transformer_model.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"transformer_model.ipynb","version":"0.3.2","provenance":[],"collapsed_sections":[]},"kernelspec":{"name":"python3","display_name":"Python 3"},"accelerator":"GPU"},"cells":[{"metadata":{"id":"iK4_ICF0ZQDW","colab_type":"code","outputId":"b246957e-bcba-49ed-d52f-147789d96ef5","executionInfo":{"status":"ok","timestamp":1554613382546,"user_tz":420,"elapsed":9695,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":289}},"cell_type":"code","source":["!pip install sentencepiece\n","!pip install segtok"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Collecting sentencepiece\n","\u001b[?25l  Downloading https://files.pythonhosted.org/packages/7e/8a/0e4a10bc00a0263db8d45d0062c83892598eb58e8091f439c63926e9b107/sentencepiece-0.1.81-cp36-cp36m-manylinux1_x86_64.whl (1.0MB)\n","\u001b[K    100% |████████████████████████████████| 1.0MB 19.1MB/s \n","\u001b[?25hInstalling collected packages: sentencepiece\n","Successfully installed sentencepiece-0.1.81\n","Collecting segtok\n","  Downloading https://files.pythonhosted.org/packages/1d/59/6ed78856ab99d2da04084b59e7da797972baa0efecb71546b16d48e49d9b/segtok-1.5.7.tar.gz\n","Requirement already satisfied: regex in /usr/local/lib/python3.6/dist-packages (from segtok) (2018.1.10)\n","Building wheels for collected packages: segtok\n","  Building wheel for segtok (setup.py) ... \u001b[?25ldone\n","\u001b[?25h  Stored in directory: /root/.cache/pip/wheels/15/ee/a8/6112173f1386d33eebedb3f73429cfa41a4c3084556bcee254\n","Successfully built segtok\n","Installing collected packages: segtok\n","Successfully installed segtok-1.5.7\n"],"name":"stdout"}]},{"metadata":{"id":"hsIwgSmeZXsI","colab_type":"code","outputId":"1544ab58-8c7e-49a1-d4f3-f79d9e2494ad","executionInfo":{"status":"ok","timestamp":1554613404050,"user_tz":420,"elapsed":30951,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":127}},"cell_type":"code","source":["from google.colab import drive\n","drive.mount('/content/gdrive')"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Go to this URL in a browser: https://accounts.google.com/o/oauth2/auth?client_id=947318989803-6bn6qk8qdgf4n4g3pfee6491hc0brc4i.apps.googleusercontent.com&redirect_uri=urn%3Aietf%3Awg%3Aoauth%3A2.0%3Aoob&scope=email%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdocs.test%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fdrive.photos.readonly%20https%3A%2F%2Fwww.googleapis.com%2Fauth%2Fpeopleapi.readonly&response_type=code\n","\n","Enter your authorization code:\n","··········\n","Mounted at /content/gdrive\n"],"name":"stdout"}]},{"metadata":{"id":"WX0ngYn5Z44L","colab_type":"code","colab":{}},"cell_type":"code","source":["import sentencepiece as spm\n","import tensorflow as tf\n","import numpy as np\n","import os, sys\n","import json\n","import matplotlib.pyplot as plt\n","import pandas as pd\n","import string\n","\n","sys.path.insert(0, \"/content/gdrive/My Drive/Project-Lion/\") # This enables us to import Python libraries in the folder.\n","\n","from transformer import Transformer\n","import capita\n","\n","root_folder = \"/content/gdrive/My Drive/Project-Lion/\""],"execution_count":0,"outputs":[]},{"metadata":{"id":"qAAZ5PR5Zewp","colab_type":"code","colab":{}},"cell_type":"code","source":["# Load the word piece model that will be used to tokenize the texts into\n","# word pieces with a vocabulary size of 10000\n","\n","sp = spm.SentencePieceProcessor()\n","sp.Load(root_folder+\"dataset/wp_vocab10000.model\")\n","\n","vocab = [line.split('\\t')[0] for line in open(root_folder+\"dataset/wp_vocab10000.vocab\", \"r\")]\n","pad_index = vocab.index('#')"],"execution_count":0,"outputs":[]},{"metadata":{"id":"wpFZQX6NbxuP","colab_type":"code","colab":{}},"cell_type":"code","source":["df = pd.read_csv(root_folder + 'data/kaggle/shortjokes.csv')\n","df = df.loc[df[\"Joke\"] != \"\"]"],"execution_count":0,"outputs":[]},{"metadata":{"id":"4hBmUZx5cGwR","colab_type":"code","colab":{}},"cell_type":"code","source":["def clean_data(lst, in_place=True, keep_punc=\"\"):\n","    if in_place:\n","        new_lst = lst\n","    else:\n","        new_lst = lst.copy()\n","        \n","    for i in range(len(new_lst)):\n","        joke = lst[i]\n","        joke = joke.replace(\"\\r\", \" \").replace(\"\\n\", \" \").replace(\"/\", \" or \")\n","        joke = joke.translate(str.maketrans(\"\", \"\", string.punctuation.replace(keep_punc, \"\"))) # Remove punctuation\n","        joke = joke.lower() # Lowercase\n","        joke = ''.join(char for char in joke if not char.isdigit()) # Remove numbers\n","        joke = joke.strip() # Remove leading and ending whitespace\n","        new_lst[i] = joke\n","    return new_lst"],"execution_count":0,"outputs":[]},{"metadata":{"id":"3su6cKOPdMVD","colab_type":"code","colab":{}},"cell_type":"code","source":["data = clean_data(df[\"Joke\"].tolist(), keep_punc=\"?\")"],"execution_count":0,"outputs":[]},{"metadata":{"id":"oNvJj6jUdTVv","colab_type":"code","outputId":"caa6a5d6-a886-4f58-b14b-c80d76b2c23d","executionInfo":{"status":"ok","timestamp":1554613416232,"user_tz":420,"elapsed":42004,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":53}},"cell_type":"code","source":["qa_all = [[s.strip() for s in list(filter(None, joke.split(\"?\")))] for joke in data if len(list(filter(None, joke.split(\"?\")))) == 2]\n","print(\"Size of entire dataset:\", len(qa_all))\n","print(qa_all[0])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Size of entire dataset: 93782\n","['why cant barbie get pregnant', 'because ken comes in a different box heyooooooo']\n"],"name":"stdout"}]},{"metadata":{"id":"JRdCL3KFjcBe","colab_type":"code","outputId":"6129a0b1-7113-4962-84ae-1907ae304bc6","executionInfo":{"status":"ok","timestamp":1554613416770,"user_tz":420,"elapsed":42357,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":53}},"cell_type":"code","source":["min_qlen, max_qlen = 3, 15\n","min_alen, max_alen = 3, 15\n","qa = [(q, a) for q, a in qa_all if len(q.split()) >= min_qlen and len(q.split()) <= max_qlen and len(a.split()) >= min_alen and len(a.split()) <= max_alen]\n","print(\"Size of dataset:\", len(qa))\n","print(qa[0])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Size of dataset: 59999\n","('why cant barbie get pregnant', 'because ken comes in a different box heyooooooo')\n"],"name":"stdout"}]},{"metadata":{"id":"ZMeNKfZ4i6yq","colab_type":"code","colab":{}},"cell_type":"code","source":["def pad_sequence(numerized, pad_index, to_length):\n","    pad = numerized[:to_length]\n","    padded = pad + [pad_index] * (to_length - len(pad))\n","    mask = [w != pad_index for w in padded]\n","    return padded, mask"],"execution_count":0,"outputs":[]},{"metadata":{"id":"w9yW16IwizC-","colab_type":"code","colab":{}},"cell_type":"code","source":["input_length = max_qlen\n","output_length = max_alen\n","\n","dataset = []\n","for q, a in qa:\n","  d = {}\n","  # Tokenize, numerize\n","  d['input'] = sp.EncodeAsIds(capita.preprocess_capitalization(q))\n","  d['output'] = sp.EncodeAsIds(capita.preprocess_capitalization(a))\n","\n","  # Padding    \n","  d['input'], d['input_mask'] =  pad_sequence(d['input'],  pad_index, input_length)\n","  d['output'], d['output_mask'] = pad_sequence(d['output'], pad_index, output_length)\n","  dataset.append(d)"],"execution_count":0,"outputs":[]},{"metadata":{"id":"KarV3qGBhBrx","colab_type":"code","outputId":"d1f0bb37-4b01-4dd1-cb47-9bd497b0f261","executionInfo":{"status":"ok","timestamp":1554613438380,"user_tz":420,"elapsed":63302,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":35}},"cell_type":"code","source":["dataset_size = len(dataset)\n","print(dataset_size)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["59999\n"],"name":"stdout"}]},{"metadata":{"id":"4KfkU9zWhkwk","colab_type":"code","outputId":"93f71610-97b1-4674-824a-888ad7f531f1","executionInfo":{"status":"ok","timestamp":1554613438382,"user_tz":420,"elapsed":62940,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":35}},"cell_type":"code","source":["training_size = (dataset_size * 9) // 10\n","\n","d_train = dataset[:training_size]\n","d_valid = dataset[training_size:]\n","\n","len(d_train), len(d_valid)"],"execution_count":0,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(53999, 6000)"]},"metadata":{"tags":[]},"execution_count":13}]},{"metadata":{"id":"5Nsl8T8BhQaM","colab_type":"code","colab":{}},"cell_type":"code","source":["def build_batch(dataset, batch_size):\n","    indices = list(np.random.randint(0, len(dataset), size=batch_size))\n","    \n","    batch = [dataset[i] for i in indices]\n","    batch_input = np.array([a['input'] for a in batch])\n","    batch_input_mask = np.array([a['input_mask'] for a in batch])\n","    batch_output = np.array([a['output'] for a in batch])\n","    batch_output_mask = np.array([a['output_mask'] for a in batch])\n","    \n","    return batch_input, batch_input_mask, batch_output, batch_output_mask"],"execution_count":0,"outputs":[]},{"metadata":{"id":"57xhDbShbU7D","colab_type":"code","colab":{}},"cell_type":"code","source":["# We are giving you the trainer, as it is similar to the one\n","# you created in the Language Modeling assignment.\n","\n","class TransformerTrainer():\n","\n","    def __init__(self, vocab_size, d_model, input_length, output_length, n_layers, d_filter, learning_rate=1e-3):\n","\n","        self.source_sequence = tf.placeholder(tf.int32,shape=(None,input_length), name=\"source_sequence\")\n","        self.target_sequence = tf.placeholder(tf.int32, shape=(None,output_length),name=\"target_sequence\")\n","        self.encoder_mask = tf.placeholder(tf.bool,shape=(None,input_length),name=\"encoder_mask\")\n","        self.decoder_mask = tf.placeholder(tf.bool, shape=(None,output_length),name=\"decoder_mask\")\n","\n","        self.model = Transformer(vocab_size=vocab_size, d_model=d_model, n_layers=n_layers, d_filter=d_filter)\n","\n","        self.decoded_logits = self.model(self.source_sequence, self.target_sequence, encoder_mask=self.encoder_mask, decoder_mask=self.decoder_mask)\n","        self.global_step = tf.Variable(0, trainable=False)\n","        learning_rate = tf.train.exponential_decay(learning_rate, self.global_step, 100000, 0.96)\n","        \n","        # Summarization loss\n","        self.loss = tf.losses.sparse_softmax_cross_entropy(self.target_sequence, self.decoded_logits, tf.cast(self.decoder_mask, tf.float32))\n","        self.optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate, beta1=0.9, beta2=0.98, epsilon=1e-9)\n","        self.train_op = self.optimizer.minimize(self.loss, global_step=self.global_step)\n","        self.saver = tf.train.Saver()"],"execution_count":0,"outputs":[]},{"metadata":{"id":"ZUT-8rc0Eir4","colab_type":"code","colab":{}},"cell_type":"code","source":["def predict(model, model_file, questions=[], size=10, start_idx=None):\n","  \n","  sample = questions == []\n","  if sample:\n","    if start_idx == None:\n","      start_idx = np.random.randint(training_size, len(qa) - size)\n","    questions = [q for q, a in qa[start_idx:start_idx + size]]\n","    answers = [a for q, a in qa[start_idx:start_idx + size]]\n","\n","  with tf.Session() as sess:\n","\n","    model.saver.restore(sess, model_file)\n","    \n","    if sample:\n","      print(\"\\nJokes starting from index:\", start_idx)\n","\n","    for i in range(len(questions)):\n","      question = questions[i]\n","\n","      # Process the capitalization with the preprocess_capitalization of the capita package.\n","      question = capita.preprocess_capitalization(question)\n","\n","      # Numerize the tokens of the processed text using the loaded sentencepiece model.\n","      numerized = sp.EncodeAsIds(question)\n","      # Pad the sequence and keep the mask of the input\n","      padded, mask = pad_sequence(numerized, pad_index, input_len)\n","\n","      # Making the news article into a batch of size one, to be fed to the neural network.\n","      encoder_input = np.array([padded])\n","      encoder_mask = np.array([mask])\n","\n","      decoded_so_far = [0]\n","\n","      for j in range(output_len):\n","        padded_decoder_input, decoder_mask = pad_sequence(decoded_so_far, pad_index, output_len)\n","        padded_decoder_input = [padded_decoder_input]\n","        decoder_mask = [decoder_mask]\n","        #print(\"========================\")\n","        #print(padded_decoder_input)\n","        # Use the model to find the distrbution over the vocabulary for the next word\n","        logits = sess.run(model.decoded_logits, feed_dict={model.source_sequence: encoder_input, model.encoder_mask: encoder_mask, model.target_sequence: padded_decoder_input, model.decoder_mask: decoder_mask})\n","\n","        chosen_words = np.argmax(logits, axis=2) # Take the argmax, getting the most likely next word\n","        decoded_so_far.append(int(chosen_words[0, j])) # We add it to the summary so far\n","\n","      print(\"\\n=================================================\\n\")\n","      print(\"Question:\", question)\n","      if sample:\n","        answer = answers[i]\n","        print(\"Answer:\", answer)\n","      print(\"Predicted answer:\", \"\".join([vocab[i] for i in decoded_so_far]).replace(\"▁\", \" \"))"],"execution_count":0,"outputs":[]},{"metadata":{"id":"G-OwVG853adl","colab_type":"text"},"cell_type":"markdown","source":["# Initialize new model"]},{"metadata":{"id":"BA_OQG5_URar","colab_type":"code","colab":{}},"cell_type":"code","source":["# Dataset related parameters\n","vocab_size = len(vocab)\n","input_len = max_qlen # Length of the article\n","output_len = max_alen # Length of the summaries\n","\n","# Model related parameters, feel free to modify these.