VQA-Med-tf.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from data_utils import Utility"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Valid data size= 500\n",
      "Train data size= 5413\n"
     ]
    }
   ],
   "source": [
    "val_images, val_questions, val_answers = Utility.read_dataset(\"Valid\")\n",
    "train_images, train_questions, train_answers = Utility.read_dataset(\"Train\")\n",
    "#test_images, test_questions = read_dataset(\"Test\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Image name : dataset/VQAMed2018Train/VQAMed2018Train-images/AIAN-14-313-g002.jpg\n",
      "Question   : where does axial section mri abdomen show hypoechoic mass ?\n",
      "Answer     : in distal pancreas\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAADWCAYAAADfCUmxAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n0z7t0z494/T/Az9Ol6qrPZ0/AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f47c7b009b0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "id =1\n",
    "Utility.show_image(id, train_images, train_questions, train_answers )"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Tensorflow"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "import yaml\n",
    "import sys\n",
    "import math\n",
    "import time\n",
    "import argparse\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "from random import shuffle\n",
    "from shutil import copyfile\n",
    "#from bleu import moses_multi_bleu\n",
    "from time import strftime, localtime\n",
    "from gensim.models.wrappers import FastText\n",
    "from tensorflow.python.layers import core as layers_core\n",
    "\n",
    "tf.set_random_seed(123)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(tf.__version__)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf.reset_default_graph()\n",
    "sess = tf.InteractiveSession() # config=tf.ConfigProto(log_device_placement=True))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_hidden_units = 300\n",
    "decoder_hidden_units = encoder_hidden_units\n",
    "output_dropout = 1.0\n",
    "emb_size = 150\n",
    "vocab_size = 100000\n",
    "num_layers = 1\n",
    "input_dropout = 0.9\n",
    "epochs = 100\n",
    "batch_size = 128\n",
    "bidirectional = 1\n",
    "residual = 0\n",
    "attention = 0\n",
    "attention_nodes = emb_size\n",
    "beam_width = 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "vectors = 'dataset/vqa_vec150'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load FastText Model and create lookup embeddings\n",
    "print(\"Load embeddings from: \"+vectors)\n",
    "all_model = FastText()\n",
    "all_model = all_model.load_fasttext_format(vectors)\n",
    "all_word_vectors = all_model.wv\n",
    "embeddings = np.array([all_model[k] for k in all_word_vectors.vocab.keys()])\n",
    "\n",
    "# Add PAD, EOS and UNK to the word embeddings\n",
    "pre = np.concatenate((np.ones(emb_size)*-1, np.ones(emb_size), np.zeros(emb_size)))\n",
    "all_vocab_size = len(all_word_vectors.vocab)\n",
    "embeddings = np.reshape(np.insert(embeddings, 0, pre), (all_vocab_size+3, emb_size))\n",
    "wor2tok = {'PAD': 0, 'EOS': 1, 'UNK': 2}\n",
    "tok2wor = {0: 'PAD', 1: 'EOS', 2: 'UNK'}\n",
    "ind = 3\n",
    "for k in all_word_vectors.vocab.keys():\n",
    "    wor2tok[k] = ind\n",
    "    tok2wor[ind] = k\n",
    "    ind += 1\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tf_embeddings = tf.get_variable(name=\"emb\", shape=embeddings.shape, dtype=tf.float32,\n",
    "                                initializer=tf.constant_initializer(embeddings),\n",
    "                                trainable=False)\n",
    "\n",
    "encoder_inputs = tf.placeholder(shape=(None, None), dtype=tf.int32,\n",
    "                                name='encoder_inputs')\n",
