├── README.md ├── bert_text_classification_demo └── bert_classification.ipynb └── glove ├── .idea ├── .gitignore ├── glove.iml ├── inspectionProfiles │ └── profiles_settings.xml ├── misc.xml ├── modules.xml └── vcs.xml ├── corpus ├── text8.dev.txt ├── text8.test.txt └── text8.train.txt ├── logs ├── glove-50-1.log ├── glove-50-2.log ├── glove-50-3.log └── glove-50.log ├── model ├── GloveModel.py ├── GloveTrain.py ├── WordEmbeddingDataset.py ├── __pycache__ │ ├── GloveModel.cpython-36.pyc │ ├── GloveModel.cpython-37.pyc │ ├── WordEmbeddingDataset.cpython-36.pyc │ └── WordEmbeddingDataset.cpython-37.pyc └── record-50.log └── weights ├── glove-50-1.th ├── glove-50-2.th ├── glove-50-3.th └── glove-50.th /README.md: -------------------------------------------------------------------------------- 1 | # deep-learning-examples 2 | 深度学习相关的实例 3 | -------------------------------------------------------------------------------- /bert_text_classification_demo/bert_classification.ipynb: -------------------------------------------------------------------------------- 1 | { 2 | "cells": [ 3 | { 4 | "cell_type": "code", 5 | "execution_count": 1, 6 | "metadata": {}, 7 | "outputs": [], 8 | "source": [ 9 | "#part2: bert feature-base\n", 10 | "import numpy as np\n", 11 | "import pandas as pd\n", 12 | "from sklearn.model_selection import train_test_split\n", 13 | "from sklearn.linear_model import LogisticRegression\n", 14 | "from sklearn.model_selection import GridSearchCV\n", 15 | "from sklearn.model_selection import cross_val_score\n", 16 | "import torch\n", 17 | "import transformers as tfs\n", 18 | "import warnings\n", 19 | "\n", 20 | "warnings.filterwarnings('ignore')" 21 | ] 22 | }, 23 | { 24 | "cell_type": "code", 25 | "execution_count": 5, 26 | "metadata": {}, 27 | "outputs": [], 28 | "source": [ 29 | "train_df = pd.read_csv('https://github.com/clairett/pytorch-sentiment-classification/raw/master/data/SST2/train.tsv', delimiter='\\t', header=None)" 30 | ] 31 | }, 32 | { 33 | "cell_type": "code", 34 | "execution_count": 6, 35 | "metadata": {}, 36 | "outputs": [ 37 | { 38 | "name": "stdout", 39 | "output_type": "stream", 40 | "text": [ 41 | "Train set shape: (3000, 2)\n" 42 | ] 43 | } 44 | ], 45 | "source": [ 46 | "train_set = train_df[:3000]\n", 47 | "\n", 48 | "print(\"Train set shape:\", train_set.shape)" 49 | ] 50 | }, 51 | { 52 | "cell_type": "code", 53 | "execution_count": 7, 54 | "metadata": {}, 55 | "outputs": [ 56 | { 57 | "data": { 58 | "text/plain": [ 59 | "1 1565\n", 60 | "0 1435\n", 61 | "Name: 1, dtype: int64" 62 | ] 63 | }, 64 | "execution_count": 7, 65 | "metadata": {}, 66 | "output_type": "execute_result" 67 | } 68 | ], 69 | "source": [ 70 | "train_set[1].value_counts()" 71 | ] 72 | }, 73 | { 74 | "cell_type": "code", 75 | "execution_count": 12, 76 | "metadata": {}, 77 | "outputs": [], 78 | "source": [ 79 | "model_class, tokenizer_class, pretrained_weights = (tfs.BertModel, tfs.BertTokenizer, 'bert-base-uncased')\n", 80 | "tokenizer = tokenizer_class.from_pretrained(pretrained_weights)\n", 81 | "model = model_class.from_pretrained(pretrained_weights)" 82 | ] 83 | }, 84 | { 85 | "cell_type": "code", 86 | "execution_count": 13, 87 | "metadata": {}, 88 | "outputs": [], 89 | "source": [ 90 | "train_tokenized = train_set[0].apply((lambda x: tokenizer.encode(x, add_special_tokens=True)))" 91 | ] 92 | }, 93 | { 94 | "cell_type": "code", 95 | "execution_count": 14, 96 | "metadata": {}, 97 | "outputs": [ 98 | { 99 | "name": "stdout", 100 | "output_type": "stream", 101 | "text": [ 102 | "train set shape: (3000, 66)\n" 103 | ] 104 | } 105 | ], 106 | "source": [ 107 | "train_max_len = 0\n", 108 | "for i in train_tokenized.values:\n", 109 | " if len(i) > train_max_len:\n", 110 | " train_max_len = len(i)\n", 111 | "\n", 112 | "train_padded = np.array([i + [0] * (train_max_len-len(i)) for i in train_tokenized.values])\n", 113 | "print(\"train set shape:\",train_padded.shape)" 114 | ] 115 | }, 116 | { 117 | "cell_type": "code", 118 | "execution_count": 15, 119 | "metadata": {}, 120 | "outputs": [ 121 | { 122 | "name": "stdout", 123 | "output_type": "stream", 124 | "text": [ 125 | "[ 101 1037 18385 1010 6057 1998 2633 18276 2128 16603 1997 5053\n", 126 | " 1998 1996 6841 1998 5687 5469 3152 102 0 0 0 0\n", 127 | " 0 0 0 0 0 0 0 0 0 0 0 0\n", 128 | " 0 0 0 0 0 0 0 0 0 0 0 0\n", 129 | " 0 0 0 0 0 0 0 0 0 0 0 0\n", 130 | " 0 0 0 0 0 0]\n", 131 | "[1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0\n", 132 | " 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]\n" 133 | ] 134 | } 135 | ], 136 | "source": [ 137 | "print(train_padded[0])\n", 138 | "train_attention_mask = np.where(train_padded != 0, 1, 0)\n", 139 | "print(train_attention_mask[0])" 140 | ] 141 | }, 142 | { 143 | "cell_type": "code", 144 | "execution_count": 16, 145 | "metadata": {}, 146 | "outputs": [], 147 | "source": [ 148 | "# 训练集\n", 149 | "train_input_ids = torch.tensor(train_padded).long()\n", 150 | "train_attention_mask = torch.tensor(train_attention_mask).long()\n", 151 | "with torch.no_grad():\n", 152 | " train_last_hidden_states = model(train_input_ids, attention_mask=train_attention_mask)" 153 | ] 154 | }, 155 | { 156 | "cell_type": "code", 157 | "execution_count": 17, 158 | "metadata": {}, 159 | "outputs": [ 160 | { 161 | "data": { 162 | "text/plain": [ 163 | "torch.Size([3000, 66, 768])" 164 | ] 165 | }, 166 | "execution_count": 17, 167 | "metadata": {}, 168 | "output_type": "execute_result" 169 | } 170 | ], 171 | "source": [ 172 | "train_last_hidden_states[0].size()" 173 | ] 174 | }, 175 | { 176 | "cell_type": "code", 177 | "execution_count": 18, 178 | "metadata": {}, 179 | "outputs": [], 180 | "source": [ 181 | "train_features = train_last_hidden_states[0][:,0,:].numpy()\n", 182 | "train_labels = train_set[1]" 183 | ] 184 | }, 185 | { 186 | "cell_type": "code", 187 | "execution_count": 19, 188 | "metadata": {}, 189 | "outputs": [], 190 | "source": [ 191 | "train_features, test_features, train_labels, test_labels = train_test_split(train_features, train_labels)" 192 | ] 193 | }, 194 | { 195 | "cell_type": "code", 196 | "execution_count": 20, 197 | "metadata": {}, 198 | "outputs": [ 199 | { 200 | "data": { 201 | "text/plain": [ 202 | "LogisticRegression(C=1.0, class_weight=None, dual=False, fit_intercept=True,\n", 203 | " intercept_scaling=1, max_iter=100, multi_class='warn',\n", 204 | " n_jobs=None, penalty='l2', random_state=None, solver='warn',\n", 205 | " tol=0.0001, verbose=0, warm_start=False)" 206 | ] 207 | }, 208 | "execution_count": 20, 209 | "metadata": {}, 210 | "output_type": "execute_result" 211 | } 212 | ], 213 | "source": [ 214 | "lr_clf = LogisticRegression()\n", 215 | "lr_clf.fit(train_features, train_labels)" 216 | ] 217 | }, 218 | { 219 | "cell_type": "code", 220 | "execution_count": 21, 221 | "metadata": {}, 222 | "outputs": [ 223 | { 224 | "data": { 225 | "text/plain": [ 226 | "0.8306666666666667" 227 | ] 228 | }, 229 | "execution_count": 21, 230 | "metadata": {}, 231 | "output_type": "execute_result" 232 | } 233 | ], 234 | "source": [ 235 | "lr_clf.score(test_features, test_labels)" 236 | ] 237 | }, 238 | { 239 | "cell_type": "code", 240 | "execution_count": 23, 241 | "metadata": {}, 242 | "outputs": [], 243 | "source": [ 244 | "#part 2 - bert fine-tuned\n", 245 | "import torch\n", 246 | "from torch import nn\n", 247 | "from torch import optim\n", 248 | "import transformers as tfs\n", 249 | "import math\n", 250 | "\n", 251 | "class BertClassificationModel(nn.Module):\n", 252 | " def __init__(self):\n", 253 | " super(BertClassificationModel, self).__init__() \n", 254 | " model_class, tokenizer_class, pretrained_weights = (tfs.BertModel, tfs.BertTokenizer, 'bert-base-uncased') \n", 255 | " self.tokenizer = tokenizer_class.from_pretrained(pretrained_weights)\n", 256 | " self.bert = model_class.from_pretrained(pretrained_weights)\n", 257 | " self.dense = nn.Linear(768, 2) #bert默认的隐藏单元数是768， 输出单元是2，表示二分类\n", 258 | " \n", 259 | " def forward(self, batch_sentences):\n", 260 | " #print(batch_sentences[4])\n", 261 | " batch_tokenized = self.tokenizer.batch_encode_plus(batch_sentences, add_special_tokens=True,\n", 262 | " max_len=66, pad_to_max_length=True) #tokenize、add special token、pad\n", 263 | " input_ids = torch.tensor(batch_tokenized['input_ids'])\n", 264 | " attention_mask = torch.tensor(batch_tokenized['attention_mask'])\n", 265 | " bert_output = self.bert(input_ids, attention_mask=attention_mask)\n", 266 | " bert_cls_hidden_state = bert_output[0][:,0,:] #提取[CLS]对应的隐藏状态\n", 267 | " linear_output = self.dense(bert_cls_hidden_state)\n", 268 | " return linear_output" 269 | ] 270 | }, 271 | { 272 | "cell_type": "code", 273 | "execution_count": 24, 274 | "metadata": {}, 275 | "outputs": [], 276 | "source": [ 277 | "sentences = train_set[0].values\n", 278 | "targets = train_set[1].values\n", 279 | "train_inputs, test_inputs, train_targets, test_targets = train_test_split(sentences, targets)\n", 280 | "\n", 281 | "batch_size = 64\n", 282 | "batch_count = int(len(train_inputs) / batch_size)\n", 283 | "batch_train_inputs, batch_train_targets = [], []\n", 284 | "for i in range(batch_count):\n", 285 | " batch_train_inputs.append(train_inputs[i*batch_size : (i+1)*batch_size])\n", 286 | " batch_train_targets.append(train_targets[i*batch_size : (i+1)*batch_size])" 287 | ] 288 | }, 289 | { 290 | "cell_type": "code", 291 | "execution_count": 28, 292 | "metadata": {}, 293 | "outputs": [ 294 | { 295 | "name": "stdout", 296 | "output_type": "stream", 297 | "text": [ 298 | "Batch: 10, Loss: 0.7067\n", 299 | "Batch: 20, Loss: 0.6043\n", 300 | "Batch: 30, Loss: 0.5483\n", 301 | "Batch: 10, Loss: 0.5091\n", 302 | "Batch: 20, Loss: 0.3184\n", 303 | "Batch: 30, Loss: 0.3790\n", 304 | "Batch: 10, Loss: 