iter[8] Batch [2500]#011Speed: 19885.89 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:26 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2500 train rmse <loss>=0.866835208591\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:26 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2500 train mse <loss>=0.751403278853\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:26 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=2500 train absolute_loss <loss>=0.641206432573\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:31 INFO 139853468579648] Iter[8] Batch [3000]#011Speed: 20337.26 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:31 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3000 train rmse <loss>=0.865786678457\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:31 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3000 train mse <loss>=0.749586572593\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:31 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3000 train absolute_loss <loss>=0.640080669165\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] Iter[8] Batch [3500]#011Speed: 20198.76 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3500 train rmse <loss>=0.865444873431\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3500 train mse <loss>=0.748994828948\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, batch=3500 train absolute_loss <loss>=0.639609794693\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:33:36.463] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 18, \"duration\": 34921, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, train rmse <loss>=0.866127012441\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, train mse <loss>=0.75017600168\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=8, train absolute_loss <loss>=0.639880603761\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34923.874855041504, \"sum\": 34923.874855041504, \"min\": 34923.874855041504}}, \"EndTime\": 1593639216.463646, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639181.536929}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #progress_metric: host=algo-1, completed 90 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 31564, \"sum\": 31564.0, \"min\": 31564}, \"Total Records Seen\": {\"count\": 1, \"max\": 6312269, \"sum\": 6312269.0, \"min\": 6312269}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 10, \"sum\": 10.0, \"min\": 10}}, \"EndTime\": 1593639216.463928, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 8}, \"StartTime\": 1593639181.539734}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20081.705217 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=0 train rmse <loss>=0.867464852184\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=0 train mse <loss>=0.752495269775\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:36 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=0 train absolute_loss <loss>=0.657159805298\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:41 INFO 139853468579648] Iter[9] Batch [500]#011Speed: 20357.54 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=500 train rmse <loss>=0.845869498124\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=500 train mse <loss>=0.715495207857\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:41 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=500 train absolute_loss <loss>=0.623406031336\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:46 INFO 139853468579648] Iter[9] Batch [1000]#011Speed: 20246.14 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1000 train rmse <loss>=0.838832092281\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1000 train mse <loss>=0.703639279041\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:46 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1000 train absolute_loss <loss>=0.615416760326\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:51 INFO 139853468579648] Iter[9] Batch [1500]#011Speed: 20333.80 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1500 train rmse <loss>=0.835112900555\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1500 train mse <loss>=0.697413556673\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:51 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=1500 train absolute_loss <loss>=0.610949413851\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:56 INFO 139853468579648] Iter[9] Batch [2000]#011Speed: 20766.59 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2000 train rmse <loss>=0.832840565479\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2000 train mse <loss>=0.693623407508\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:33:56 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2000 train absolute_loss <loss>=0.608151470815\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:00 INFO 139853468579648] Iter[9] Batch [2500]#011Speed: 20364.99 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:00 