\n","n_layers = 6\n","d_model  = 256\n","d_filter = 416\n","\n","tf.reset_default_graph()\n","model = TransformerTrainer(vocab_size, d_model, input_len, output_len, n_layers, d_filter, learning_rate=1e-3)"],"execution_count":0,"outputs":[]},{"metadata":{"id":"eGyy2173EfeM","colab_type":"text"},"cell_type":"markdown","source":["# Initial Training"]},{"metadata":{"id":"na3yBYgMbnaH","colab_type":"code","outputId":"5624c945-14a9-49b5-ac4c-fef3ae37edaa","executionInfo":{"status":"error","timestamp":1554594729063,"user_tz":420,"elapsed":12471,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":6075}},"cell_type":"code","source":["model_file = root_folder+\"models/transformer_model\"\n","\n","batch_size = 50\n","epochs = 10\n","\n","iter_per_epoch = len(d_train) // batch_size\n","total_iterations = iter_per_epoch * epochs\n","print_per_iter = 10\n","eval_per_iter = 50\n","save_per_iter = 100\n","\n","print(\"Total iterations:\", total_iterations)\n","\n","train_losses = []\n","val_losses = []\n","iterations = []\n","\n","with tf.Session() as sess:\n","    sess.run(tf.global_variables_initializer())\n","    iteration = 1\n","    for epoch in range(1, epochs + 1):\n","        print(\"Epoch:\", epoch)\n","        for _ in range(iter_per_epoch):\n","            batch_input, batch_input_mask, batch_output, batch_output_mask = build_batch(d_train, batch_size)\n","            feed = {model.source_sequence: batch_input, model.target_sequence: batch_output,\n","                              model.encoder_mask: batch_input_mask, model.decoder_mask: batch_output_mask}\n","\n","            train_loss, _, step = sess.run([model.loss, model.train_op, model.global_step], feed_dict=feed)\n","\n","            if iteration % print_per_iter == 0:\n","                print(\"Iteration {0} / {1}: Training loss - {2}\".format(iteration, total_iterations, train_loss))\n","\n","            if iteration % eval_per_iter == 0:\n","                batch_input, batch_input_mask, batch_output, batch_output_mask = build_batch(d_valid, batch_size)\n","                feed = {model.source_sequence: batch_input, model.target_sequence: batch_output,\n","                              model.encoder_mask: batch_input_mask, model.decoder_mask: batch_output_mask}\n","                eval_loss = sess.run(model.loss, feed_dict=feed)\n","                print(\"Evaluation loss:\", eval_loss)\n","                val_losses.append(eval_loss)\n","                train_losses.append(train_loss)\n","                iterations.append(iteration)\n","                \n","            if iteration % save_per_iter == 0:\n","                model.saver.save(sess, model_file)\n","                np.save(root_folder + \"models/transformer_model_train_losses\", np.array(train_losses))\n","                np.save(root_folder + \"models/transformer_model_val_losses\", np.array(val_losses))\n","                np.save(root_folder + \"models/transformer_model_iterations\", np.array(iterations))\n","                \n","            iteration += 1"],"execution_count":0,"outputs":[{"output_type":"stream","text":["Total iterations: 10790\n","Epoch: 1\n","Iteration 10 / 10790: Training loss - 9.302825927734375\n","Iteration 20 / 10790: Training loss - 8.529587745666504\n","Iteration 30 / 10790: Training loss - 8.6222562789917\n","Iteration 40 / 10790: Training loss - 8.441946029663086\n","Iteration 50 / 10790: Training loss - 8.565862655639648\n","Evaluation loss: 8.226765\n","Iteration 60 / 10790: Training loss - 8.36087703704834\n","Iteration 70 / 10790: Training loss - 8.175726890563965\n","Iteration 80 / 10790: Training loss - 8.144606590270996\n","Iteration 90 / 10790: Training loss - 8.354171752929688\n","Iteration 100 / 10790: Training loss - 8.107654571533203\n","Evaluation loss: 8.264207\n","Iteration 110 / 10790: Training loss - 8.176652908325195\n","Iteration 120 / 10790: Training loss - 7.914579391479492\n","Iteration 130 / 10790: Training loss - 8.302800178527832\n","Iteration 140 / 10790: Training loss - 8.044289588928223\n","Iteration 150 / 10790: Training loss - 7.804849624633789\n","Evaluation loss: 7.8036613\n","Iteration 160 / 10790: Training loss - 7.817606449127197\n","Iteration 170 / 10790: Training loss - 7.58344841003418\n","Iteration 180 / 10790: Training loss - 7.8089070320129395\n","Iteration 190 / 10790: Training loss - 7.844474792480469\n","Iteration 200 / 10790: Training loss - 7.748758792877197\n","Evaluation loss: 7.7675714\n","Iteration 210 / 10790: Training loss - 7.642362594604492\n","Iteration 220 / 10790: Training loss - 7.341987133026123\n","Iteration 230 / 10790: Training loss - 7.208065509796143\n","Iteration 240 / 10790: Training loss - 7.605661869049072\n","Iteration 250 / 10790: Training loss - 7.316489219665527\n","Evaluation loss: 7.0339265\n","Iteration 260 / 10790: Training loss - 7.333079814910889\n","Iteration 270 / 10790: Training loss - 7.1616058349609375\n","Iteration 280 / 10790: Training loss - 7.305466175079346\n","Iteration 290 / 10790: Training loss - 7.242315292358398\n","Iteration 300 / 10790: Training loss - 6.707549095153809\n","Evaluation loss: 7.145681\n","Iteration 310 / 10790: Training loss - 6.860950946807861\n","Iteration 320 / 10790: Training loss - 7.211257457733154\n","Iteration 330 / 10790: Training loss - 7.188359260559082\n","Iteration 340 / 10790: Training loss - 6.740680694580078\n","Iteration 350 / 10790: Training loss - 6.945425987243652\n","Evaluation loss: 7.09036\n","Iteration 360 / 10790: Training loss - 6.770499229431152\n","Iteration 370 / 10790: Training loss - 6.884030342102051\n","Iteration 380 / 10790: Training loss - 6.8821821212768555\n","Iteration 390 / 10790: Training loss - 6.535759925842285\n","Iteration 400 / 10790: Training loss - 6.605838298797607\n","Evaluation loss: 7.018873\n","Iteration 410 / 10790: Training loss - 7.129065990447998\n","Iteration 420 / 10790: Training loss - 6.9036149978637695\n","Iteration 430 / 10790: Training loss - 6.732597827911377\n","Iteration 440 / 10790: Training loss - 6.747409343719482\n","Iteration 450 / 10790: Training loss - 6.854428768157959\n","Evaluation loss: 7.024675\n","Iteration 460 / 10790: Training loss - 6.962937355041504\n","Iteration 470 / 10790: Training loss - 6.3652825355529785\n","Iteration 480 / 10790: Training loss - 6.822086334228516\n","Iteration 490 / 10790: Training loss - 6.526680946350098\n","Iteration 500 / 10790: Training loss - 6.691030502319336\n","Evaluation loss: 6.677663\n","Iteration 510 / 10790: Training loss - 6.229783535003662\n","Iteration 520 / 10790: Training loss - 6.374362468719482\n","Iteration 530 / 10790: Training loss - 6.759851932525635\n","Iteration 540 / 10790: Training loss - 6.474534511566162\n","Iteration 550 / 10790: Training loss - 6.3078837394714355\n","Evaluation loss: 6.4679894\n","Iteration 560 / 10790: Training loss - 6.306136131286621\n","Iteration 570 / 10790: Training loss - 6.822094917297363\n","Iteration 580 / 10790: Training loss - 6.493320465087891\n","Iteration 590 / 10790: Training loss - 6.690279483795166\n","Iteration 600 / 10790: Training loss - 6.20764684677124\n","Evaluation loss: 6.5082836\n","Iteration 610 / 10790: Training loss - 6.724930286407471\n","Iteration 620 / 10790: Training loss - 6.468964099884033\n","Iteration 630 / 10790: Training loss - 6.431088924407959\n","Iteration 640 / 10790: Training loss - 6.130812644958496\n","Iteration 650 / 10790: Training loss - 6.740591526031494\n","Evaluation loss: 6.1770205\n","Iteration 660 / 10790: Training loss - 6.494958400726318\n","Iteration 670 / 10790: Training loss - 6.420289993286133\n","Iteration 680 / 10790: Training loss - 6.401297092437744\n","Iteration 690 / 10790: Training loss - 5.975399971008301\n","Iteration 700 / 10790: Training loss - 6.29823637008667\n","Evaluation loss: 5.988657\n","Iteration 710 / 10790: Training loss - 6.109473705291748\n","Iteration 720 / 10790: Training loss - 6.448416709899902\n","Iteration 730 / 10790: Training loss - 6.132540702819824\n","Iteration 740 / 10790: Training loss - 6.434757232666016\n","Iteration 750 / 10790: Training loss - 6.168873310089111\n","Evaluation loss: 6.214454\n","Iteration 760 / 10790: Training loss - 6.432871341705322\n","Iteration 770 / 10790: Training loss - 6.304025173187256\n","Iteration 780 / 10790: Training loss - 6.156185150146484\n","Iteration 790 / 10790: Training loss - 6.376759052276611\n","Iteration 800 / 10790: Training loss - 6.457028865814209\n","Evaluation loss: 6.389462\n","Iteration 810 / 10790: Training loss - 5.932460784912109\n","Iteration 820 / 10790: Training loss - 5.709438323974609\n","Iteration 830 / 10790: Training loss - 6.281424522399902\n","Iteration 840 / 10790: Training loss - 5.836631774902344\n","Iteration 850 / 10790: Training loss - 5.997111797332764\n","Evaluation loss: 6.451051\n","Iteration 860 / 10790: Training loss - 6.459415435791016\n","Iteration 870 / 10790: Training loss - 5.912067413330078\n","Iteration 880 / 10790: Training loss - 6.206467151641846\n","Iteration 890 / 10790: Training loss - 6.224895477294922\n","Iteration 900 / 10790: Training loss - 5.7402496337890625\n","Evaluation loss: 6.452026\n","Iteration 910 / 10790: Training loss - 6.038321018218994\n","Iteration 920 / 10790: Training loss - 6.104475975036621\n","Iteration 930 / 10790: Training loss - 6.125846862792969\n","Iteration 940 / 10790: Training loss - 5.938140869140625\n","Iteration 950 / 10790: Training loss - 5.438774108886719\n","Evaluation loss: 5.942171\n","Iteration 960 / 10790: Training loss - 6.14271879196167\n","Iteration 970 / 10790: Training loss - 5.757171154022217\n","Iteration 980 / 10790: Training loss - 5.816136360168457\n","Iteration 990 / 10790: Training loss - 6.10329532623291\n","Iteration 1000 / 10790: Training loss - 6.093562602996826\n","Evaluation loss: 6.028907\n","Iteration 1010 / 10790: Training loss - 6.037572383880615\n","Iteration 1020 / 10790: Training loss - 6.208059787750244\n","Iteration 1030 / 10790: Training loss - 6.156642913818359\n","Iteration 1040 / 10790: Training loss - 6.351938724517822\n","Iteration 1050 / 10790: Training loss - 5.656499862670898\n","Evaluation loss: 6.376241\n","Iteration 1060 / 10790: Training loss - 5.492630958557129\n","Iteration 1070 / 10790: Training loss - 6.039140701293945\n","Epoch: 2\n","Iteration 1080 / 10790: Training loss - 6.115725517272949\n","Iteration 1090 / 10790: Training loss - 5.719571590423584\n","Iteration 1100 / 10790: Training loss - 5.579849720001221\n","Evaluation loss: 6.011796\n","Iteration 1110 / 10790: Training loss - 5.884855270385742\n","Iteration 1120 / 10790: Training loss - 5.694131374359131\n","Iteration 1130 / 10790: Training loss - 5.905986309051514\n","Iteration 1140 / 10790: Training loss - 6.220251560211182\n","Iteration 1150 / 10790: Training loss - 5.9191718101501465\n","Evaluation loss: 6.1266823\n","Iteration 1160 / 10790: Training loss - 6.03022575378418\n","Iteration 1170 / 10790: Training loss - 6.323237895965576\n","Iteration 1180 / 10790: Training loss - 5.825687408447266\n","Iteration 1190 / 10790: Training loss - 5.795993328094482\n","Iteration 1200 / 10790: Training loss - 5.956028938293457\n","Evaluation loss: 6.206112\n","Iteration 1210 / 10790: Training loss - 5.712897777557373\n","Iteration 1220 / 10790: Training loss - 5.61391544342041\n","Iteration 1230 / 10790: Training loss - 6.100117206573486\n","Iteration 1240 / 10790: Training loss - 5.417026996612549\n","Iteration 1250 / 10790: Training loss - 5.810816764831543\n","Evaluation loss: 6.6538897\n","Iteration 1260 / 10790: Training loss - 5.795987129211426\n","Iteration 1270 / 10790: Training loss - 5.83464241027832\n","Iteration 1280 / 10790: Training loss - 5.389614105224609\n","Iteration 1290 / 10790: Training loss - 5.792092323303223\n","Iteration 1300 / 10790: Training loss - 5.834153652191162\n","Evaluation loss: 6.0400524\n","Iteration 1310 / 10790: Training loss - 5.862401485443115\n","Iteration 1320 / 10790: Training loss - 5.840102672576904\n","Iteration 1330 / 10790: Training loss - 5.849576950073242\n","Iteration 1340 / 10790: Training loss - 5.17535400390625\n","Iteration 1350 / 10790: Training loss - 5.8843841552734375\n","Evaluation loss: 5.740416\n","Iteration 1360 / 10790: Training loss - 5.652781009674072\n","Iteration 1370 / 10790: Training loss - 5.647082328796387\n","Iteration 1380 / 10790: Training loss - 5.743297100067139\n","Iteration 1390 / 10790: Training loss - 5.939007759094238\n","Iteration 1400 / 10790: Training loss - 5.719480037689209\n","Evaluation loss: 6.0662956\n","Iteration 1410 / 10790: Training loss - 5.5181169509887695\n","Iteration 1420 / 10790: Training loss - 6.12783145904541\n","Iteration 1430 / 10790: Training loss - 5.61575984954834\n","Iteration 1440 / 10790: Training loss - 5.703380584716797\n","Iteration 1450 / 10790: Training loss - 5.301906585693359\n","Evaluation loss: 6.255951\n","Iteration 1460 / 10790: Training loss - 5.336485862731934\n","Iteration 1470 / 10790: Training loss - 6.2142486572265625\n","Iteration 1480 / 10790: Training loss - 5.8313398361206055\n","Iteration 1490 / 10790: Training loss - 5.670263290405273\n","Iteration 1500 / 10790: Training loss - 5.434608459472656\n","Evaluation loss: 5.775607\n","Iteration 1510 / 10790: Training loss - 5.527533054351807\n","Iteration 1520 / 10790: Training loss - 5.354048252105713\n","Iteration 1530 / 10790: Training loss - 5.845536708831787\n","Iteration 1540 / 10790: Training loss - 5.765321254730225\n","Iteration 1550 / 10790: Training loss - 5.640746116638184\n","Evaluation loss: 5.7978477\n","Iteration 1560 / 10790: Training loss - 5.7176032066345215\n","Iteration 1570 / 10790: Training loss - 5.499831199645996\n","Iteration 1580 / 10790: Training loss - 5.661712169647217\n","Iteration 1590 / 10790: Training loss - 5.296733856201172\n","Iteration 1600 / 10790: Training loss - 5.831387519836426\n","Evaluation loss: 5.9135375\n","Iteration 1610 / 10790: Training loss - 5.889135360717773\n","Iteration 1620 / 10790: Training loss - 5.55054235458374\n","Iteration 1630 / 10790: Training loss - 5.410459041595459\n","Iteration 1640 / 10790: Training loss - 5.997206687927246\n","Iteration 1650 / 10790: Training loss - 5.476696491241455\n","Evaluation loss: 5.350567\n","Iteration 1660 / 10790: Training loss - 5.720706462860107\n","Iteration 1670 / 10790: Training loss - 5.829702854156494\n","Iteration 1680 / 10790: Training loss - 5.511807441711426\n","Iteration 