    "decoder_targets = tf.placeholder(shape=(None, None), dtype=tf.int32,\n",
    "                                name='decoder_targets')\n",
    "decoder_inputs = tf.placeholder(shape=(None, None), dtype=tf.int32,\n",
    "                                name='decoder_inputs')\n",
    "\n",
    "decoder_train_length = tf.placeholder(shape=(None,), dtype=tf.int32, name='decoder_targets_length')\n",
    "batch_size_ph = tf.placeholder(tf.int32, [], name=\"bath_size\")\n",
    "input_dropout_ph = tf.placeholder(tf.float32, name=\"dropout\")\n",
    "\n",
    "encoder_vec_inputs = tf.nn.embedding_lookup(tf_embeddings, encoder_inputs)\n",
    "decoder_vec_targets = tf.nn.embedding_lookup(tf_embeddings, decoder_targets)\n",
    "decoder_vec_inputs = tf.nn.embedding_lookup(tf_embeddings, decoder_inputs)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Encoder"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Bidirectional\n",
    "if bidirectional == 1:\n",
    "    forward_cell = tf.contrib.rnn.LSTMCell(encoder_hidden_units/2)\n",
    "    backward_cell = tf.contrib.rnn.LSTMCell(encoder_hidden_units/2)\n",
    "    forward_cell = tf.contrib.rnn.DropoutWrapper(\n",
    "        forward_cell, input_keep_prob=input_dropout_ph, output_keep_prob=output_dropout)\n",
    "    backward_cell = tf.contrib.rnn.DropoutWrapper(\n",
    "        backward_cell, input_keep_prob=input_dropout_ph, output_keep_prob=output_dropout)\n",
    "\n",
    "    biencoder_outputs, biencoder_final_state = tf.nn.bidirectional_dynamic_rnn(\n",
    "          forward_cell, backward_cell, encoder_vec_inputs,time_major=False,\n",
    "          dtype=tf.float32)\n",
    "\n",
    "    outputs_concat = tf.concat([biencoder_outputs[0], biencoder_outputs[1]], 2)\n",
    "i = 0\n",
    "encoder_cells = []\n",
    "for _ in range(num_layers):\n",
    "    cell = tf.contrib.rnn.LSTMCell(encoder_hidden_units)\n",
    "    if bidirectional == 0 and i > 0:\n",
    "        cell = tf.contrib.rnn.ResidualWrapper(cell)\n",
    "    elif bidirectional == 1:\n",
    "        cell = tf.contrib.rnn.ResidualWrapper(cell)\n",
    "    cell = tf.contrib.rnn.DropoutWrapper(cell, input_keep_prob=input_dropout_ph,\n",
    "                                         output_keep_prob=output_dropout)\n",
    "    encoder_cells.append(cell)\n",
    "    i += 1\n",
    "\n",
    "encoder_cell = tf.contrib.rnn.MultiRNNCell(encoder_cells)\n",
    "\n",
    "# Encoder output and final state\n",
    "encoder_outputs, encoder_final_state = tf.nn.dynamic_rnn(\n",
    "    encoder_cell, (encoder_vec_inputs if bidirectional == 0 else outputs_concat),\n",
    "    dtype=tf.float32, time_major=False, scope=\"plain_encoder\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def build_decoder( attention, encoder_outputs, encoder_final_state, helper, beam_size, reuse):\n",
    "    with tf.variable_scope(\"decoder\", reuse=reuse):\n",
    "        if beam_size > 1:\n",
    "            encoder_outputs = tf.contrib.seq2seq.tile_batch(\n",
    "                           encoder_outputs,\n",
    "                           multiplier=beam_size)\n",
    "\n",
    "            encoder_final_state = tf.contrib.framework.nest.map_structure(\n",
    "                        lambda s: tf.contrib.seq2seq.tile_batch(s, beam_size),\n",
    "                        encoder_final_state)\n",
    "\n",
    "        decoder_cells = []\n",
    "        for _ in range(num_layers):\n",
    "            cell = tf.contrib.rnn.LSTMCell(decoder_hidden_units)\n",
    "            cell = tf.contrib.rnn.DropoutWrapper(cell, input_keep_prob=input_dropout_ph,\n",
    "                                                output_keep_prob=output_dropout)\n",
    "            decoder_cells.append(cell)\n",