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2500 train rmse <loss>=0.831537305734\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:00 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2500 train mse <loss>=0.691454290827\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:00 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=2500 train absolute_loss <loss>=0.606466704502\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:05 INFO 139853468579648] Iter[9] Batch [3000]#011Speed: 20396.85 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:05 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3000 train rmse <loss>=0.830644547468\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:05 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3000 train mse <loss>=0.689970364239\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:05 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3000 train absolute_loss <loss>=0.605490823585\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] Iter[9] Batch [3500]#011Speed: 20481.60 samples/sec\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3500 train rmse <loss>=0.830481791241\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3500 train mse <loss>=0.689700005582\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, batch=3500 train absolute_loss <loss>=0.605198552432\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:34:10.828] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/train\", \"epoch\": 20, \"duration\": 34360, \"num_examples\": 3507, \"num_bytes\": 46813936}\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, train rmse <loss>=0.83119033498\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, train mse <loss>=0.690877372964\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, epoch=9, train absolute_loss <loss>=0.605472419191\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, train rmse <loss>=0.83119033498\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, train mse <loss>=0.690877372964\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #quality_metric: host=algo-1, train absolute_loss <loss>=0.605472419191\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 34363.21496963501, \"sum\": 34363.21496963501, \"min\": 34363.21496963501}}, \"EndTime\": 1593639250.829734, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639216.463742}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #progress_metric: host=algo-1, completed 100 % of epochs\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 3507, \"sum\": 3507.0, \"min\": 3507}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Total Batches Seen\": {\"count\": 1, \"max\": 35071, \"sum\": 35071.0, \"min\": 35071}, \"Total Records Seen\": {\"count\": 1, \"max\": 7013610, \"sum\": 7013610.0, \"min\": 7013610}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 701341, \"sum\": 701341.0, \"min\": 701341}, \"Reset Count\": {\"count\": 1, \"max\": 11, \"sum\": 11.0, \"min\": 11}}, \"EndTime\": 1593639250.830036, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"training_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\", \"epoch\": 9}, \"StartTime\": 1593639216.466483}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] #throughput_metric: host=algo-1, train throughput=20409.3605429 records/second\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 WARNING 139853468579648] wait_for_all_workers will not sync workers since the kv store is not running distributed\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:10 INFO 139853468579648] Pulling entire model from kvstore to finalize\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"finalize.time\": {\"count\": 1, \"max\": 25.075912475585938, \"sum\": 25.075912475585938, \"min\": 25.075912475585938}}, \"EndTime\": 1593639250.855515, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639250.829847}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:34:12 INFO 139853468579648] Saved checkpoint to \"/tmp/tmpOPjms6/state-0001.params\"\u001b[0m\n",
      "\u001b[34m[2020-07-01 21:34:13.080] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/test\", \"epoch\": 0, \"duration\": 353808, \"num_examples\": 1, \"num_bytes\": 13380}\u001b[0m\n",
      "\n",