0.2721\n", 305 | "Batch: 20, Loss: 0.2135\n", 306 | "Batch: 30, Loss: 0.3903\n" 307 | ] 308 | } 309 | ], 310 | "source": [ 311 | "#train the model\n", 312 | "epochs = 3\n", 313 | "lr = 0.01\n", 314 | "print_every_batch = 10\n", 315 | "bert_classifier_model = BertClassificationModel()\n", 316 | "optimizer = optim.SGD(bert_classifier_model.parameters(), lr=lr)\n", 317 | "criterion = nn.CrossEntropyLoss()\n", 318 | "\n", 319 | "for epoch in range(epochs):\n", 320 | " print_avg_loss = 0\n", 321 | " for i in range(batch_count):\n", 322 | " inputs = batch_train_inputs[i]\n", 323 | " labels = torch.tensor(batch_train_targets[i])\n", 324 | " optimizer.zero_grad()\n", 325 | " outputs = bert_classifier_model(inputs)\n", 326 | " loss = criterion(outputs, labels)\n", 327 | " loss.backward()\n", 328 | " optimizer.step()\n", 329 | " \n", 330 | " print_avg_loss += loss.item()\n", 331 | " if i % print_every_batch == (print_every_batch-1):\n", 332 | " print(\"Batch: %d, Loss: %.4f\" % ((i+1), print_avg_loss/print_every_batch))\n", 333 | " print_avg_loss = 0\n", 334 | " " 335 | ] 336 | }, 337 | { 338 | "cell_type": "code", 339 | "execution_count": 29, 340 | "metadata": {}, 341 | "outputs": [ 342 | { 343 | "name": "stdout", 344 | "output_type": "stream", 345 | "text": [ 346 | "Accuracy: 90.53%\n" 347 | ] 348 | } 349 | ], 350 | "source": [ 351 | "# eval the trained model\n", 352 | "total = len(test_inputs)\n", 353 | "hit = 0\n", 354 | "with torch.no_grad():\n", 355 | " for i in range(total):\n", 356 | " outputs = bert_classifier_model([test_inputs[i]])\n", 357 | " _, predicted = torch.max(outputs, 1)\n", 358 | " if predicted == test_targets[i]:\n", 359 | " hit += 1\n", 360 | "\n", 361 | "print(\"Accuracy: %.2f%%\" % (hit / total * 100))" 362 | ] 363 | }, 364 | { 365 | "cell_type": "code", 366 | "execution_count": null, 367 | "metadata": {}, 368 | "outputs": [], 369 | "source": [] 370 | } 371 | ], 372 | "metadata": { 373 | "kernelspec": { 374 | "display_name": "Python 3", 375 | "language": "python", 376 | "name": "python3" 377 | }, 378 | "language_info": { 379 | "codemirror_mode": { 380 | "name": "ipython", 381 | "version": 3 382 | }, 383 | "file_extension": ".py", 384 | "mimetype": "text/x-python", 385 | "name": "python", 386 | "nbconvert_exporter": "python", 387 | "pygments_lexer": "ipython3", 388 | "version": "3.7.3" 389 | } 390 | }, 391 | "nbformat": 4, 392 | "nbformat_minor": 2 393 | } 394 | -------------------------------------------------------------------------------- /glove/.idea/.gitignore: -------------------------------------------------------------------------------- 1 | # Default ignored files 2 | /workspace.xml -------------------------------------------------------------------------------- /glove/.idea/glove.iml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 11 | -------------------------------------------------------------------------------- /glove/.idea/inspectionProfiles/profiles_settings.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 6 | -------------------------------------------------------------------------------- /glove/.idea/misc.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | -------------------------------------------------------------------------------- /glove/.idea/modules.