1690 / 10790: Training loss - 5.468271255493164\n","Iteration 1700 / 10790: Training loss - 5.068452835083008\n","Evaluation loss: 5.6748476\n","Iteration 1710 / 10790: Training loss - 5.331204414367676\n","Iteration 1720 / 10790: Training loss - 5.993372917175293\n","Iteration 1730 / 10790: Training loss - 5.094583988189697\n","Iteration 1740 / 10790: Training loss - 5.865930080413818\n","Iteration 1750 / 10790: Training loss - 5.599820613861084\n","Evaluation loss: 5.651089\n","Iteration 1760 / 10790: Training loss - 5.0971150398254395\n","Iteration 1770 / 10790: Training loss - 5.9439473152160645\n","Iteration 1780 / 10790: Training loss - 5.2008209228515625\n","Iteration 1790 / 10790: Training loss - 5.488217830657959\n","Iteration 1800 / 10790: Training loss - 5.558729648590088\n","Evaluation loss: 5.467077\n","Iteration 1810 / 10790: Training loss - 5.247640132904053\n","Iteration 1820 / 10790: Training loss - 5.7396159172058105\n","Iteration 1830 / 10790: Training loss - 5.1754350662231445\n","Iteration 1840 / 10790: Training loss - 5.256754398345947\n","Iteration 1850 / 10790: Training loss - 5.439314842224121\n","Evaluation loss: 6.220651\n","Iteration 1860 / 10790: Training loss - 5.258166790008545\n","Iteration 1870 / 10790: Training loss - 5.274233818054199\n","Iteration 1880 / 10790: Training loss - 5.432887077331543\n","Iteration 1890 / 10790: Training loss - 5.732486248016357\n","Iteration 1900 / 10790: Training loss - 5.514720916748047\n","Evaluation loss: 5.0509305\n","Iteration 1910 / 10790: Training loss - 5.38808536529541\n","Iteration 1920 / 10790: Training loss - 5.7854485511779785\n","Iteration 1930 / 10790: Training loss - 5.444493293762207\n","Iteration 1940 / 10790: Training loss - 5.234443187713623\n","Iteration 1950 / 10790: Training loss - 5.507816791534424\n","Evaluation loss: 5.5985494\n","Iteration 1960 / 10790: Training loss - 5.4874725341796875\n","Iteration 1970 / 10790: Training loss - 5.709356307983398\n","Iteration 1980 / 10790: Training loss - 5.406312465667725\n","Iteration 1990 / 10790: Training loss - 5.44323205947876\n","Iteration 2000 / 10790: Training loss - 5.5007195472717285\n","Evaluation loss: 5.211809\n","Iteration 2010 / 10790: Training loss - 5.281378269195557\n","Iteration 2020 / 10790: Training loss - 5.014746189117432\n","Iteration 2030 / 10790: Training loss - 5.29028844833374\n","Iteration 2040 / 10790: Training loss - 5.613513469696045\n","Iteration 2050 / 10790: Training loss - 5.83323860168457\n","Evaluation loss: 5.4951677\n","Iteration 2060 / 10790: Training loss - 5.577224254608154\n","Iteration 2070 / 10790: Training loss - 5.001086235046387\n","Iteration 2080 / 10790: Training loss - 5.564448356628418\n","Iteration 2090 / 10790: Training loss - 5.117267608642578\n","Iteration 2100 / 10790: Training loss - 5.201091289520264\n","Evaluation loss: 5.1416216\n","Iteration 2110 / 10790: Training loss - 5.7683916091918945\n","Iteration 2120 / 10790: Training loss - 5.300034523010254\n","Iteration 2130 / 10790: Training loss - 5.502989768981934\n","Iteration 2140 / 10790: Training loss - 6.167428016662598\n","Iteration 2150 / 10790: Training loss - 5.515735149383545\n","Evaluation loss: 5.7604566\n","Epoch: 3\n","Iteration 2160 / 10790: Training loss - 4.914018630981445\n","Iteration 2170 / 10790: Training loss - 5.389779567718506\n","Iteration 2180 / 10790: Training loss - 5.2851691246032715\n","Iteration 2190 / 10790: Training loss - 5.632180690765381\n","Iteration 2200 / 10790: Training loss - 5.6478800773620605\n","Evaluation loss: 5.4779124\n","Iteration 2210 / 10790: Training loss - 5.073143005371094\n","Iteration 2220 / 10790: Training loss - 5.191973686218262\n","Iteration 2230 / 10790: Training loss - 5.285637378692627\n","Iteration 2240 / 10790: Training loss - 5.416747093200684\n","Iteration 2250 / 10790: Training loss - 5.309338569641113\n","Evaluation loss: 5.7824774\n","Iteration 2260 / 10790: Training loss - 5.437018394470215\n","Iteration 2270 / 10790: Training loss - 5.476558685302734\n","Iteration 2280 / 10790: Training loss - 5.161159515380859\n","Iteration 2290 / 10790: Training loss - 4.800003528594971\n","Iteration 2300 / 10790: Training loss - 5.555710315704346\n","Evaluation loss: 5.696549\n","Iteration 2310 / 10790: Training loss - 5.506260871887207\n","Iteration 2320 / 10790: Training loss - 4.839799880981445\n","Iteration 2330 / 10790: Training loss - 5.621054649353027\n","Iteration 2340 / 10790: Training loss - 5.407857418060303\n","Iteration 2350 / 10790: Training loss - 5.717645168304443\n","Evaluation loss: 5.343521\n","Iteration 2360 / 10790: Training loss - 5.327334403991699\n","Iteration 2370 / 10790: Training loss - 5.527626037597656\n","Iteration 2380 / 10790: Training loss - 5.435171127319336\n","Iteration 2390 / 10790: Training loss - 4.928947448730469\n","Iteration 2400 / 10790: Training loss - 4.974211692810059\n","Evaluation loss: 5.0487742\n","Iteration 2410 / 10790: Training loss - 5.276504993438721\n","Iteration 2420 / 10790: Training loss - 5.250876426696777\n","Iteration 2430 / 10790: Training loss - 5.500267505645752\n","Iteration 2440 / 10790: Training loss - 4.88502836227417\n","Iteration 2450 / 10790: Training loss - 5.409289360046387\n","Evaluation loss: 5.7215476\n","Iteration 2460 / 10790: Training loss - 5.648573398590088\n","Iteration 2470 / 10790: Training loss - 4.9987592697143555\n","Iteration 2480 / 10790: Training loss - 4.791787624359131\n","Iteration 2490 / 10790: Training loss - 5.363884449005127\n","Iteration 2500 / 10790: Training loss - 5.563549995422363\n","Evaluation loss: 5.3594394\n","Iteration 2510 / 10790: Training loss - 5.242987632751465\n","Iteration 2520 / 10790: Training loss - 5.528505325317383\n","Iteration 2530 / 10790: Training loss - 5.318052291870117\n","Iteration 2540 / 10790: Training loss - 5.031051158905029\n","Iteration 2550 / 10790: Training loss - 5.25830602645874\n","Evaluation loss: 5.3646545\n","Iteration 2560 / 10790: Training loss - 5.573902130126953\n","Iteration 2570 / 10790: Training loss - 4.779969215393066\n","Iteration 2580 / 10790: Training loss - 5.087569236755371\n","Iteration 2590 / 10790: Training loss - 5.146025657653809\n","Iteration 2600 / 10790: Training loss - 5.001367568969727\n","Evaluation loss: 5.298655\n"],"name":"stdout"},{"output_type":"error","ename":"KeyboardInterrupt","evalue":"ignored","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)","\u001b[0;32m<ipython-input-43-494abc7dd528>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     42\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     43\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0miteration\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0msave_per_iter\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 44\u001b[0;31m                 \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msaver\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msess\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel_file\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     45\u001b[0m                 \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mroot_folder\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m\"models/transformer_model_train_losses\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtrain_losses\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     46\u001b[0m                 \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msave\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mroot_folder\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m\"models/transformer_model_val_losses\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0marray\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mval_losses\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/training/saver.py\u001b[0m in \u001b[0;36msave\u001b[0;34m(self, sess, save_path, global_step, latest_filename, meta_graph_suffix, write_meta_graph, write_state, strip_default_attrs)\u001b[0m\n\u001b[1;32m   1194\u001b[0m         \u001b[0;32mwith\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mgraph\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_default\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1195\u001b[0m           self.export_meta_graph(\n\u001b[0;32m-> 1196\u001b[0;31m               meta_graph_filename, strip_default_attrs=strip_default_attrs)\n\u001b[0m\u001b[1;32m   1197\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1198\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_is_empty\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mKeyboardInterrupt\u001b[0m: "]}]},{"metadata":{"id":"0LAz8I983XEc","colab_type":"text"},"cell_type":"markdown","source":["# Continue Training Model"]},{"metadata":{"id":"YVt8HoCNnzUz","colab_type":"code","outputId":"83220fdb-2ca4-4147-c163-3cdd57cb56f5","executionInfo":{"status":"error","timestamp":1554616425452,"user_tz":420,"elapsed":2777548,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":20241}},"cell_type":"code","source":["model_file = root_folder + \"models/transformer_model\"\n","\n","batch_size = 50\n","epochs = 10\n","\n","iter_per_epoch = len(d_train) // batch_size\n","total_iterations = iter_per_epoch * epochs\n","print_per_iter = 10\n","eval_per_iter = 50\n","save_per_iter = 100\n","\n","train_losses = np.load(root_folder + \"models/transformer_model_train_losses.npy\").tolist()\n","val_losses = np.load(root_folder + \"models/transformer_model_val_losses.npy\").tolist()\n","iterations = np.load(root_folder + \"models/transformer_model_iterations.npy\").tolist()\n","\n","with tf.Session() as sess:\n","    model.saver.restore(sess, model_file)\n","    print(\"Continue running for\", total_iterations, \"iterations\")\n","    iteration = iterations[-1] + 1\n","    total_iterations = iteration + total_iterations - 1\n","    \n","    for epoch in range(1, epochs + 1):\n","        print(\"Epoch:\", epoch)\n","        for _ in range(iter_per_epoch):\n","            batch_input, batch_input_mask, batch_output, batch_output_mask = build_batch(d_train, batch_size)\n","            feed = {model.source_sequence: batch_input, model.target_sequence: batch_output,\n","                              model.encoder_mask: batch_input_mask, model.decoder_mask: batch_output_mask}\n","\n","            train_loss, _, step = sess.run([model.loss, model.train_op, model.global_step], feed_dict=feed)\n","\n","            if iteration % print_per_iter == 0:\n","                print(\"Iteration {0} / {1}: Training loss - {2}\".format(iteration, total_iterations, train_loss))\n","\n","            if iteration % eval_per_iter == 0:\n","                batch_input, batch_input_mask, batch_output, batch_output_mask = build_batch(d_valid, batch_size)\n","                feed = {model.source_sequence: batch_input, model.target_sequence: batch_output,\n","                              model.encoder_mask: batch_input_mask, model.decoder_mask: batch_output_mask}\n","                eval_loss = sess.run(model.loss, feed_dict=feed)\n","                print(\"Evaluation loss:\", eval_loss)\n","                val_losses.append(eval_loss)\n","                train_losses.append(train_loss)\n","                iterations.append(iteration)\n","                \n","            if iteration % save_per_iter == 0:\n","                model.saver.save(sess, model_file)\n","                np.save(root_folder + \"models/transformer_model_train_losses\", np.array(train_losses))\n","                np.save(root_folder + \"models/transformer_model_val_losses\", np.array(val_losses))\n","                np.save(root_folder + \"models/transformer_model_iterations\", np.array(iterations))\n","                \n","            iteration += 1"],"execution_count":0,"outputs":[{"output_type":"stream","text":["INFO:tensorflow:Restoring parameters from /content/gdrive/My Drive/Project-Lion/models/transformer_model\n","Continue running for 10790 iterations\n","Epoch: 1\n","Iteration 6610 / 17390: Training loss - 4.230494976043701\n","Iteration 6620 / 17390: Training loss - 4.12554931640625\n","Iteration 6630 / 17390: Training loss - 4.20909309387207\n","Iteration 6640 / 17390: Training loss - 4.141358375549316\n","Iteration 6650 / 17390: Training loss - 4.076033592224121\n","Evaluation loss: 4.7744427\n","Iteration 6660 / 17390: Training loss - 3.572007894515991\n","Iteration 6670 / 17390: Training loss - 4.504037857055664\n","Iteration 6680 / 17390: Training loss - 4.499098777770996\n","Iteration 6690 / 17390: Training loss - 4.149064064025879\n","Iteration 6700 / 17390: Training loss - 4.52952766418457\n","Evaluation loss: 4.6322284\n","Iteration 6710 / 17390: Training loss - 3.8780694007873535\n","Iteration 6720 / 17390: Training loss - 4.042656898498535\n","Iteration 6730 / 17390: Training loss - 4.074113368988037\n","Iteration 6740 / 17390: Training loss - 3.805556535720825\n","Iteration 6750 / 17390: Training loss - 4.337742805480957\n","Evaluation loss: 5.4408607\n","Iteration 6760 / 17390: Training loss - 4.011147975921631\n","Iteration 6770 / 17390: Training loss - 4.332068920135498\n","Iteration 6780 / 17390: Training loss - 4.322481155395508\n","Iteration 6790 / 17390: Training loss - 4.471817493438721\n","Iteration 6800 / 17390: Training loss - 3.8918421268463135\n","Evaluation loss: 5.5852804\n","Iteration 6810 / 17390: Training loss - 4.124536991119385\n","Iteration 6820 / 17390: Training loss - 4.5045952796936035\n","Iteration 6830 / 17390: Training loss - 4.539609432220459\n","Iteration 6840 / 17390: Training loss - 4.129691123962402\n","Iteration 6850 / 17390: Training loss - 3.993307113647461\n","Evaluation loss: 5.271112\n","Iteration 6860 / 17390: Training loss - 4.563857078552246\n","Iteration 6870 / 17390: Training loss - 4.125427722930908\n","Iteration 6880 / 17390: Training loss - 3.6313257217407227\n","Iteration 6890 / 17390: Training loss - 4.661851406097412\n","Iteration 6900 / 17390: Training loss - 4.245063781738281\n","Evaluation loss: 5.342045\n","Iteration 6910 / 17390: Training loss - 4.282598495483398\n","Iteration 