      "2020-07-01 21:40:14 Uploading - Uploading generated training model\u001b[34m[2020-07-01 21:40:11.955] [tensorio] [info] epoch_stats={\"data_pipeline\": \"/opt/ml/input/data/test\", \"epoch\": 1, \"duration\": 358874, \"num_examples\": 1169, \"num_bytes\": 15603916}\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"Max Batches Seen Between Resets\": {\"count\": 1, \"max\": 1169, \"sum\": 1169.0, \"min\": 1169}, \"Number of Batches Since Last Reset\": {\"count\": 1, \"max\": 1169, \"sum\": 1169.0, \"min\": 1169}, \"Number of Records Since Last Reset\": {\"count\": 1, \"max\": 233781, \"sum\": 233781.0, \"min\": 233781}, \"Total Batches Seen\": {\"count\": 1, \"max\": 1169, \"sum\": 1169.0, \"min\": 1169}, \"Total Records Seen\": {\"count\": 1, \"max\": 233781, \"sum\": 233781.0, \"min\": 233781}, \"Max Records Seen Between Resets\": {\"count\": 1, \"max\": 233781, \"sum\": 233781.0, \"min\": 233781}, \"Reset Count\": {\"count\": 1, \"max\": 1, \"sum\": 1.0, \"min\": 1}}, \"EndTime\": 1593639611.955279, \"Dimensions\": {\"Host\": \"algo-1\", \"Meta\": \"test_data_iter\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639253.080411}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #test_score (algo-1) : ('rmse', 1.0622683112630258)\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #test_score (algo-1) : ('mse', 1.1284139651136005)\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #test_score (algo-1) : ('absolute_loss', 0.8553062813383824)\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #quality_metric: host=algo-1, test rmse <loss>=1.06226831126\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #quality_metric: host=algo-1, test mse <loss>=1.12841396511\u001b[0m\n",
      "\u001b[34m[07/01/2020 21:40:11 INFO 139853468579648] #quality_metric: host=algo-1, test absolute_loss <loss>=0.855306281338\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"totaltime\": {\"count\": 1, \"max\": 712751.8517971039, \"sum\": 712751.8517971039, \"min\": 712751.8517971039}, \"setuptime\": {\"count\": 1, \"max\": 56.96582794189453, \"sum\": 56.96582794189453, \"min\": 56.96582794189453}}, \"EndTime\": 1593639611.957232, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"factorization-machines\"}, \"StartTime\": 1593639250.855649}\n",
      "\u001b[0m\n",
      "\n",
      "2020-07-01 21:41:11 Completed - Training job completed\n",
      "Training seconds: 836\n",
      "Billable seconds: 836\n"
     ]
    }
   ],
   "source": [
    "fm.fit({'train': s3_train_path,'test': s3_val_path}, wait=True) # 'wait = False' to NOT forece cell to finish to go another"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 8. Check the loss"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For Loss function, Let's check the 'RMSE'(Root Mean Squared Error) of the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "training_job_name = fm._current_job_name\n",
    "metric_name = 'train:rmse:epoch'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'factorization-machines-2020-07-01-21-25-04-664'"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "training_job_name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The training job is completed\n"
     ]
    }
   ],
   "source": [
    "# run this cell to check current status of training job\n",
    "fm_training_job_result = smclient.describe_training_job(TrainingJobName=training_job_name)\n",
    "\n",
    "status = fm_training_job_result['TrainingJobStatus']\n",
    "if status != 'Completed':\n",
    "    print('Reminder: the training job has not been completed.')\n",
    "else:\n",
    "    print('The training job is completed')"
   ]
  },
  {
   "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>timestamp</th>\n",
       "      <th>metric_name</th>\n",
       "      <th>value</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>1.332740</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>60.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>1.130839</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>120.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>1.075495</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>180.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>0.999515</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>240.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>0.940895</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>300.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>0.884451</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>360.0</td>\n",
       "      <td>train:rmse:epoch</td>\n",
       "      <td>0.831190</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   timestamp       metric_name     value\n",
       "0        0.0  train:rmse:epoch  1.332740\n",
       "1       60.0  train:rmse:epoch  1.130839\n",
       "2      120.0  train:rmse:epoch  1.075495\n",
       "3      180.0  train:rmse:epoch  0.999515\n",
       "4      240.0  train:rmse:epoch  0.940895\n",
       "5      300.0  train:rmse:epoch  0.884451\n",
       "6      360.0  train:rmse:epoch  0.831190"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# plug-in the training job name and metrics to be captured\n",
    "metrics_dataframe = TrainingJobAnalytics(training_job_name=training_job_name,metric_names=[metric_name]).dataframe()\n",
    "metrics_dataframe"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x360 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt = metrics_dataframe.plot(kind='line', figsize=(12,5), x='timestamp', y='value', style='b.', legend=False)\n",