    "        initial_state = [state for state in encoder_final_state]\n",
    "        # Attention\n",
    "        if attention == 1:\n",
    "\n",
    "            attention_mechanism = tf.contrib.seq2seq.BahdanauAttention(\n",
    "                                    decoder_hidden_units,\n",
    "                                    encoder_outputs,\n",
    "                                    name=\"BahdanauAttention\")\n",
    "\n",
    "            decoder_cells[-1]= tf.contrib.seq2seq.AttentionWrapper(\n",
    "                                    decoder_cells[-1],\n",
    "                                    attention_mechanism,\n",
    "                                    attention_layer_size=emb_size,\n",
    "                                    name=\"AttentionWrapper\")\n",
    "\n",
    "            initial_state[-1] = decoder_cells[-1].zero_state(\n",
    "                                    batch_size=batch_size_ph * beam_size,\n",
    "                                    dtype=tf.float32)\n",
    "\n",
    "        initial_state = tuple(initial_state)\n",
    "        projection_layer = layers_core.Dense(vocab_size, use_bias=False)\n",
    "        decoder_cell = tf.contrib.rnn.MultiRNNCell(decoder_cells)\n",
    "\n",
    "        if helper !=None:\n",
    "            if helper==\"greedy\":\n",
    "                helper = tf.contrib.seq2seq.GreedyEmbeddingHelper(\n",
    "                        tf_embeddings,\n",
    "                        tf.fill([batch_size_ph], 1),\n",
    "                        1)\n",
    "            else:\n",
    "                helper = tf.contrib.seq2seq.TrainingHelper(\n",
    "                                        decoder_vec_inputs,\n",
    "                                        decoder_train_length,\n",
    "                                        name = \"TrainingHelper\")\n",
    "\n",
    "            print(\"Helper =\",helper)\n",
    "            decoder_outputs = tf.contrib.seq2seq.BasicDecoder(\n",
    "                                cell = decoder_cell,\n",
    "                                helper = helper,\n",
    "                                initial_state = initial_state,\n",
    "                                output_layer=projection_layer)\n",
    "\n",
    "        else:\n",
    "            decoder_outputs= tf.contrib.seq2seq.BeamSearchDecoder(\n",
    "                                    cell=decoder_cell,\n",
    "                                    embedding=tf_embeddings,\n",
    "                                    start_tokens= tf.fill([batch_size_ph], 1),\n",
    "                                    end_token=1,\n",
    "                                    beam_width=beam_size,\n",
    "                                    initial_state=initial_state,\n",
    "                                    output_layer=projection_layer,\n",
    "                                    length_penalty_weight=0.0)\n",
    "\n",
    "        decoder_outputs,_,_ = tf.contrib.seq2seq.dynamic_decode(\n",
    "                                    decoder_outputs,\n",
    "                                    maximum_iterations=30)\n",
    "    return decoder_outputs\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_decoder = build_decoder(\n",
    "                    attention,\n",
    "                    encoder_outputs,\n",
    "                    encoder_final_state,\n",
    "                    helper = \"train\",\n",
    "                    beam_size = 1,\n",
    "                    reuse = False)\n",
    "\n",
    "greedy_decoder = build_decoder(\n",
    "                    attention,\n",
    "                    encoder_outputs,\n",
    "                    encoder_final_state,\n",