6920 / 17390: Training loss - 3.816692590713501\n","Iteration 6930 / 17390: Training loss - 4.211063385009766\n","Iteration 6940 / 17390: Training loss - 4.6415629386901855\n","Iteration 6950 / 17390: Training loss - 4.390877723693848\n","Evaluation loss: 5.3874464\n","Iteration 6960 / 17390: Training loss - 5.019842147827148\n","Iteration 6970 / 17390: Training loss - 4.10995626449585\n","Iteration 6980 / 17390: Training loss - 4.2494120597839355\n","Iteration 6990 / 17390: Training loss - 4.521683692932129\n","Iteration 7000 / 17390: Training loss - 4.208292484283447\n","Evaluation loss: 4.633122\n","Iteration 7010 / 17390: Training loss - 4.309473991394043\n","Iteration 7020 / 17390: Training loss - 4.3499064445495605\n","Iteration 7030 / 17390: Training loss - 4.3448896408081055\n","Iteration 7040 / 17390: Training loss - 4.060855388641357\n","Iteration 7050 / 17390: Training loss - 3.809206008911133\n","Evaluation loss: 4.955728\n","Iteration 7060 / 17390: Training loss - 4.414841651916504\n","Iteration 7070 / 17390: Training loss - 3.779301881790161\n","Iteration 7080 / 17390: Training loss - 3.6866729259490967\n","Iteration 7090 / 17390: Training loss - 4.43890905380249\n","Iteration 7100 / 17390: Training loss - 4.293843746185303\n","Evaluation loss: 4.6561112\n","Iteration 7110 / 17390: Training loss - 3.7363827228546143\n","Iteration 7120 / 17390: Training loss - 3.8390424251556396\n","Iteration 7130 / 17390: Training loss - 3.8231332302093506\n","Iteration 7140 / 17390: Training loss - 3.816429615020752\n","Iteration 7150 / 17390: Training loss - 4.2977375984191895\n","Evaluation loss: 4.575856\n","Iteration 7160 / 17390: Training loss - 4.091454029083252\n","Iteration 7170 / 17390: Training loss - 3.9121506214141846\n","Iteration 7180 / 17390: Training loss - 3.6332151889801025\n","Iteration 7190 / 17390: Training loss - 4.293827533721924\n","Iteration 7200 / 17390: Training loss - 3.819704532623291\n","Evaluation loss: 5.6338387\n","Iteration 7210 / 17390: Training loss - 4.3201422691345215\n","Iteration 7220 / 17390: Training loss - 4.486114978790283\n","Iteration 7230 / 17390: Training loss - 4.151120185852051\n","Iteration 7240 / 17390: Training loss - 4.661431789398193\n","Iteration 7250 / 17390: Training loss - 3.9151594638824463\n","Evaluation loss: 5.267808\n","Iteration 7260 / 17390: Training loss - 4.082564830780029\n","Iteration 7270 / 17390: Training loss - 3.7152040004730225\n","Iteration 7280 / 17390: Training loss - 3.738962411880493\n","Iteration 7290 / 17390: Training loss - 4.169947624206543\n","Iteration 7300 / 17390: Training loss - 3.6197924613952637\n","Evaluation loss: 5.1545634\n","Iteration 7310 / 17390: Training loss - 4.2820258140563965\n","Iteration 7320 / 17390: Training loss - 4.430201530456543\n","Iteration 7330 / 17390: Training loss - 4.1257147789001465\n","Iteration 7340 / 17390: Training loss - 4.000830173492432\n","Iteration 7350 / 17390: Training loss - 3.4113829135894775\n","Evaluation loss: 4.987265\n","Iteration 7360 / 17390: Training loss - 3.90035343170166\n","Iteration 7370 / 17390: Training loss - 3.867050886154175\n","Iteration 7380 / 17390: Training loss - 3.9681262969970703\n","Iteration 7390 / 17390: Training loss - 4.157931804656982\n","Iteration 7400 / 17390: Training loss - 4.081980228424072\n","Evaluation loss: 5.5068493\n","Iteration 7410 / 17390: Training loss - 4.111236572265625\n","Iteration 7420 / 17390: Training loss - 4.476561546325684\n","Iteration 7430 / 17390: Training loss - 4.240701675415039\n","Iteration 7440 / 17390: Training loss - 4.386781692504883\n","Iteration 7450 / 17390: Training loss - 3.6554300785064697\n","Evaluation loss: 4.90106\n","Iteration 7460 / 17390: Training loss - 4.1374831199646\n","Iteration 7470 / 17390: Training loss - 4.264580249786377\n","Iteration 7480 / 17390: Training loss - 3.7526302337646484\n","Iteration 7490 / 17390: Training loss - 4.4741339683532715\n","Iteration 7500 / 17390: Training loss - 4.260255336761475\n","Evaluation loss: 4.9410844\n","Iteration 7510 / 17390: Training loss - 4.03109884262085\n","Iteration 7520 / 17390: Training loss - 3.6361377239227295\n","Iteration 7530 / 17390: Training loss - 4.462881088256836\n","Iteration 7540 / 17390: Training loss - 3.990722179412842\n","Iteration 7550 / 17390: Training loss - 3.8113152980804443\n","Evaluation loss: 4.97811\n","Iteration 7560 / 17390: Training loss - 3.8031413555145264\n","Iteration 7570 / 17390: Training loss - 3.438600778579712\n","Iteration 7580 / 17390: Training loss - 4.340465545654297\n","Iteration 7590 / 17390: Training loss - 4.110655307769775\n","Iteration 7600 / 17390: Training loss - 3.670128583908081\n","Evaluation loss: 5.2008896\n","Iteration 7610 / 17390: Training loss - 3.968550682067871\n","Iteration 7620 / 17390: Training loss - 3.5670177936553955\n","Iteration 7630 / 17390: Training loss - 4.23529052734375\n","Iteration 7640 / 17390: Training loss - 4.055710315704346\n","Iteration 7650 / 17390: Training loss - 3.9766759872436523\n","Evaluation loss: 5.019658\n","Iteration 7660 / 17390: Training loss - 4.34740686416626\n","Iteration 7670 / 17390: Training loss - 4.1261467933654785\n","Epoch: 2\n","Iteration 7680 / 17390: Training loss - 3.6981794834136963\n","Iteration 7690 / 17390: Training loss - 4.032618999481201\n","Iteration 7700 / 17390: Training loss - 3.9555437564849854\n","Evaluation loss: 4.875604\n","Iteration 7710 / 17390: Training loss - 4.156682968139648\n","Iteration 7720 / 17390: Training loss - 4.2012481689453125\n","Iteration 7730 / 17390: Training loss - 3.8644306659698486\n","Iteration 7740 / 17390: Training loss - 4.253974914550781\n","Iteration 7750 / 17390: Training loss - 4.344287395477295\n","Evaluation loss: 5.1343865\n","Iteration 7760 / 17390: Training loss - 4.091521263122559\n","Iteration 7770 / 17390: Training loss - 4.102182388305664\n","Iteration 7780 / 17390: Training loss - 3.6830949783325195\n","Iteration 7790 / 17390: Training loss - 3.78967547416687\n","Iteration 7800 / 17390: Training loss - 3.6935949325561523\n","Evaluation loss: 4.487245\n","Iteration 7810 / 17390: Training loss - 3.9055135250091553\n","Iteration 7820 / 17390: Training loss - 3.6598474979400635\n","Iteration 7830 / 17390: Training loss - 4.194036960601807\n","Iteration 7840 / 17390: Training loss - 4.064937591552734\n","Iteration 7850 / 17390: Training loss - 4.267115116119385\n","Evaluation loss: 5.073651\n","Iteration 7860 / 17390: Training loss - 4.153618335723877\n","Iteration 7870 / 17390: Training loss - 3.8997602462768555\n","Iteration 7880 / 17390: Training loss - 4.219396591186523\n","Iteration 7890 / 17390: Training loss - 3.9686641693115234\n","Iteration 7900 / 17390: Training loss - 4.514742851257324\n","Evaluation loss: 4.6567845\n","Iteration 7910 / 17390: Training loss - 4.252290725708008\n","Iteration 7920 / 17390: Training loss - 4.33362340927124\n","Iteration 7930 / 17390: Training loss - 3.418301820755005\n","Iteration 7940 / 17390: Training loss - 4.0018086433410645\n","Iteration 7950 / 17390: Training loss - 3.812756061553955\n","Evaluation loss: 5.1049895\n","Iteration 7960 / 17390: Training loss - 3.9974992275238037\n","Iteration 7970 / 17390: Training loss - 3.859297752380371\n","Iteration 7980 / 17390: Training loss - 3.766913890838623\n","Iteration 7990 / 17390: Training loss - 4.024269104003906\n","Iteration 8000 / 17390: Training loss - 4.055183410644531\n","Evaluation loss: 4.919616\n","Iteration 8010 / 17390: Training loss - 3.7139244079589844\n","Iteration 8020 / 17390: Training loss - 3.817310094833374\n","Iteration 8030 / 17390: Training loss - 3.981440782546997\n","Iteration 8040 / 17390: Training loss - 3.868903875350952\n","Iteration 8050 / 17390: Training loss - 3.740598201751709\n","Evaluation loss: 5.3171926\n","Iteration 8060 / 17390: Training loss - 3.882054090499878\n","Iteration 8070 / 17390: Training loss - 4.026693820953369\n","Iteration 8080 / 17390: Training loss - 4.0829243659973145\n","Iteration 8090 / 17390: Training loss - 3.6863861083984375\n","Iteration 8100 / 17390: Training loss - 4.218307018280029\n","Evaluation loss: 4.7449465\n","Iteration 8110 / 17390: Training loss - 3.827387809753418\n","Iteration 8120 / 17390: Training loss - 3.3712475299835205\n","Iteration 8130 / 17390: Training loss - 3.7970006465911865\n","Iteration 8140 / 17390: Training loss - 3.820383310317993\n","Iteration 8150 / 17390: Training loss - 3.7451350688934326\n","Evaluation loss: 4.5264034\n","Iteration 8160 / 17390: Training loss - 4.000439167022705\n","Iteration 8170 / 17390: Training loss - 4.451685428619385\n","Iteration 8180 / 17390: Training loss - 3.7588162422180176\n","Iteration 8190 / 17390: Training loss - 3.650580406188965\n","Iteration 8200 / 17390: Training loss - 4.039997100830078\n","Evaluation loss: 5.282469\n","Iteration 8210 / 17390: Training loss - 4.65266227722168\n","Iteration 8220 / 17390: Training loss - 3.755403518676758\n","Iteration 8230 / 17390: Training loss - 4.05975866317749\n","Iteration 8240 / 17390: Training loss - 4.1637468338012695\n","Iteration 8250 / 17390: Training loss - 3.9007201194763184\n","Evaluation loss: 5.79487\n","Iteration 8260 / 17390: Training loss - 3.914677143096924\n","Iteration 8270 / 17390: Training loss - 3.9667513370513916\n","Iteration 8280 / 17390: Training loss - 3.373015880584717\n","Iteration 8290 / 17390: Training loss - 4.147475719451904\n","Iteration 8300 / 17390: Training loss - 3.890704393386841\n","Evaluation loss: 4.894937\n","Iteration 8310 / 17390: Training loss - 3.795409917831421\n","Iteration 8320 / 17390: Training loss - 3.606304168701172\n","Iteration 8330 / 17390: Training loss - 3.7314212322235107\n","Iteration 8340 / 17390: Training loss - 4.055309295654297\n","Iteration 8350 / 17390: Training loss - 3.8771581649780273\n","Evaluation loss: 4.4085855\n","Iteration 8360 / 17390: Training loss - 3.762103319168091\n","Iteration 8370 / 17390: Training loss - 4.374059200286865\n","Iteration 8380 / 17390: Training loss - 4.506633281707764\n","Iteration 8390 / 17390: Training loss - 3.6156601905822754\n","Iteration 8400 / 17390: Training loss - 4.275811195373535\n","Evaluation loss: 4.668943\n","Iteration 8410 / 17390: Training loss - 3.128438711166382\n","Iteration 8420 / 17390: Training loss - 3.5438380241394043\n","Iteration 8430 / 17390: Training loss - 4.012973785400391\n","Iteration 8440 / 17390: Training loss - 3.8089599609375\n","Iteration 8450 / 17390: Training loss - 3.986325979232788\n","Evaluation loss: 4.8353877\n","Iteration 8460 / 17390: Training loss - 3.9901351928710938\n","Iteration 8470 / 17390: Training loss - 4.098666191101074\n","Iteration 8480 / 17390: Training loss - 4.557943820953369\n","Iteration 8490 / 17390: Training loss - 3.948892593383789\n","Iteration 8500 / 17390: Training loss - 4.401364326477051\n","Evaluation loss: 4.951353\n","Iteration 8510 / 17390: Training loss - 4.032670974731445\n","Iteration 8520 / 17390: Training loss - 4.12148380279541\n","Iteration 8530 / 17390: Training loss - 3.642528533935547\n","Iteration 8540 / 17390: Training loss - 3.761672258377075\n","Iteration 8550 / 17390: Training loss - 4.142852783203125\n","Evaluation loss: 4.733473\n","Iteration 8560 / 17390: Training loss - 3.7631397247314453\n","Iteration 8570 / 17390: Training loss - 3.7737131118774414\n","Iteration 8580 / 17390: Training loss - 3.6303396224975586\n","Iteration 8590 / 17390: Training loss - 3.7061142921447754\n","Iteration 8600 / 17390: Training loss - 3.532102346420288\n","Evaluation loss: 5.1929007\n","Iteration 8610 / 17390: Training loss - 3.8982815742492676\n","Iteration 8620 / 17390: Training loss - 4.166408061981201\n","Iteration 8630 / 17390: Training loss - 3.8301827907562256\n","Iteration 8640 / 17390: Training loss - 3.787903070449829\n","Iteration 8650 / 17390: Training loss - 3.3011929988861084\n","Evaluation loss: 4.952182\n","Iteration 8660 / 17390: Training loss - 4.194302082061768\n","Iteration 8670 / 17390: Training loss - 3.8796262741088867\n","Iteration 8680 / 17390: Training loss - 4.082573890686035\n","Iteration 8690 / 17390: Training loss - 3.8704817295074463\n","Iteration 8700 / 17390: Training loss - 3.4154202938079834\n","Evaluation loss: 4.976448\n","Iteration 8710 / 17390: Training loss - 4.086291790008545\n","Iteration 8720 / 17390: Training loss - 3.533745050430298\n","Iteration 8730 / 17390: Training loss - 3.9662680625915527\n","Iteration 8740 / 17390: Training loss - 4.071219444274902\n","Iteration 8750 / 17390: Training loss - 3.864570140838623\n","Evaluation loss: 5.027122\n","Epoch: 3\n","Iteration 8760 / 17390: Training loss - 3.753291368484497\n","Iteration 8770 / 17390: Training loss - 3.671560287475586\n","Iteration 8780 / 17390: Training loss - 3.6651148796081543\n","Iteration 8790 / 17390: Training loss - 4.263564109802246\n","Iteration 8800 / 17390: Training loss - 3.6455655097961426\n","Evaluation loss: 5.0406075\n","Iteration 8810 / 17390: Training loss - 3.8932056427001953\n","Iteration 8820 / 17390: Training loss - 3.601039171218872\n","Iteration 8830 / 17390: Training loss - 3.9993557929992676\n","Iteration 8840 / 17390: Training loss - 4.077028274536133\n","Iteration 8850 / 17390: Training loss - 4.228513240814209\n","Evaluation loss: 4.383585\n","Iteration 8860 / 17390: Training loss - 3.790966272354126\n","Iteration 8870 / 17390: Training loss - 4.265119552612305\n","Iteration 8880 / 17390: Training loss - 3.9976823329925537\n","Iteration 8890 / 17390: Training loss - 4.107676029205322\n","Iteration 8900 / 17390: Training loss - 4.140815258026123\n","Evaluation loss: 4.953264\n","Iteration 8910 / 17390: Training loss - 3.512058734893799\n","Iteration 8920 / 17390: Training loss - 3.6501965522766113\n","Iteration 8930 / 17390: Training loss - 4.297004222869873\n","Iteration 8940 / 17390: Training loss - 3.978513240814209\n","Iteration 8950 / 17390: Training loss - 3.6433265209198\n","Evaluation loss: 5.6125445\n","Iteration 8960 / 17390: Training loss - 4.0386128425598145\n","Iteration 8970 / 17390: Training loss - 4.258336067199707\n","Iteration 8980 / 17390: Training loss - 3.7285354137420654\n","Iteration 8990 / 17390: Training loss - 3.882011651992798\n","Iteration 9000 / 17390: Training loss - 4.0737223625183105\n","Evaluation loss: 5.3314233\n","Iteration 9010 / 17390: Training loss - 3.775965452194214\n","Iteration 9020 / 17390: Training loss - 3.8620872497558594\n","Iteration 9030 / 17390: Training loss - 3.818528890609741\n","Iteration 9040 / 17390: Training loss - 4.186148643493652\n","Iteration 9050 / 17390: Training loss - 4.418323040008545\n","Evaluation loss: 5.7443633\n","Iteration 9060 / 17390: Training loss - 3.879992961883545\n","Iteration 9070 / 17390: Training loss - 3.351916790008545\n","Iteration 9080 / 17390: Training loss - 4.642425537109375\n","Iteration 9090 / 17390: Training loss - 3.821633815765381\n","Iteration 9100 / 17390: Training loss - 3.791231155395508\n","Evaluation loss: 4.8258977\n","Iteration 9110 / 17390: Training loss - 4.025415897369385\n","Iteration 9120 / 17390: Training loss - 3.721656084060669\n","Iteration 9130 / 17390: Training loss - 4.530725479125977\n","Iteration 9140 / 17390: Training loss - 4.176238536834717\n","Iteration 9150 / 17390: Training loss - 4.338430404663086\n","Evaluation loss: 4.5297365\n","Iteration 9160 / 17390: Training loss - 3.77577805519104\n","Iteration 9170 / 17390: Training loss - 3.7628250122070312\n","Iteration 9180 / 17390: Training loss - 3.9108970165252686\n","Iteration 9190 / 17390: Training loss - 4.166551113128662\n","Iteration 9200 / 17390: Training loss - 4.099533557891846\n","Evaluation loss: 5.04282\n","Iteration 9210 / 17390: Training loss - 3.445835590362549\n","Iteration 9220 / 17390: Training loss - 4.586597442626953\n","Iteration 9230 / 17390: Training loss - 4.28306770324707\n","Iteration 9240 / 17390: Training loss - 3.6353342533111572\n","Iteration 9250 / 17390: Training loss - 4.418817520141602\n","Evaluation loss: 5.039524\n","Iteration 9260 / 17390: Training loss - 4.223890781402588\n","Iteration 9270 / 17390: Training loss - 3.867918014526367\n","Iteration 9280 / 17390: Training loss - 3.1346192359924316\n","Iteration 9290 / 17390: Training loss - 4.033656120300293\n","Iteration 9300 / 17390: Training loss - 4.038361549377441\n","Evaluation loss: 4.720341\n","Iteration 9310 / 17390: Training loss - 3.6689395904541016\n","Iteration 9320 / 17390: Training loss - 4.065887451171875\n","Iteration 9330 / 17390: Training loss - 3.8274972438812256\n","Iteration 9340 / 17390: Training loss - 3.3653337955474854\n","Iteration 9350 / 17390: Training loss - 3.911900520324707\n","Evaluation loss: 4.474083\n","Iteration 9360 / 17390: Training loss - 3.9083847999572754\n","Iteration 9370 / 17390: Training loss - 3.783770799636841\n","Iteration 9380 / 17390: Training loss - 4.126155376434326\n","Iteration 9390 / 17390: Training loss - 3.191678762435913\n","Iteration 9400 / 17390: Training loss - 3.876403570175171\n","Evaluation loss: 5.7434344\n","Iteration 9410 / 17390: Training loss - 3.6502373218536377\n","Iteration 9420 / 17390: Training loss - 4.132349491119385\n","Iteration 9430 / 17390: Training loss - 3.648599624633789\n","Iteration 9440 / 17390: Training loss - 3.7839183807373047\n","Iteration 9450 / 17390: Training loss - 3.6744039058685303\n","Evaluation loss: 4.566737\n","Iteration 9460 / 17390: Training loss - 3.838521957397461\n","Iteration 9470 / 17390: Training loss - 3.3939759731292725\n","Iteration 9480 / 17390: Training loss - 4.03765869140625\n","Iteration 9490 / 17390: Training loss - 3.913360595703125\n","Iteration 9500 / 17390: Training loss - 3.7057230472564697\n","Evaluation loss: 5.456215\n","Iteration 9510 / 17390: Training loss - 3.7820093631744385\n","Iteration 9520 / 17390: Training loss - 3.753481149673462\n","Iteration 9530 / 17390: Training loss - 3.918088912963867\n","Iteration 9540 / 17390: Training loss - 3.7417056560516357\n","Iteration 9550 / 17390: Training loss - 3.9754531383514404\n","Evaluation loss: 5.1978836\n","Iteration 9560 / 17390: Training loss - 4.013535022735596\n","Iteration 9570 / 17390: Training loss - 3.7354238033294678\n","Iteration 9580 / 17390: Training loss - 3.8203368186950684\n","Iteration 9590 / 17390: Training loss - 3.4347846508026123\n","Iteration 9600 / 17390: Training loss - 4.0990777015686035\n","Evaluation loss: 5.2233744\n","Iteration 9610 / 17390: Training loss - 3.8413994312286377\n","Iteration 9620 / 17390: Training loss - 3.582540273666382\n","Iteration 9630 / 17390: Training loss - 3.3743736743927\n","Iteration 9640 / 17390: Training loss - 3.0796282291412354\n","Iteration 9650 / 17390: Training loss - 3.3187601566314697\n","Evaluation loss: 4.847568\n","Iteration 9660 / 17390: Training loss - 3.934833288192749\n","Iteration 9670 / 17390: Training loss - 3.5102193355560303\n","Iteration 9680 / 17390: Training loss - 3.6037392616271973\n","Iteration 9690 / 17390: Training loss - 3.685150384902954\n","Iteration 9700 / 17390: Training loss - 3.8776280879974365\n","Evaluation loss: 5.6098537\n","Iteration 9710 / 17390: Training loss - 4.200991630554199\n","Iteration 9720 / 17390: Training loss - 3.3817505836486816\n","Iteration 9730 / 17390: Training loss - 3.493267297744751\n","Iteration 9740 / 17390: Training loss - 3.1950571537017822\n","Iteration 9750 / 17390: Training loss - 3.541729211807251\n","Evaluation loss: 4.9959545\n","Iteration 9760 / 17390: Training loss - 3.3188867568969727\n","Iteration 9770 / 17390: Training loss - 4.209226131439209\n","Iteration 9780 / 17390: Training loss - 3.305680751800537\n","Iteration 9790 / 17390: Training loss - 3.7368624210357666\n","Iteration 9800 / 17390: Training loss - 3.713965654373169\n","Evaluation loss: 4.4075537\n","Iteration 9810 / 17390: Training loss - 4.200976848602295\n","Iteration 9820 / 17390: Training loss - 3.1947948932647705\n","Iteration 9830 / 17390: Training loss - 3.833315849304199\n","Epoch: 4\n","Iteration 9840 / 17390: Training loss - 3.657672643661499\n","Iteration 9850 / 17390: Training loss - 4.456517696380615\n","Evaluation loss: 5.135185\n","Iteration 9860 / 17390: Training loss - 3.0098931789398193\n","Iteration 9870 / 17390: Training loss - 3.5309081077575684\n","Iteration 9880 / 17390: Training loss - 4.038206100463867\n","Iteration 9890 / 17390: Training loss - 3.626422643661499\n","Iteration 9900 / 17390: Training loss - 3.9050137996673584\n","Evaluation loss: 5.205984\n","Iteration 9910 / 17390: Training loss - 3.5657968521118164\n","Iteration 9920 / 17390: Training loss - 3.5444512367248535\n","Iteration 9930 / 17390: Training loss - 3.613416910171509\n","Iteration 9940 / 17390: Training loss - 3.7297728061676025\n","Iteration 9950 / 17390: Training loss - 3.5077133178710938\n","Evaluation loss: 5.1543503\n","Iteration 9960 / 17390: Training loss - 3.82550048828125\n","Iteration 9970 / 17390: Training loss - 3.5338685512542725\n","Iteration 9980 / 17390: Training loss - 3.751638174057007\n","Iteration 9990 / 17390: Training loss - 3.774566650390625\n","Iteration 10000 / 17390: Training loss - 3.9289357662200928\n","Evaluation loss: 5.296621\n","Iteration 10010 / 17390: Training loss - 3.6305994987487793\n","Iteration 10020 / 17390: Training loss - 3.660038709640503\n","Iteration 10030 / 17390: Training loss - 3.8173515796661377\n","Iteration 10040 / 17390: Training loss - 3.7198619842529297\n","Iteration 10050 / 17390: Training loss - 3.1595938205718994\n","Evaluation loss: 5.257894\n","Iteration 10060 / 17390: Training loss - 3.8215298652648926\n","Iteration 10070 / 17390: Training loss - 3.699922800064087\n","Iteration 10080 / 17390: Training loss - 3.233811616897583\n","Iteration 10090 / 17390: Training loss - 3.600266933441162\n","Iteration 10100 / 17390: Training loss - 3.732126474380493\n","Evaluation loss: 5.6756644\n","Iteration 10110 / 17390: Training loss - 3.846613883972168\n","Iteration 10120 / 17390: Training loss - 3.524522066116333\n","Iteration 10130 / 17390: Training loss - 3.881929397583008\n","Iteration 10140 / 17390: Training loss - 3.6914520263671875\n","Iteration 10150 / 17390: Training loss - 3.678280830383301\n","Evaluation loss: 4.587881\n","Iteration 10160 / 17390: Training loss - 4.277448654174805\n","Iteration 10170 / 17390: Training loss - 4.2392964363098145\n","Iteration 10180 / 17390: Training loss - 3.791825294494629\n","Iteration 10190 / 17390: Training loss - 3.592054605484009\n","Iteration 10200 / 17390: Training loss - 4.0541205406188965\n","Evaluation loss: 4.8056483\n","Iteration 10210 / 17390: Training loss - 3.9606165885925293\n","Iteration 10220 / 17390: Training loss - 3.4917540550231934\n","Iteration 10230 / 17390: Training loss - 3.7055253982543945\n","Iteration 10240 / 17390: Training loss - 3.305417537689209\n","Iteration 10250 / 17390: Training loss - 4.474937438964844\n","Evaluation loss: 5.1237645\n","Iteration 10260 / 17390: Training loss - 3.5876235961914062\n","Iteration 10270 / 17390: Training loss - 3.550506114959717\n","Iteration 10280 / 17390: Training loss - 3.567370653152466\n","Iteration 10290 / 17390: Training loss - 3.7127771377563477\n","Iteration 10300 / 17390: Training loss - 3.901185989379883\n","Evaluation loss: 5.3024273\n","Iteration 10310 / 17390: Training loss - 3.046501398086548\n","Iteration 10320 / 17390: Training loss - 3.1086432933807373\n","Iteration 10330 / 17390: Training loss - 3.802340030670166\n","Iteration 10340 / 17390: Training loss - 3.8839893341064453\n","Iteration 10350 / 17390: Training loss - 3.670597553253174\n","Evaluation loss: 5.199759\n","Iteration 10360 / 17390: Training loss - 3.230776309967041\n","Iteration 10370 / 17390: Training loss - 3.4863638877868652\n","Iteration 10380 / 17390: Training loss - 3.5458457469940186\n","Iteration 10390 / 17390: Training loss - 4.121236324310303\n","Iteration 10400 / 17390: Training loss - 3.572679281234741\n","Evaluation loss: 5.3382807\n","Iteration 10410 / 17390: Training loss - 3.9945452213287354\n","Iteration 10420 / 17390: Training loss - 3.4241485595703125\n","Iteration 10430 / 17390: Training loss - 3.3508076667785645\n","Iteration 10440 / 17390: Training loss - 3.3196017742156982\n","Iteration 10450 / 17390: Training loss - 3.300856351852417\n","Evaluation loss: 4.8884506\n","Iteration 10460 / 17390: Training loss - 4.125894546508789\n","Iteration 10470 / 17390: Training loss - 3.908944845199585\n","Iteration 10480 / 17390: Training loss - 4.065185546875\n","Iteration 10490 / 17390: Training loss - 3.8363120555877686\n","Iteration 10500 / 17390: Training loss - 3.627164602279663\n","Evaluation loss: 4.994597\n","Iteration 10510 / 17390: Training loss - 3.7292118072509766\n","Iteration 10520 / 17390: Training loss - 3.378722906112671\n","Iteration 10530 / 17390: Training loss - 3.8365626335144043\n","Iteration 10540 / 17390: Training loss - 3.633920907974243\n","Iteration 10550 / 17390: Training loss - 3.8456878662109375\n","Evaluation loss: 5.480506\n","Iteration 10560 / 17390: Training loss - 4.2091474533081055\n","Iteration 10570 / 17390: Training loss - 3.662768840789795\n","Iteration 10580 / 17390: Training loss - 3.5894088745117188\n","Iteration 10590 / 17390: Training loss - 3.5403571128845215\n","Iteration 10600 / 17390: Training loss - 3.1837809085845947\n","Evaluation loss: 4.963902\n","Iteration 10610 / 17390: Training loss - 3.1380720138549805\n","Iteration 10620 / 17390: Training loss - 3.652111530303955\n","Iteration 10630 / 17390: Training loss - 4.153295040130615\n","Iteration 10640 / 17390: Training loss - 3.1028785705566406\n","Iteration 10650 / 17390: Training loss - 3.18855357170105\n","Evaluation loss: 4.6603436\n","Iteration 10660 / 17390: Training loss - 4.134255409240723\n","Iteration 10670 / 17390: Training loss - 3.573047637939453\n","Iteration 10680 / 17390: Training loss - 3.7574830055236816\n","Iteration 10690 / 17390: Training loss - 4.352236747741699\n","Iteration 10700 / 17390: Training loss - 3.5654377937316895\n","Evaluation loss: 4.981101\n","Iteration 10710 / 17390: Training loss - 3.69291615486145\n","Iteration 10720 / 17390: Training loss - 3.473876714706421\n","Iteration 10730 / 17390: Training loss - 3.5688047409057617\n","Iteration 10740 / 17390: Training loss - 4.119623184204102\n","Iteration 10750 / 17390: Training loss - 3.3125972747802734\n","Evaluation loss: 5.145262\n","Iteration 10760 / 17390: Training loss - 3.3341422080993652\n","Iteration 10770 / 17390: Training loss - 3.322559356689453\n","Iteration 10780 / 17390: Training loss - 3.450721263885498\n","Iteration 10790 / 17390: Training loss - 3.876356363296509\n","Iteration 10800 / 17390: Training loss - 3.49718976020813\n","Evaluation loss: 5.365985\n","Iteration 10810 / 17390: Training loss - 3.82828688621521\n","Iteration 10820 / 17390: Training loss - 3.8592565059661865\n","Iteration 10830 / 17390: Training loss - 3.930072069168091\n","Iteration 10840 / 17390: Training loss - 3.965668201446533\n","Iteration 10850 / 17390: Training loss - 3.6529829502105713\n","Evaluation loss: 5.549068\n","Iteration 10860 / 17390: Training loss - 2.976322889328003\n","Iteration 10870 / 17390: Training loss - 3.743023633956909\n","Iteration 10880 / 17390: Training loss - 3.7152884006500244\n","Iteration 10890 / 17390: Training loss - 4.095010280609131\n","Iteration 10900 / 17390: Training loss - 3.3760416507720947\n","Evaluation loss: 5.0318255\n","Iteration 10910 / 17390: Training loss - 3.135364532470703\n","Epoch: 5\n","Iteration 10920 / 17390: Training loss - 4.013958930969238\n","Iteration 10930 / 17390: Training loss - 4.073272228240967\n","Iteration 10940 / 17390: Training loss - 3.8863046169281006\n","Iteration 10950 / 17390: Training loss - 3.7722623348236084\n","Evaluation loss: 4.6611824\n","Iteration 10960 / 17390: Training loss - 3.0161244869232178\n","Iteration 10970 / 17390: Training loss - 3.438187837600708\n","Iteration 10980 / 17390: Training loss - 3.1796815395355225\n","Iteration 10990 / 17390: Training loss - 3.756406784057617\n","Iteration 11000 / 17390: Training loss - 3.946852684020996\n","Evaluation loss: 5.391489\n","Iteration 11010 / 17390: Training loss - 3.270202875137329\n","Iteration 11020 / 17390: Training loss - 3.5106794834136963\n","Iteration 11030 / 17390: Training loss - 3.890383720397949\n","Iteration 11040 / 17390: Training loss - 3.689530372619629\n","Iteration 11050 / 17390: Training loss - 3.920872688293457\n","Evaluation loss: 5.0892096\n","Iteration 11060 / 17390: Training loss - 3.867825508117676\n","Iteration 11070 / 17390: Training loss - 3.5334928035736084\n","Iteration 11080 / 17390: Training loss - 3.421793222427368\n","Iteration 11090 / 17390: Training loss - 3.834505796432495\n","Iteration 11100 / 17390: Training loss - 3.478151559829712\n","Evaluation loss: 4.6291423\n","Iteration 11110 / 17390: Training loss - 3.865335464477539\n","Iteration 11120 / 17390: Training loss - 3.725109100341797\n","Iteration 11130 / 17390: Training loss - 3.434401035308838\n","Iteration 11140 / 17390: Training loss - 3.521792411804199\n","Iteration 11150 / 17390: Training loss - 3.7509684562683105\n","Evaluation loss: 4.8510785\n","Iteration 11160 / 17390: Training loss - 3.2488229274749756\n","Iteration 11170 / 17390: Training loss - 3.0524823665618896\n","Iteration 11180 / 17390: Training loss - 3.332141160964966\n","Iteration 11190 / 17390: Training loss - 3.537008285522461\n","Iteration 11200 / 17390: Training loss - 2.950953245162964\n","Evaluation loss: 4.7062526\n","Iteration 11210 / 17390: Training loss - 3.6774661540985107\n","Iteration 11220 / 17390: Training loss - 3.3754518032073975\n","Iteration 11230 / 17390: Training loss - 3.640977621078491\n","Iteration 11240 / 17390: Training loss - 3.293947696685791\n","Iteration 11250 / 17390: Training loss - 4.0567145347595215\n","Evaluation loss: 5.2335873\n","Iteration 11260 / 17390: Training loss - 3.5778732299804688\n","Iteration 11270 / 17390: Training loss - 3.38826847076416\n","Iteration 11280 / 17390: Training loss - 3.07327938079834\n","Iteration 11290 / 17390: Training loss - 3.45320463180542\n","Iteration 11300 / 17390: Training loss - 3.233774423599243\n","Evaluation loss: 4.943617\n","Iteration 11310 / 17390: Training loss - 3.4080851078033447\n","Iteration 11320 / 17390: Training loss - 3.5736865997314453\n","Iteration 11330 / 17390: Training loss - 3.1115376949310303\n","Iteration 11340 / 17390: Training loss - 3.0630884170532227\n","Iteration 11350 / 17390: Training loss - 3.562082529067993\n","Evaluation loss: 5.15518\n","Iteration 11360 / 17390: Training loss - 3.7115418910980225\n","Iteration 11370 / 17390: Training loss - 3.5171709060668945\n","Iteration 11380 / 17390: Training loss - 3.6401309967041016\n","Iteration 11390 / 17390: Training loss - 3.2418265342712402\n","Iteration 11400 / 17390: Training loss - 3.4251952171325684\n","Evaluation loss: 5.4116297\n","Iteration 11410 / 17390: Training loss - 3.8005876541137695\n","Iteration 11420 / 17390: Training loss - 3.8188767433166504\n","Iteration 11430 / 17390: Training loss - 3.639714002609253\n","Iteration 11440 / 17390: Training loss - 2.8102588653564453\n","Iteration 11450 / 17390: Training loss - 3.38173246383667\n","Evaluation loss: 4.6329017\n","Iteration 11460 / 17390: Training loss - 3.7276368141174316\n","Iteration 11470 / 17390: Training loss - 3.4847021102905273\n","Iteration 11480 / 17390: Training loss - 3.4345247745513916\n","Iteration 11490 / 17390: Training loss - 3.059468984603882\n","Iteration 11500 / 17390: Training loss - 3.7126684188842773\n","Evaluation loss: 4.7359767\n","Iteration 11510 / 17390: Training loss - 3.7524375915527344\n","Iteration 11520 / 17390: Training loss - 3.726445198059082\n","Iteration 11530 / 17390: Training loss - 2.9388389587402344\n","Iteration 11540 / 17390: Training loss - 4.119894504547119\n","Iteration 11550 / 17390: Training loss - 3.2656614780426025\n","Evaluation loss: 4.779849\n","Iteration 11560 / 17390: Training loss - 3.1620771884918213\n","Iteration 11570 / 17390: Training loss - 3.3819472789764404\n","Iteration 11580 / 17390: Training loss - 3.1570560932159424\n","Iteration 11590 / 17390: Training loss - 3.870380401611328\n","Iteration 11600 / 17390: Training loss - 3.4426045417785645\n","Evaluation loss: 4.843679\n","Iteration 11610 / 17390: Training loss - 3.558039665222168\n","Iteration 11620 / 17390: Training loss - 3.8012020587921143\n","Iteration 11630 / 17390: Training loss - 3.7086222171783447\n","Iteration 11640 / 17390: Training loss - 3.4312548637390137\n","Iteration 11650 / 17390: Training loss - 3.3297369480133057\n","Evaluation loss: 4.918646\n","Iteration 11660 / 17390: Training loss - 3.8393092155456543\n","Iteration 11670 / 17390: Training loss - 3.5225183963775635\n","Iteration 11680 / 17390: Training loss - 3.243131160736084\n","Iteration 11690 / 17390: Training loss - 3.538648843765259\n","Iteration 11700 / 17390: Training loss - 3.4818220138549805\n","Evaluation loss: 5.1691294\n","Iteration 11710 / 17390: Training loss - 3.8413023948669434\n","Iteration 11720 / 17390: Training loss - 4.017261505126953\n","Iteration 11730 / 17390: Training loss - 3.8296947479248047\n","Iteration 11740 / 17390: Training loss - 3.459364175796509\n","Iteration 11750 / 17390: Training loss - 3.6579818725585938\n","Evaluation loss: 4.631519\n","Iteration 11760 / 17390: Training loss - 3.4309420585632324\n","Iteration 11770 / 17390: Training loss - 3.4011270999908447\n","Iteration 11780 / 17390: Training loss - 3.543933153152466\n","Iteration 11790 / 17390: Training loss - 3.443023204803467\n","Iteration 11800 / 17390: Training loss - 3.8197357654571533\n","Evaluation loss: 5.1966863\n","Iteration 11810 / 17390: Training loss - 3.353003740310669\n","Iteration 11820 / 17390: Training loss - 3.329120397567749\n","Iteration 11830 / 17390: Training loss - 3.696887254714966\n","Iteration 11840 / 17390: Training loss - 3.236801862716675\n","Iteration 11850 / 17390: Training loss - 3.212043523788452\n","Evaluation loss: 4.897563\n","Iteration 11860 / 17390: Training loss - 3.648279905319214\n","Iteration 11870 / 17390: Training loss - 3.7871217727661133\n","Iteration 11880 / 17390: Training loss - 3.350435256958008\n","Iteration 11890 / 17390: Training loss - 3.4965715408325195\n","Iteration 11900 / 17390: Training loss - 3.215207576751709\n","Evaluation loss: 4.5294976\n","Iteration 11910 / 17390: Training loss - 2.901942729949951\n","Iteration 11920 / 17390: Training loss - 3.001706838607788\n","Iteration 11930 / 17390: Training loss - 3.9544458389282227\n","Iteration 11940 / 17390: Training loss - 2.998037338256836\n","Iteration 11950 / 17390: Training loss - 3.003793239593506\n","Evaluation loss: 4.767143\n","Iteration 11960 / 17390: Training loss - 3.8729658126831055\n","Iteration 11970 / 17390: Training loss - 3.102617025375366\n","Iteration 11980 / 17390: Training loss - 3.862532377243042\n","Iteration 11990 / 17390: Training loss - 3.5060176849365234\n","Epoch: 6\n","Iteration 12000 / 17390: Training loss - 3.9504008293151855\n","Evaluation loss: 5.138483\n","Iteration 12010 / 17390: Training loss - 3.0859365463256836\n","Iteration 12020 / 17390: Training loss - 3.394529342651367\n","Iteration 12030 / 17390: Training loss - 3.4355812072753906\n","Iteration 12040 / 17390: Training loss - 3.5523056983947754\n","Iteration 12050 / 17390: Training loss - 3.9504029750823975\n","Evaluation loss: 4.948091\n","Iteration 12060 / 17390: Training loss - 3.341853380203247\n","Iteration 12070 / 17390: Training loss - 3.37807559967041\n","Iteration 12080 / 17390: Training loss - 3.5234272480010986\n","Iteration 12090 / 17390: Training loss - 3.6544716358184814\n","Iteration 12100 / 17390: Training loss - 3.751368999481201\n","Evaluation loss: 5.276953\n","Iteration 12110 / 17390: Training loss - 3.228098154067993\n","Iteration 12120 / 17390: Training loss - 3.905449151992798\n","Iteration 12130 / 17390: Training loss - 3.480152130126953\n","Iteration 12140 / 17390: Training loss - 3.6049888134002686\n","Iteration 12150 / 17390: Training loss - 3.6163036823272705\n","Evaluation loss: 5.261781\n","Iteration 12160 / 17390: Training loss - 2.780017137527466\n","Iteration 12170 / 17390: Training loss - 3.6957623958587646\n","Iteration 12180 / 17390: Training loss - 3.295351266860962\n","Iteration 12190 / 17390: Training loss - 3.490335464477539\n","Iteration 12200 / 17390: Training loss - 3.6910507678985596\n","Evaluation loss: 5.017096\n","Iteration 12210 / 17390: Training loss - 2.8697509765625\n","Iteration 12220 / 17390: Training loss - 3.1716437339782715\n","Iteration 12230 / 17390: Training loss - 2.920654535293579\n","Iteration 12240 / 17390: Training loss - 3.2837796211242676\n","Iteration 12250 / 17390: Training loss - 2.9528636932373047\n","Evaluation loss: 5.765223\n","Iteration 12260 / 17390: Training loss - 3.080212354660034\n","Iteration 12270 / 17390: Training loss - 3.8357431888580322\n","Iteration 12280 / 17390: Training loss - 3.3015220165252686\n","Iteration 12290 / 17390: Training loss - 3.2691454887390137\n","Iteration 12300 / 17390: Training loss - 3.4549903869628906\n","Evaluation loss: 4.3062496\n","Iteration 12310 / 17390: Training loss - 3.4237618446350098\n","Iteration 12320 / 17390: Training loss - 3.6854419708251953\n","Iteration 12330 / 17390: Training loss - 3.0305919647216797\n","Iteration 12340 / 17390: Training loss - 3.5733609199523926\n","Iteration 12350 / 17390: Training loss - 3.753690242767334\n","Evaluation loss: 5.2860994\n","Iteration 12360 / 17390: Training loss - 3.4767162799835205\n","Iteration 12370 / 17390: Training loss - 3.139893054962158\n","Iteration 12380 / 17390: Training loss - 3.469235897064209\n","Iteration 12390 / 17390: Training loss - 3.2038686275482178\n","Iteration 12400 / 17390: Training loss - 3.6312856674194336\n","Evaluation loss: 5.1714373\n","Iteration 12410 / 17390: Training loss - 3.459724187850952\n","Iteration 12420 / 17390: Training loss - 3.264491319656372\n","Iteration 12430 / 17390: Training loss - 3.184577703475952\n","Iteration 12440 / 17390: Training loss - 3.460376501083374\n","Iteration 12450 / 17390: Training loss - 3.5667123794555664\n","Evaluation loss: 5.447425\n","Iteration 12460 / 17390: Training loss - 3.525111198425293\n","Iteration 12470 / 17390: Training loss - 3.23990797996521\n","Iteration 12480 / 17390: Training loss - 3.943185329437256\n","Iteration 12490 / 17390: Training loss - 3.079096794128418\n","Iteration 12500 / 17390: Training loss - 2.9873697757720947\n","Evaluation loss: 5.553499\n","Iteration 12510 / 17390: Training loss - 3.3229222297668457\n","Iteration 12520 / 17390: Training loss - 3.452347755432129\n","Iteration 12530 / 17390: Training loss - 3.055729389190674\n","Iteration 12540 / 17390: Training loss - 3.7752437591552734\n","Iteration 12550 / 17390: Training loss - 3.605936050415039\n","Evaluation loss: 4.5139675\n","Iteration 12560 / 17390: Training loss - 3.550668239593506\n","Iteration 12570 / 17390: Training loss - 3.328674554824829\n","Iteration 12580 / 17390: Training loss - 2.8793323040008545\n","Iteration 12590 / 17390: Training loss - 3.1354119777679443\n","Iteration 12600 / 17390: Training loss - 3.0187442302703857\n","Evaluation loss: 4.2798576\n","Iteration 12610 / 17390: Training loss - 3.3755459785461426\n","Iteration 12620 / 17390: Training loss - 3.6187894344329834\n","Iteration 12630 / 17390: Training loss - 3.1354379653930664\n","Iteration 12640 / 17390: Training loss - 3.4476120471954346\n","Iteration 12650 / 17390: Training loss - 3.17860746383667\n","Evaluation loss: 5.3636456\n","Iteration 12660 / 17390: Training loss - 3.8054068088531494\n","Iteration 12670 / 17390: Training loss - 3.0839333534240723\n","Iteration 12680 / 17390: Training loss - 3.637291431427002\n","Iteration 12690 / 17390: Training loss - 2.6754937171936035\n","Iteration 12700 / 17390: Training loss - 3.249539613723755\n","Evaluation loss: 5.6626463\n","Iteration 12710 / 17390: Training loss - 2.66312313079834\n","Iteration 12720 / 17390: Training loss - 3.3069348335266113\n","Iteration 12730 / 17390: Training loss - 3.0265214443206787\n","Iteration 12740 / 17390: Training loss - 3.289398193359375\n","Iteration 12750 / 17390: Training loss - 2.934769868850708\n","Evaluation loss: 5.096901\n","Iteration 12760 / 17390: Training loss - 3.841378927230835\n","Iteration 12770 / 17390: Training loss - 3.303319215774536\n","Iteration 12780 / 17390: Training loss - 3.3591012954711914\n","Iteration 12790 / 17390: Training loss - 3.1475446224212646\n","Iteration 12800 / 17390: Training loss - 3.1291275024414062\n","Evaluation loss: 5.1408386\n","Iteration 12810 / 17390: Training loss - 3.853773355484009\n","Iteration 12820 / 17390: Training loss - 3.1159794330596924\n","Iteration 12830 / 17390: Training loss - 3.608891487121582\n","Iteration 12840 / 17390: Training loss - 3.0367493629455566\n","Iteration 12850 / 17390: Training loss - 3.478783130645752\n","Evaluation loss: 5.4337664\n","Iteration 12860 / 17390: Training loss - 3.3705101013183594\n","Iteration 12870 / 17390: Training loss - 3.268551826477051\n","Iteration 12880 / 17390: Training loss - 3.4742531776428223\n","Iteration 12890 / 17390: Training loss - 3.532168388366699\n","Iteration 12900 / 17390: Training loss - 3.5276525020599365\n","Evaluation loss: 5.2450647\n","Iteration 12910 / 17390: Training loss - 3.1945078372955322\n","Iteration 12920 / 17390: Training loss - 3.621743679046631\n","Iteration 12930 / 17390: Training loss - 3.333604574203491\n","Iteration 12940 / 17390: Training loss - 3.2928149700164795\n","Iteration 12950 / 17390: Training loss - 3.263731002807617\n","Evaluation loss: 5.2740784\n","Iteration 12960 / 17390: Training loss - 2.9918110370635986\n","Iteration 12970 / 17390: Training loss - 3.265357255935669\n","Iteration 12980 / 17390: Training loss - 3.4143292903900146\n","Iteration 12990 / 17390: Training loss - 3.1566097736358643\n","Iteration 13000 / 17390: Training loss - 3.141556739807129\n","Evaluation loss: 5.2615943\n","Iteration 13010 / 17390: Training loss - 3.7888851165771484\n","Iteration 13020 / 17390: Training loss - 3.4220168590545654\n","Iteration 13030 / 17390: Training loss - 3.671233892440796\n","Iteration 13040 / 17390: Training loss - 3.098719596862793\n","Iteration 