    "plt.set_ylabel(metric_name);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 9. Deploy the model (Inference)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Since the model is trained, all it takes to deploy the model is a Sagemaker API call deploy() that creates the model package, sets up endpoint configuration and finally creates the endpoint."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_predictor = fm.deploy(instance_type='ml.t2.medium', \n",
    "                         initial_instance_count=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### (1) Real-time Inference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Predictions could be done by sending HTTP POST requests from a separate web service, but to keep things easy, we'll just use the .predict() method from the SageMaker Python SDK. The API expects JSON or RecordIO format for request and JSON for response data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 45,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Helper functions\n",
    "def convert_to_protobuf(X, Y=None):\n",
    "    buf = io.BytesIO()\n",
    "    smac.write_spmatrix_to_sparse_tensor(buf, X, Y)\n",
    "    buf.seek(0)\n",
    "    return buf\n",
    "\n",
    "def convert_sparse_matrix_X(df, nb_rows, nb_customer, nb_products):\n",
    "    # dataframe to array\n",
    "    df_val = df.values\n",
    "\n",
    "    # determine feature size\n",
    "    nb_cols = nb_customer + nb_products\n",
    "    \n",
    "    # extract customers and ratings\n",
    "    df_X = df_val[:,0:2]\n",
    "    # Features are one-hot encoded in a sparse matrix\n",
    "    X = lil_matrix((nb_rows, nb_cols)).astype('float32')\n",
    "    df_X[:,1] = nb_customer + df_X[:,1]\n",
    "    coords = df_X[:,0:2]\n",
    "    X[np.arange(nb_rows), coords[:, 0]] = 1\n",
    "    X[np.arange(nb_rows), coords[:, 1]] = 1\n",
    "\n",
    "    # validate size and shape\n",
    "    assert X.shape == (nb_rows, nb_cols)\n",
    "    \n",
    "    return X"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_predictor.content_type = 'application/x-recordio-protobuf'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_pb = convert_to_protobuf(test_X[1000:1010]).getvalue()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "response = fm_predictor.predict(test_pb)\n",
    "response"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predicted = [round(r['score'], 2) for r in json.loads(response)['predictions']]\n",
    "predicted"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "results_df = pd.DataFrame(zip(test_Y[1000:1010], predicted), columns = [\"actual_rating\", \"predicted_rating\"])\n",
    "results_df"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### (2) Batch Inference"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Doing Batch inference on the test data set prepared earlier (chunking into multiple protobuf files). To run batch transform, create a model package for the transform endpoint"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Create the model from the training estimator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_model = fm.create_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "2. Perform batch inference on the test data set and save results to S3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_transformer = fm_model.transformer(\n",
    "    instance_type='ml.c4.xlarge', \n",
    "    instance_count=1, \n",
    "    strategy=\"MultiRecord\", \n",
    "    output_path=\"s3://{}/transform/\".format(bucket)\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_transformer.transform(\n",
    "    data=\"s3://{}/prepare/test/\".format(bucket), \n",
    "    data_type='S3Prefix', \n",
    "    content_type=\"application/x-recordio-protobuf\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Batch-Transform job name: ' + fm_transformer.latest_transform_job.job_name)\n",
    "fm_transformer.wait()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "3. Inference results will be stored in a separate file for each test file chunk. Let's download the results from S3 and merge them"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def download_from_s3(bucket, key):\n",
    "    s3 = boto3.resource('s3')\n",
    "    obj = s3.Object( bucket, key)\n",
    "    content = obj.get()['Body'].read()\n",
    "    return content"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_preds = []\n",
    "for i in range(N):\n",
    "    key = 'transform/test_' + str(i) + '.protobuf.out'\n",
    "    response = download_from_s3(bucket, key)\n",
    "    result = [json.loads(row)[\"score\"] for row in response.split(\"\\n\") if len(row) > 0]\n",
    "    test_preds.extend(result)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_preds = np.array(test_preds)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_preds.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_preds[0:10]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 10. Model Evaluation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's start by calculating a naive baseline to approximate how well our model is doing. The simplest estimate would be to assume every user item rating is just the average rating over all ratings."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Naive MSE:', np.mean((test_df['star_rating'] - np.mean(train_df['star_rating'])) ** 2))\n",