    "                    helper = \"greedy\",\n",
    "                    beam_size = 1,\n",
    "                    reuse = True)\n",
    "\n",
    "beam_decoder = build_decoder(\n",
    "                    attention,\n",
    "                    encoder_outputs,\n",
    "                    encoder_final_state,\n",
    "                    helper = None,\n",
    "                    beam_size = beam_width,\n",
    "                    reuse = True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_translations= train_decoder.sample_id\n",
    "greedy_translations = greedy_decoder.sample_id\n",
    "beam_translations = beam_decoder.predicted_ids\n",
    "beam_scores = beam_decoder.beam_search_decoder_output.scores\n",
    "output = train_decoder.rnn_output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "loss = tf.reduce_mean(tf.nn.sparse_softmax_cross_entropy_with_logits(labels=decoder_targets, logits=output))\n",
    "train_op = tf.train.AdamOptimizer().minimize(loss)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Inference"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def test(sess, sent, id, actual):\n",
    "    \n",
    "    X = [sent.strip()]\n",
    "    X_tok = [[word for word in x.lower().strip().split()] for x in X]\n",
    "\n",
    "    ei = np.array([[wor2tok[w] if w in wor2tok.keys() else wor2tok['UNK'] for w in x]\n",
    "                + [wor2tok['EOS']] for x in X_tok]).T\n",
    "\n",
    "    greedy_out, beam_out, beam_score = sess.run(\n",
    "                [greedy_translations, beam_translations, beam_scores],\n",
    "                feed_dict={encoder_inputs: ei.T, input_dropout_ph: 1.0, batch_size_ph :1})\n",
    "\n",
    "    greedy_sent= \" \".join([tok2wor[i] for i in greedy_out[0][:-1]])\n",
    "\n",
    "    beam_sents = [ [tok2wor[j] for j in beam_out[0].T[i] if j!=-1 and j!=1] for i in range(beam_width)]\n",
    "    beam_sents_scores = [beam_score[0].T[i] for i in range(beam_width)]\n",
    "    list_of_preds = [\" \".join(r) for r in beam_sents]\n",
    "    if id < 30:\n",
    "        print(\"Sentence: \"+ sent)\n",
    "        print(\"Actual: \"+actual)\n",
    "        print(\"Greedy predicted: \"+ greedy_sent)\n",
    "       # printf(\"Beam predicted: \"+ list_of_preds)\n",
    "        for i in range(len(list_of_preds)):\n",
    "            print(\"Beam predicted #\"+str(i+1)+\": \"+ list_of_preds[i])\n",
    "    else:\n",
    "        sys.stdout.write('\\r')\n",
    "        sys.stdout.write(\"testing: \" + str(id) + \" out of \" + str(len(train_questions)))\n",
    "        sys.stdout.flush()\n",
    "\n",
    "    return greedy_sent, list_of_preds[0], list_of_preds[-1]\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Feeding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def next_feed(batch_size, i, isTrain):\n",
    "\n",
    "    if isTrain:\n",
    "        X = train_questions[batch_size*i:batch_size*(i+1)]\n",
    "        Y = train_answers[batch_size*i:batch_size*(i+1)]\n",
    "\n",
    "    src_max = 0 \n",
    "    trg_max = 0\n",
    "    sentence_lengths=[]\n",
    "\n",
    "    for x, y in zip(X, Y):\n",
    "        if len(y.split()) > trg_max:\n",
    "            trg_max = len(y.split())\n",
    "        if len(x.split()) > src_max:\n",
    "            src_max = len(x.split())\n",
    "\n",
    "    sentence_lengths= [trg_max+1 for s in range(batch_size)]\n",
    "    sentence_lengths= np.array(sentence_lengths)\n",
    "\n",
    "    X_tok = [[word for word in x.lower().strip().split()] for x in X]\n",
    "    Y_tok = [[word for word in y.lower().strip().split()] for y in Y]\n",
    "\n",
    "    ei = np.array([[wor2tok[w] for w in x] + [wor2tok['EOS']] +\n",
    "                   [wor2tok['PAD']] * (src_max - len(x)) for x in X_tok]).T\n",