13050 / 17390: Training loss - 3.4492125511169434\n","Evaluation loss: 5.2378154\n","Iteration 13060 / 17390: Training loss - 3.9539713859558105\n","Iteration 13070 / 17390: Training loss - 2.8245699405670166\n","Epoch: 7\n","Iteration 13080 / 17390: Training loss - 3.364396810531616\n","Iteration 13090 / 17390: Training loss - 3.271357536315918\n","Iteration 13100 / 17390: Training loss - 3.347787618637085\n","Evaluation loss: 4.979747\n","Iteration 13110 / 17390: Training loss - 3.4676687717437744\n","Iteration 13120 / 17390: Training loss - 3.3915598392486572\n","Iteration 13130 / 17390: Training loss - 2.827017307281494\n","Iteration 13140 / 17390: Training loss - 3.2094836235046387\n","Iteration 13150 / 17390: Training loss - 3.2873032093048096\n","Evaluation loss: 5.4898863\n","Iteration 13160 / 17390: Training loss - 3.769212245941162\n","Iteration 13170 / 17390: Training loss - 3.2641491889953613\n","Iteration 13180 / 17390: Training loss - 3.2043309211730957\n","Iteration 13190 / 17390: Training loss - 2.8935954570770264\n","Iteration 13200 / 17390: Training loss - 3.3331103324890137\n","Evaluation loss: 5.4388957\n","Iteration 13210 / 17390: Training loss - 2.955734968185425\n","Iteration 13220 / 17390: Training loss - 3.0533041954040527\n","Iteration 13230 / 17390: Training loss - 3.218043804168701\n","Iteration 13240 / 17390: Training loss - 3.1090846061706543\n","Iteration 13250 / 17390: Training loss - 3.266592025756836\n","Evaluation loss: 5.096052\n","Iteration 13260 / 17390: Training loss - 3.5652475357055664\n","Iteration 13270 / 17390: Training loss - 3.631500482559204\n","Iteration 13280 / 17390: Training loss - 3.8525726795196533\n","Iteration 13290 / 17390: Training loss - 3.265859842300415\n","Iteration 13300 / 17390: Training loss - 3.497699499130249\n","Evaluation loss: 5.691436\n","Iteration 13310 / 17390: Training loss - 3.1554737091064453\n","Iteration 13320 / 17390: Training loss - 3.176628589630127\n","Iteration 13330 / 17390: Training loss - 2.8735480308532715\n","Iteration 13340 / 17390: Training loss - 3.4681012630462646\n","Iteration 13350 / 17390: Training loss - 3.311636209487915\n","Evaluation loss: 5.6286263\n","Iteration 13360 / 17390: Training loss - 3.1788454055786133\n","Iteration 13370 / 17390: Training loss - 3.485213279724121\n","Iteration 13380 / 17390: Training loss - 3.1108720302581787\n","Iteration 13390 / 17390: Training loss - 3.5276477336883545\n","Iteration 13400 / 17390: Training loss - 2.925184726715088\n","Evaluation loss: 5.5519485\n","Iteration 13410 / 17390: Training loss - 3.162749767303467\n","Iteration 13420 / 17390: Training loss - 3.414325714111328\n","Iteration 13430 / 17390: Training loss - 3.331817150115967\n","Iteration 13440 / 17390: Training loss - 3.7962257862091064\n","Iteration 13450 / 17390: Training loss - 3.6046371459960938\n","Evaluation loss: 5.755331\n","Iteration 13460 / 17390: Training loss - 3.4944374561309814\n","Iteration 13470 / 17390: Training loss - 3.6441805362701416\n","Iteration 13480 / 17390: Training loss - 3.3971598148345947\n","Iteration 13490 / 17390: Training loss - 3.4238665103912354\n","Iteration 13500 / 17390: Training loss - 3.018624782562256\n","Evaluation loss: 5.055215\n","Iteration 13510 / 17390: Training loss - 3.583117723464966\n","Iteration 13520 / 17390: Training loss - 3.0595500469207764\n","Iteration 13530 / 17390: Training loss - 3.5930380821228027\n","Iteration 13540 / 17390: Training loss - 3.2761683464050293\n","Iteration 13550 / 17390: Training loss - 3.4680533409118652\n","Evaluation loss: 5.5042596\n","Iteration 13560 / 17390: Training loss - 3.531728506088257\n","Iteration 13570 / 17390: Training loss - 2.8084306716918945\n","Iteration 13580 / 17390: Training loss - 3.0715184211730957\n","Iteration 13590 / 17390: Training loss - 3.9579732418060303\n","Iteration 13600 / 17390: Training loss - 3.1279823780059814\n","Evaluation loss: 5.1546063\n","Iteration 13610 / 17390: Training loss - 3.0809481143951416\n","Iteration 13620 / 17390: Training loss - 3.3301568031311035\n","Iteration 13630 / 17390: Training loss - 2.8956754207611084\n","Iteration 13640 / 17390: Training loss - 3.200681686401367\n","Iteration 13650 / 17390: Training loss - 3.34161376953125\n","Evaluation loss: 4.665187\n","Iteration 13660 / 17390: Training loss - 3.308047294616699\n","Iteration 13670 / 17390: Training loss - 3.2772061824798584\n","Iteration 13680 / 17390: Training loss - 3.249694347381592\n","Iteration 13690 / 17390: Training loss - 2.719402313232422\n","Iteration 13700 / 17390: Training loss - 3.7209911346435547\n","Evaluation loss: 5.620393\n","Iteration 13710 / 17390: Training loss - 3.2973897457122803\n","Iteration 13720 / 17390: Training loss - 3.428372383117676\n","Iteration 13730 / 17390: Training loss - 3.1948108673095703\n","Iteration 13740 / 17390: Training loss - 2.8179244995117188\n","Iteration 13750 / 17390: Training loss - 2.934762954711914\n","Evaluation loss: 4.144801\n","Iteration 13760 / 17390: Training loss - 3.236454725265503\n","Iteration 13770 / 17390: Training loss - 3.9343342781066895\n","Iteration 13780 / 17390: Training loss - 3.761059522628784\n","Iteration 13790 / 17390: Training loss - 3.148772716522217\n","Iteration 13800 / 17390: Training loss - 3.3703320026397705\n","Evaluation loss: 4.9707537\n","Iteration 13810 / 17390: Training loss - 3.170682907104492\n","Iteration 13820 / 17390: Training loss - 2.960845708847046\n","Iteration 13830 / 17390: Training loss - 2.8854548931121826\n","Iteration 13840 / 17390: Training loss - 3.360262155532837\n","Iteration 13850 / 17390: Training loss - 3.3913416862487793\n","Evaluation loss: 4.902902\n","Iteration 13860 / 17390: Training loss - 3.1965363025665283\n","Iteration 13870 / 17390: Training loss - 3.560826063156128\n","Iteration 13880 / 17390: Training loss - 3.225402355194092\n","Iteration 13890 / 17390: Training loss - 2.929124355316162\n","Iteration 13900 / 17390: Training loss - 3.5989301204681396\n","Evaluation loss: 4.909915\n","Iteration 13910 / 17390: Training loss - 2.8313963413238525\n","Iteration 13920 / 17390: Training loss - 3.2680492401123047\n","Iteration 13930 / 17390: Training loss - 3.3110411167144775\n","Iteration 13940 / 17390: Training loss - 3.274198293685913\n","Iteration 13950 / 17390: Training loss - 3.730374336242676\n","Evaluation loss: 5.4350057\n","Iteration 13960 / 17390: Training loss - 3.4693527221679688\n","Iteration 13970 / 17390: Training loss - 3.2734920978546143\n","Iteration 13980 / 17390: Training loss - 2.699258327484131\n","Iteration 13990 / 17390: Training loss - 2.8163464069366455\n","Iteration 14000 / 17390: Training loss - 3.5279593467712402\n","Evaluation loss: 4.1563907\n","Iteration 14010 / 17390: Training loss - 2.7645249366760254\n","Iteration 14020 / 17390: Training loss - 3.2192375659942627\n","Iteration 14030 / 17390: Training loss - 3.1898200511932373\n","Iteration 14040 / 17390: Training loss - 3.1437554359436035\n","Iteration 14050 / 17390: Training loss - 2.8391928672790527\n","Evaluation loss: 5.2157125\n","Iteration 14060 / 17390: Training loss - 3.1481103897094727\n","Iteration 14070 / 17390: Training loss - 3.274474620819092\n","Iteration 14080 / 17390: Training loss - 3.652921199798584\n","Iteration 14090 / 17390: Training loss - 3.3454501628875732\n","Iteration 14100 / 17390: Training loss - 3.1502411365509033\n","Evaluation loss: 4.9834948\n","Iteration 14110 / 17390: Training loss - 3.1966657638549805\n","Iteration 14120 / 17390: Training loss - 3.273826837539673\n","Iteration 14130 / 17390: Training loss - 3.3660800457000732\n","Iteration 14140 / 17390: Training loss - 3.3219399452209473\n","Iteration 14150 / 17390: Training loss - 3.3323071002960205\n","Evaluation loss: 4.966421\n","Epoch: 8\n","Iteration 14160 / 17390: Training loss - 2.781062602996826\n","Iteration 14170 / 17390: Training loss - 3.4408884048461914\n","Iteration 14180 / 17390: Training loss - 2.9690144062042236\n","Iteration 14190 / 17390: Training loss - 3.3267340660095215\n","Iteration 14200 / 17390: Training loss - 2.923185110092163\n","Evaluation loss: 5.645192\n","Iteration 14210 / 17390: Training loss - 2.8700790405273438\n","Iteration 14220 / 17390: Training loss - 3.622044563293457\n","Iteration 14230 / 17390: Training loss - 2.897892951965332\n","Iteration 14240 / 17390: Training loss - 2.9920361042022705\n","Iteration 14250 / 17390: Training loss - 3.0688223838806152\n","Evaluation loss: 5.784621\n","Iteration 14260 / 17390: Training loss - 3.2243237495422363\n","Iteration 14270 / 17390: Training loss - 3.1312196254730225\n","Iteration 14280 / 17390: Training loss - 3.2947640419006348\n","Iteration 14290 / 17390: Training loss - 3.6092309951782227\n","Iteration 14300 / 17390: Training loss - 3.5282623767852783\n","Evaluation loss: 5.931561\n","Iteration 14310 / 17390: Training loss - 3.2737672328948975\n","Iteration 14320 / 17390: Training loss - 3.407949686050415\n","Iteration 14330 / 17390: Training loss - 3.147991180419922\n","Iteration 14340 / 17390: Training loss - 3.2768218517303467\n","Iteration 14350 / 17390: Training loss - 3.2824149131774902\n","Evaluation loss: 5.351649\n","Iteration 14360 / 17390: Training loss - 2.960449695587158\n","Iteration 14370 / 17390: Training loss - 3.3305282592773438\n","Iteration 14380 / 17390: Training loss - 3.1035380363464355\n","Iteration 14390 / 17390: Training loss - 3.526533842086792\n","Iteration 14400 / 17390: Training loss - 3.584303379058838\n","Evaluation loss: 5.0396705\n","Iteration 14410 / 17390: Training loss - 3.0996487140655518\n","Iteration 14420 / 17390: Training loss - 3.164987325668335\n","Iteration 14430 / 17390: Training loss - 3.023897409439087\n","Iteration 14440 / 17390: Training loss - 3.463549852371216\n","Iteration 14450 / 17390: Training loss - 3.266406297683716\n","Evaluation loss: 5.2750864\n","Iteration 14460 / 17390: Training loss - 2.979465961456299\n","Iteration 14470 / 17390: Training loss - 2.9252426624298096\n","Iteration 14480 / 17390: Training loss - 2.5477724075317383\n","Iteration 14490 / 17390: Training loss - 2.5338797569274902\n","Iteration 14500 / 17390: Training loss - 2.9834444522857666\n","Evaluation loss: 5.5671754\n","Iteration 14510 / 17390: Training loss - 2.8339240550994873\n","Iteration 14520 / 17390: Training loss - 3.289315938949585\n","Iteration 14530 / 17390: Training loss - 3.2504494190216064\n","Iteration 14540 / 17390: Training loss - 2.913093328475952\n","Iteration 14550 / 17390: Training loss - 2.914891481399536\n","Evaluation loss: 4.5427675\n","Iteration 14560 / 17390: Training loss - 3.309478998184204\n","Iteration 14570 / 17390: Training loss - 3.124160051345825\n","Iteration 14580 / 17390: Training loss - 3.2967207431793213\n","Iteration 14590 / 17390: Training loss - 3.0697295665740967\n","Iteration 14600 / 17390: Training loss - 2.9753520488739014\n","Evaluation loss: 4.6850796\n","Iteration 14610 / 17390: Training loss - 3.2634940147399902\n","Iteration 14620 / 17390: Training loss - 2.8663313388824463\n","Iteration 14630 / 17390: Training loss - 3.106985092163086\n","Iteration 14640 / 17390: Training loss - 3.1963024139404297\n","Iteration 14650 / 17390: Training loss - 2.348802328109741\n","Evaluation loss: 5.3710914\n","Iteration 14660 / 17390: Training loss - 3.0037693977355957\n","Iteration 14670 / 17390: Training loss - 3.211700916290283\n","Iteration 14680 / 17390: Training loss - 3.4247334003448486\n","Iteration 14690 / 17390: Training loss - 3.2323789596557617\n","Iteration 14700 / 17390: Training loss - 2.6772923469543457\n","Evaluation loss: 4.718474\n","Iteration 14710 / 17390: Training loss - 3.289855718612671\n","Iteration 14720 / 17390: Training loss - 3.1771132946014404\n","Iteration 14730 / 17390: Training loss - 2.8521530628204346\n","Iteration 14740 / 17390: Training loss - 3.6150262355804443\n","Iteration 14750 / 17390: Training loss - 3.1908671855926514\n","Evaluation loss: 5.124423\n","Iteration 14760 / 17390: Training loss - 3.387796640396118\n","Iteration 14770 / 17390: Training loss - 2.9675467014312744\n","Iteration 14780 / 17390: Training loss - 2.958967685699463\n","Iteration 14790 / 17390: Training loss - 3.5779337882995605\n","Iteration 14800 / 17390: Training loss - 3.2342472076416016\n","Evaluation loss: 4.5428333\n","Iteration 14810 / 17390: Training loss - 3.403014659881592\n","Iteration 14820 / 17390: Training loss - 3.3355162143707275\n","Iteration 14830 / 17390: Training loss - 3.028878688812256\n","Iteration 14840 / 17390: Training loss - 2.991122245788574\n","Iteration 14850 / 17390: Training loss - 2.992488384246826\n","Evaluation loss: 6.0727057\n","Iteration 14860 / 17390: Training loss - 2.771838903427124\n","Iteration 14870 / 17390: Training loss - 2.76642107963562\n","Iteration 14880 / 17390: Training loss - 3.143773078918457\n","Iteration 14890 / 17390: Training loss - 3.087188720703125\n","Iteration 14900 / 17390: Training loss - 2.965052366256714\n","Evaluation loss: 4.849159\n","Iteration 14910 / 17390: Training loss - 3.132474184036255\n","Iteration 14920 / 17390: Training loss - 3.0960469245910645\n","Iteration 14930 / 17390: Training loss - 3.404555559158325\n","Iteration 14940 / 17390: Training loss - 2.889172077178955\n","Iteration 14950 / 17390: Training loss - 3.209082841873169\n","Evaluation loss: 5.055988\n","Iteration 14960 / 17390: Training loss - 3.0230953693389893\n","Iteration 14970 / 17390: Training loss - 2.8378407955169678\n","Iteration 14980 / 17390: Training loss - 3.040992259979248\n","Iteration 14990 / 17390: Training loss - 3.30178165435791\n","Iteration 15000 / 17390: Training loss - 2.5798308849334717\n","Evaluation loss: 5.407699\n","Iteration 15010 / 17390: Training loss - 3.4877333641052246\n","Iteration 15020 / 17390: Training loss - 3.413456916809082\n","Iteration 15030 / 17390: Training loss - 3.1126210689544678\n","Iteration 15040 / 17390: Training loss - 3.0194952487945557\n","Iteration 15050 / 17390: Training loss - 3.081350326538086\n","Evaluation loss: 5.786244\n","Iteration 15060 / 17390: Training loss - 3.2304253578186035\n","Iteration 