    "print('MSE:', np.mean((test_Y - test_preds) ** 2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can see that our neural network and embedding model produces substantially better results (~1.44 vs 1.13 on the mean square error).\n",
    "\n",
    "For recommender systems, **subjective accuracy** also matters. Let's get some recommendations for a random user to see if they make intuitive sense."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "Let's look at customer number '6' and the rating the person gave."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_customer_6 = reduced_df[reduced_df['customer'] == 6].sort_values(['star_rating', 'product'], ascending=[False, True])\n",
    "pd.concat((df_customer_6.head(10), df_customer_6.tail(10)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Looks like the customer #6 liked dramamtic television series and sci-fi, but they dislike silly comedies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def create_payload(cust_id, nb_customer, nb_products, product_index):\n",
    "    # prepare payload for user #6\n",
    "    c = [cust_id] * nb_products\n",
    "    p = product_index['product'].values\n",
    "    x = pd.DataFrame(zip(c,p))\n",
    "    p_x = convert_sparse_matrix_X(x, x.shape[0], nb_customer, nb_products)\n",
    "    x_pb = convert_to_protobuf(p_x)\n",
    "    return x_pb"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_pb = create_payload(6, nb_customer, nb_products, product_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# make predictions using end-point created in Real-Time Inference\n",
    "response = fm_predictor.predict(x_pb)\n",
    "predictions = [round(r['score'], 2) for r in json.loads(response)['predictions']]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions_df = pd.DataFrame({'product': product_index['product'],\n",
    "                            'prediction': predictions})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_results_cust_6 = df_customer_6.merge(predictions_df, on=['product'])[['customer', 'customer_id', 'product', 'product_id', 'product_title', 'star_rating', 'prediction']]\n",
    "df_results_cust_6.sort_values(['prediction', 'product'], ascending=[False, True])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Indeed, our predicted highly rated shows have some well-reviewed TV dramas and some sci-fi. Meanwhile, our bottom rated shows include goofball comedies."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "What about cusomter number '7'?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "x_pb = create_payload(7, nb_customer, nb_products, product_index)\n",
    "response = fm_predictor.predict(x_pb)\n",
    "# get predictions\n",
    "predictions_user7 = [round(r['score'], 2) for r in json.loads(response)['predictions']]\n",
    "\n",
    "plt.scatter(predictions_df['prediction'], np.array(predictions_user7))\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 11. Model Hypyerparmeter Tuning"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Let's see if we can tune some hyperparameters of Factorization Machine to improve model accuracy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "output_location = 's3://{}/train/'.format(bucket)\n",
    "s3_train_path = 's3://{}/prepare/train/train.protobuf'.format(bucket)\n",
    "s3_val_path = 's3://{}/prepare/validate/validate.protobuf'.format(bucket)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Same as above"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_estimator = sagemaker.estimator.Estimator(container,\n",
    "                                   role, \n",
    "                                   train_instance_count=1, \n",
    "                                   train_instance_type='ml.c5.4xlarge',\n",
    "                                   output_path=output_location,\n",
    "                                   sagemaker_session=sess)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_estimator.set_hyperparameters(\n",
    "    feature_dim=feature_dim,\n",
    "    predictor_type='regressor',\n",
    "    mini_batch_size=200,\n",
    "    num_factors=512,\n",
    "    bias_lr=0.02,\n",
    "    epochs=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# We do grid serach for two paratmers here\n",
    "# factors_lr: The learning rate \n",
    "# factors_init_sigma: The standard deviation for initialization of factorization terms. Takes effect if factors_init_method is set to normal.\n",
    "hyperparameter_ranges=  {\n",
    "    \"factors_lr\": ContinuousParameter(0.0001, 0.2),\n",
    "    \"factors_init_sigma\": ContinuousParameter(0.0001, 1)\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "objective_metric_name = \"test:rmse\"\n",