    "    di = np.array([[wor2tok['EOS']] +\n",
    "                    [wor2tok[w] if wor2tok[w] < vocab_size else wor2tok['UNK'] for w in x]\n",
    "                    + [wor2tok['PAD']] * (trg_max - len(x)) for x in Y_tok]).T\n",
    "    dt = np.array([[wor2tok[w] if wor2tok[w] < vocab_size else wor2tok['UNK']\n",
    "                    for w in x] + [wor2tok['EOS']] +\n",
    "                    [wor2tok['PAD']]*(trg_max - len(x)) for x in Y_tok]).T\n",
    "\n",
    "\n",
    "    return {\n",
    "        encoder_inputs: ei.T,\n",
    "        decoder_inputs: di.T,\n",
    "        decoder_targets: dt.T,\n",
    "        input_dropout_ph: input_dropout,\n",
    "        batch_size_ph : batch_size,\n",
    "        decoder_train_length: sentence_lengths\n",
    "    }, X, Y\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "sess.run(tf.global_variables_initializer())\n",
    "saver = tf.train.Saver(tf.global_variables(), max_to_keep=100)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "print(\"Start training\")\n",
    "for iteration in range(epochs):\n",
    "    start=time.time()\n",
    "    avg_cost = 0.\n",
    "    total_batch = int(len(train_questions)/batch_size)\n",
    "    for i in range(total_batch):\n",
    "        fd, X, Y = next_feed(batch_size, i, True)\n",
    "        _, cost = sess.run([train_op, loss], feed_dict=fd)\n",
    "        avg_cost += cost/total_batch\n",
    "        if i > 0 and i % 100 == 0:\n",
    "            print(\"Partial Iteration \" + str(i) + \"/\" + str(total_batch) +\n",
    "                  \" :  \"+ '%04d' % (iteration + 1)+ \" cost= \"+ \"{:.9f}\".format(avg_cost))\n",
    "    end=time.time()\n",
    "    \n",
    "    print(\"Training Latency: \"+str(round((end-start),2))+\" seconds\"+\"\\n\")\n",
    "    if (iteration+1) %5 == 0  and iteration > 0:\n",
    "        print(\"\\nTesting...\")\n",
    "        out = [test(sess, val_questions[i], i, val_answers[i]) for i in range(len(val_questions))]\n",
    "        ref_arr = np.array(val_answers)\n",
    "\n",
    "        greedies = np.array([i[0] for i in out])\n",
    "        first_beams = np.array([i[1] for i in out])\n",
    "        last_beams = np.array([i[2] for i in out])\n",
    "\n",
    "        try:\n",
    "            bleu_geedy = moses_multi_bleu(greedies, ref_arr)\n",
    "            bleu_first_beam = moses_multi_bleu(first_beams, ref_arr)\n",
    "            bleu_last_beam = moses_multi_bleu(last_beams, ref_arr)\n",
    "\n",
    "            print(\"bleu test  (greedy):     \" + str(bleu_geedy))\n",
    "            print(\"bleu test  (first beam): \" + str(bleu_first_beam))\n",
    "            print(\"bleu test  (last beam):  \" + str(bleu_last_beam))\n",
    "\n",
    "            with open(saver_folder + \"/bleu.txt\", \"a\") as f:\n",
    "                f.write(\"Test \" + str(iteration+1) + \"\\n\")\n",
    "                f.write(\"- bleu test  (greedy)    : \" + str(bleu_geedy)+\"\\n\")\n",
    "                f.write(\"- bleu test  (first beam): \" + str(bleu_first_beam)+\"\\n\")\n",
    "                f.write(\"- bleu test  (last beam) : \" + str(bleu_last_beam)+\"\\n\\n\")\n",
    "        except ZeroDivisionError:\n",
    "            print (\"Exception ZeroDivisionError: float division by zero !\")\n",
    "\n",
    "        saver.save(sess, saver_folder+'/translate'+str(iteration+1))\n",
    "        print(\"Model Saved!\")\n",
    "    print(\"Iteration: \"+ '%04d' % (iteration + 1)+ \" cost= \"+ \"{:.9f}\".format(avg_cost))\n",
    "    tf.summary.scalar('loss', avg_cost)\n",
    "\n",
    "print(\"Tuning Completed!\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.5.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}