15070 / 17390: Training loss - 3.4547362327575684\n","Iteration 15080 / 17390: Training loss - 3.213435173034668\n","Iteration 15090 / 17390: Training loss - 3.4936673641204834\n","Iteration 15100 / 17390: Training loss - 2.847193717956543\n","Evaluation loss: 4.310959\n","Iteration 15110 / 17390: Training loss - 3.484419345855713\n","Iteration 15120 / 17390: Training loss - 2.814255952835083\n","Iteration 15130 / 17390: Training loss - 3.1872477531433105\n","Iteration 15140 / 17390: Training loss - 3.0583112239837646\n","Iteration 15150 / 17390: Training loss - 2.9988160133361816\n","Evaluation loss: 4.5976834\n","Iteration 15160 / 17390: Training loss - 3.246021032333374\n","Iteration 15170 / 17390: Training loss - 2.727867603302002\n","Iteration 15180 / 17390: Training loss - 2.7399487495422363\n","Iteration 15190 / 17390: Training loss - 2.5712506771087646\n","Iteration 15200 / 17390: Training loss - 3.262568473815918\n","Evaluation loss: 5.283307\n","Iteration 15210 / 17390: Training loss - 2.995795726776123\n","Iteration 15220 / 17390: Training loss - 2.7461650371551514\n","Iteration 15230 / 17390: Training loss - 3.5650746822357178\n","Epoch: 9\n","Iteration 15240 / 17390: Training loss - 3.13541841506958\n","Iteration 15250 / 17390: Training loss - 2.52308988571167\n","Evaluation loss: 5.5390525\n","Iteration 15260 / 17390: Training loss - 3.4850330352783203\n","Iteration 15270 / 17390: Training loss - 3.0235345363616943\n","Iteration 15280 / 17390: Training loss - 3.2351105213165283\n","Iteration 15290 / 17390: Training loss - 3.09586763381958\n","Iteration 15300 / 17390: Training loss - 2.986984968185425\n","Evaluation loss: 5.4205046\n","Iteration 15310 / 17390: Training loss - 3.4354498386383057\n","Iteration 15320 / 17390: Training loss - 2.981553554534912\n","Iteration 15330 / 17390: Training loss - 2.866913318634033\n","Iteration 15340 / 17390: Training loss - 2.850511074066162\n","Iteration 15350 / 17390: Training loss - 3.053699016571045\n","Evaluation loss: 5.1323657\n","Iteration 15360 / 17390: Training loss - 3.050546884536743\n","Iteration 15370 / 17390: Training loss - 2.841439962387085\n","Iteration 15380 / 17390: Training loss - 2.9138660430908203\n","Iteration 15390 / 17390: Training loss - 2.9209094047546387\n","Iteration 15400 / 17390: Training loss - 3.264573335647583\n","Evaluation loss: 5.0726\n","Iteration 15410 / 17390: Training loss - 2.9108211994171143\n"],"name":"stdout"},{"output_type":"error","ename":"KeyboardInterrupt","evalue":"ignored","traceback":["\u001b[0;31m---------------------------------------------------------------------------\u001b[0m","\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)","\u001b[0;32m<ipython-input-19-d2a061a0432b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     27\u001b[0m                               model.encoder_mask: batch_input_mask, model.decoder_mask: batch_output_mask}\n\u001b[1;32m     28\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 29\u001b[0;31m             \u001b[0mtrain_loss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0m_\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mstep\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0msess\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mrun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mloss\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_op\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmodel\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mglobal_step\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mfeed\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     30\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     31\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0miteration\u001b[0m \u001b[0;34m%\u001b[0m \u001b[0mprint_per_iter\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36mrun\u001b[0;34m(self, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m    927\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    928\u001b[0m       result = self._run(None, fetches, feed_dict, options_ptr,\n\u001b[0;32m--> 929\u001b[0;31m                          run_metadata_ptr)\n\u001b[0m\u001b[1;32m    930\u001b[0m       \u001b[0;32mif\u001b[0m \u001b[0mrun_metadata\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    931\u001b[0m         \u001b[0mproto_data\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtf_session\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTF_GetBuffer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mrun_metadata_ptr\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run\u001b[0;34m(self, handle, fetches, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1150\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mfinal_fetches\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mfinal_targets\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mhandle\u001b[0m \u001b[0;32mand\u001b[0m \u001b[0mfeed_dict_tensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1151\u001b[0m       results = self._do_run(handle, final_targets, final_fetches,\n\u001b[0;32m-> 1152\u001b[0;31m                              feed_dict_tensor, options, run_metadata)\n\u001b[0m\u001b[1;32m   1153\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1154\u001b[0m       \u001b[0mresults\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_run\u001b[0;34m(self, handle, target_list, fetch_list, feed_dict, options, run_metadata)\u001b[0m\n\u001b[1;32m   1326\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0mhandle\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1327\u001b[0m       return self._do_call(_run_fn, feeds, fetches, targets, options,\n\u001b[0;32m-> 1328\u001b[0;31m                            run_metadata)\n\u001b[0m\u001b[1;32m   1329\u001b[0m     \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1330\u001b[0m       \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0m_prun_fn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeeds\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetches\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_do_call\u001b[0;34m(self, fn, *args)\u001b[0m\n\u001b[1;32m   1332\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_do_call\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1333\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1334\u001b[0;31m       \u001b[0;32mreturn\u001b[0m \u001b[0mfn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0margs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   1335\u001b[0m     \u001b[0;32mexcept\u001b[0m \u001b[0merrors\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mOpError\u001b[0m \u001b[0;32mas\u001b[0m \u001b[0me\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1336\u001b[0m       \u001b[0mmessage\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mcompat\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mas_text\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmessage\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_run_fn\u001b[0;34m(feed_dict, fetch_list, target_list, options, run_metadata)\u001b[0m\n\u001b[1;32m   1317\u001b[0m       \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_extend_graph\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1318\u001b[0m       return self._call_tf_sessionrun(\n\u001b[0;32m-> 1319\u001b[0;31m           options, feed_dict, fetch_list, target_list, run_metadata)\n\u001b[0m\u001b[1;32m   1320\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1321\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_prun_fn\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;32m/usr/local/lib/python3.6/dist-packages/tensorflow/python/client/session.py\u001b[0m in \u001b[0;36m_call_tf_sessionrun\u001b[0;34m(self, options, feed_dict, fetch_list, target_list, run_metadata)\u001b[0m\n\u001b[1;32m   1405\u001b[0m     return tf_session.TF_SessionRun_wrapper(\n\u001b[1;32m   1406\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_session\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0moptions\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtarget_list\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 1407\u001b[0;31m         run_metadata)\n\u001b[0m\u001b[1;32m   1408\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   1409\u001b[0m   \u001b[0;32mdef\u001b[0m \u001b[0m_call_tf_sessionprun\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mhandle\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfeed_dict\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mfetch_list\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n","\u001b[0;31mKeyboardInterrupt\u001b[0m: "]}]},{"metadata":{"id":"3XqFDBePv53H","colab_type":"code","outputId":"25a64553-6bc8-49aa-aa29-d36cc1b24fa3","executionInfo":{"status":"ok","timestamp":1554606585829,"user_tz":420,"elapsed":674,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":347}},"cell_type":"code","source":["plt.plot(iterations, train_losses)\n","plt.plot(iterations, val_losses)\n","plt.show()"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAdkAAAFKCAYAAABRtSXvAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nia+HZAAAAABJRU5ErkJggg==\n","text/plain":["<Figure size 576x396 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"metadata":{"id":"XFdFbcXAXDQE","colab_type":"code","outputId":"93211603-48fd-4878-8693-39fcedbc157f","executionInfo":{"status":"ok","timestamp":1554616439004,"user_tz":420,"elapsed":635,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":347}},"cell_type":"code","source":["plt.plot(iterations, train_losses)\n","plt.plot(iterations, val_losses)\n","plt.show()"],"execution_count":0,"outputs":[{"output_type":"display_data","data":{"image/png":"iVBORw0KGgoAAAANSUhEUgAAAeMAAAFKCAYAAAAuZDceAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nPVasAAAAAElFTkSuQmCC\n","text/plain":["<Figure size 576x396 with 1 Axes>"]},"metadata":{"tags":[]}}]},{"metadata":{"id":"c9bsZ0L7ErQ3","colab_type":"text"},"cell_type":"markdown","source":["# Predict"]},{"metadata":{"id":"680oej7MsHrL","colab_type":"code","outputId":"0986a530-2fae-4953-ffd3-ff80e33a19f8","executionInfo":{"status":"ok","timestamp":1554616581696,"user_tz":420,"elapsed":18532,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":1171}},"cell_type":"code","source":["predict(model, model_file)"],"execution_count":0,"outputs":[{"output_type":"stream","text":["INFO:tensorflow:Restoring parameters from /content/gdrive/My Drive/Project-Lion/models/transformer_model\n","\n","Jokes starting from index: 58690\n","\n","=================================================\n","\n","Question: wonder woman whats the difference between wonder woman and the dickless wonder\n","Answer: one was played in the s by lynda the other by jimmy\n","Predicted answer: <unk> i  doncryt when  ice quarterler to to your your\n","\n","=================================================\n","\n","Question: do you want to hear a powerball joke\n","Answer: sorry odds are you wont get it\n","Predicted answer: <unk> never never mind fix you it means but are i people takes to him eat\n","\n","=================================================\n","\n","Question: what do you call a deer with big eyes\n","Answer: a bigeye deer\n","Predicted answer: <unk> a a  ctyunatning of  imim so a\n","\n","=================================================\n","\n","Question: whats the difference between eating pussy and driving in fog\n","Answer: when youre eating pussy you can see the asshole in front of you\n","Predicted answer: <unk> one one ofs them seen is the a other s oneuff reed\n","\n","=================================================\n","\n","Question: what is springy and springy\n","Answer: the easter bunny\n","Predicted answer: <unk> a a white white christmas christmas i i illegal should tont a but car\n","\n","=================================================\n","\n","Question: why do woman make terrible truck drivers\n","Answer: because you give them one good load and it takes them nine months to deliver\n","Predicted answer: <unk>  becausebecause they they are are always   buck places in\n","\n","=================================================\n","\n","Question: what do cops and mushrooms have in common\n","Answer: you feed em shit and keep em in the dark\n","Predicted answer: <unk> they therere their micro italianwave guess dim are alsont there\n","\n","=================================================\n","\n","Question: which came first the chicken or the egg\n","Answer: trick question the answer is c the farmer\n","Predicted answer: <unk> the the  rorooosterster did did the the fit\n","\n","=================================================\n","\n","Question: why was macaulay culkin an easy target for michael jackson\n","Answer: because he was home alone\n","Predicted answer: <unk>  becausebecause he he was was a a little p shellencefish\n","\n","=================================================\n","\n","Question: knock knock whos there\n","Answer: theowlgoes theowlgoes hoo\n","Predicted answer: <unk> i the eat piloter a what retail didge youd do say to\n"],"name":"stdout"}]},{"metadata":{"id":"PvcQP8Me6CTA","colab_type":"code","outputId":"c230898d-9ad0-4959-e7a9-bc77ef417e9d","executionInfo":{"status":"ok","timestamp":1554616523022,"user_tz":420,"elapsed":4982,"user":{"displayName":"Jerry Chen","photoUrl":"","userId":"09742178804226077447"}},"colab":{"base_uri":"https://localhost:8080/","height":145}},"cell_type":"code","source":["predict(model, model_file, questions=[\"why did the chicken cross the street\"])"],"execution_count":0,"outputs":[{"output_type":"stream","text":["INFO:tensorflow:Restoring parameters from /content/gdrive/My Drive/Project-Lion/models/transformer_model\n","\n","=================================================\n","\n","Question: why did the chicken cross the street\n","Predicted answer: <unk> to to get get to to the the other other side side how how to\n"],"name":"stdout"}]},{"metadata":{"id":"zf6y_6lde9w3","colab_type":"code","colab":{}},"cell_type":"code","source":[""],"execution_count":0,"outputs":[]}]}