    "objective_type = \"Minimize\"\n",
    "\n",
    "fm_tuner = HyperparameterTuner(\n",
    "    estimator=fm_estimator,\n",
    "    objective_metric_name=objective_metric_name, \n",
    "    hyperparameter_ranges=hyperparameter_ranges,\n",
    "    objective_type=objective_type,\n",
    "    max_jobs=10,\n",
    "    max_parallel_jobs=2\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "timestamp_prefix = time.strftime(\"%Y%m%d-%H%M%S\", time.gmtime())\n",
    "fm_tuner_job_name = 'hpo-fm-' + timestamp_prefix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_tuner.fit({'train': s3_train_path, 'test': s3_val_path}, job_name=fm_tuner_job_name, wait=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### We can analyze SageMaker's Hyper-Parameter Tuning Job Results"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# plug-in the training job name and metrics to be captured\n",
    "fm_tuner_analytics = HyperparameterTuningJobAnalytics(hyperparameter_tuning_job_name=fm_tuner_job_name)\n",
    "df_fm_tuner_metrics = fm_tuner_analytics.dataframe()\n",
    "df_fm_tuner_metrics"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# visualize it!\n",
    "plt = df_fm_tuner_metrics.plot(kind='line', figsize=(12,5), x='TrainingStartTime', \n",
    "                             y='FinalObjectiveValue', \n",
    "                             style='b.', legend=False)\n",
    "plt.set_ylabel(objective_metric_name);"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Find the best model after Hyper-Parameter Optimization"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"fm_tuner_job_name: \" + fm_tuner_job_name)\n",
    "fm_tuner = HyperparameterTuner.attach(fm_tuner_job_name)\n",
    "\n",
    "fm_tuner_analytics = HyperparameterTuningJobAnalytics(hyperparameter_tuning_job_name=fm_tuner_job_name)\n",
    "df_fm_tuner_metrics = fm_tuner_analytics.dataframe()\n",
    "\n",
    "fm_best_model_name = fm_tuner.best_training_job()\n",
    "print(\"fm_best_model_name: \" + fm_best_model_name)\n",
    "\n",
    "fm_model_info = smclient.describe_training_job(TrainingJobName=fm_best_model_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_fm_tuner_metrics[df_fm_tuner_metrics['TrainingJobName']==fm_best_model_name]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###### Deploy the endpoint using hyper-parameter tuning job and test the predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm = sagemaker.estimator.Estimator.attach(fm_best_model_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Finally, re-run the cells in Batch Inference and Evaluation section to evaluate the performance of the model with tuned hyper-parameters.\n",
    "\n",
    "Assuming batch inference is carried out, let's calculate predictions for our test dataset and see if we do better than the training job with default hyper-parameters."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print('Tuned MSE:', np.mean((test_Y - test_preds) ** 2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 12. Points of further improvements"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "1. Overfitting prevention (early stopping, dropout, etc.)\n",
    "2. Testing whether binarizing our target variable would improve results\n",
    "3. Including other information sources (video genres, historical ratings, time of review)\n",
    "4. Adjusting our threshold for user and item inclusion\n",
    "5. Setting the context and key value store up for distributed training\n",
    "6. Fine tuning our data ingestion (e.g. num_workers on our data iterators) to ensure we're fully utilizing our GPU"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 13. Close the SageMaker Instance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "endpoint_name_contains = ['-fm-', 'factorization-machines-','airflow-recsys','recsys']\n",
    "for name in endpoint_name_contains:\n",
    "    endpoints = smclient.list_endpoints(NameContains=name, StatusEquals='InService')\n",
    "    endpoint_names = [r['EndpointName'] for r in endpoints['Endpoints']]\n",
    "    for endpoint_name in endpoint_names:\n",
    "        print(\"Deleting endpoint: \" + endpoint_name)\n",
    "        smclient.delete_endpoint(EndpointName=endpoint_name)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "fm_predictor.delete_endpoint(predictor.endpoint)\n",
    "fm_tuner.delete_endpoint(predictor.endpoint)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Refernece\n",
    "> https://aws.amazon.com/blogs/machine-learning/build-end-to-end-machine-learning-workflows-with-amazon-sagemaker-and-apache-airflow/\n",
    "> https://github.com/aws-samples/sagemaker-ml-workflow-with-apache-airflow\n",
    "> https://github.com/awslabs/amazon-sagemaker-examples/blob/master/introduction_to_applying_machine_learning/gluon_recommender_system/gluon_recommender_system.ipynb"
   ]
  }
 ],
 "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}