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | -------------------------------------------------------------------------------- /glove/.idea/vcs.xml: -------------------------------------------------------------------------------- 1 | 2 | 3 | 4 | 5 | 6 | -------------------------------------------------------------------------------- /glove/logs/glove-50-1.log: -------------------------------------------------------------------------------- 1 | epoch: 0, iter: 10000 (0.3761%), loss: 92.86457, 0h 0m 26s (- 1h 55m 23s) 2 | epoch: 0, iter: 20000 (0.7523%), loss: 49.96140, 0h 0m 50s (- 1h 52m 1s) 3 | epoch: 0, iter: 30000 (1.1284%), loss: 27.20846, 0h 1m 15s (- 1h 50m 17s) 4 | epoch: 0, iter: 40000 (1.5046%), loss: 19.99873, 0h 1m 40s (- 1h 50m 9s) 5 | epoch: 0, iter: 50000 (1.8807%), loss: 17.49622, 0h 2m 5s (- 1h 49m 19s) 6 | epoch: 0, iter: 60000 (2.2569%), loss: 15.70824, 0h 2m 32s (- 1h 49m 44s) 7 | epoch: 0, iter: 70000 (2.6330%), loss: 14.27283, 0h 2m 58s (- 1h 49m 54s) 8 | epoch: 0, iter: 80000 (3.0092%), loss: 13.31666, 0h 3m 24s (- 1h 49m 39s) 9 | epoch: 0, iter: 90000 (3.3853%), loss: 12.25934, 0h 3m 51s (- 1h 50m 17s) 10 | epoch: 0, iter: 100000 (3.7615%), loss: 11.57755, 0h 4m 20s (- 1h 50m 59s) 11 | epoch: 0, iter: 110000 (4.1376%), loss: 10.76983, 0h 4m 47s (- 1h 51m 5s) 12 | epoch: 0, iter: 120000 (4.5138%), loss: 10.15626, 0h 5m 16s (- 1h 51m 45s) 13 | epoch: 0, iter: 130000 (4.8899%), loss: 9.58190, 0h 5m 43s (- 1h 51m 29s) 14 | epoch: 0, iter: 140000 (5.2660%), loss: 9.25488, 0h 6m 14s (- 1h 52m 15s) 15 | epoch: 0, iter: 150000 (5.6422%), loss: 8.73018, 0h 6m 42s (- 1h 52m 4s) 16 | epoch: 0, iter: 160000 (6.0183%), loss: 8.20880, 0h 7m 10s (- 1h 52m 3s) 17 | epoch: 0, iter: 170000 (6.3945%), loss: 8.14306, 0h 7m 37s (- 1h 51m 30s) 18 | epoch: 0, iter: 180000 (6.7706%), loss: 7.79843, 0h 8m 5s (- 1h 51m 32s) 19 | epoch: 0, iter: 190000 (7.1468%), loss: 7.61587, 0h 8m 34s (- 1h 51m 30s) 20 | epoch: 0, iter: 200000 (7.5229%), loss: 7.25375, 0h 9m 1s (- 1h 51m 1s) 21 | epoch: 0, iter: 210000 (7.8991%), loss: 7.19949, 0h 9m 28s (- 1h 50m 22s) 22 | epoch: 0, iter: 220000 (8.2752%), loss: 6.90067, 0h 9m 53s (- 1h 49m 38s) 23 | epoch: 0, iter: 230000 (8.6514%), loss: 6.67233, 0h 10m 19s (- 1h 48m 58s) 24 | epoch: 0, iter: 240000 (9.0275%), loss: 6.40627, 0h 10m 44s (- 1h 48m 13s) 25 | epoch: 0, iter: 250000 (9.4037%), loss: 6.53895, 0h 11m 9s (- 1h 47m 30s) 26 | epoch: 0, iter: 260000 (9.7798%), loss: 6.18961, 0h 11m 34s (- 1h 46m 47s) 27 | epoch: 1, iter: 270000 (10.1560%), loss: 2.43159, 0h 11m 59s (- 1h 46m 8s) 28 | epoch: 1, iter: 280000 (10.5321%), loss: 5.87573, 0h 12m 25s (- 1h 45m 31s) 29 | epoch: 1, iter: 290000 (10.9082%), loss: 5.87013, 0h 12m 50s (- 1h 44m 52s) 30 | epoch: 1, iter: 300000 (11.2844%), loss: 5.71348, 0h 13m 15s (- 1h 44m 12s) 31 | epoch: 1, iter: 310000 (11.6605%), loss: 5.61802, 0h 13m 41s (- 1h 43m 43s) 32 | epoch: 1, iter: 320000 (12.0367%), loss: 5.53479, 0h 14m 8s (- 1h 43m 18s) 33 | epoch: 1, iter: 330000 (12.4128%), loss: 5.53895, 0h 14m 36s (- 1h 43m 2s) 34 | epoch: 1, iter: 340000 (12.7890%), loss: 5.45641, 0h 15m 3s (- 1h 42m 37s) 35 | epoch: 1, iter: 350000 (13.1651%), loss: 5.28671, 0h 15m 30s (- 1h 42m 19s) 36 | epoch: 1, iter: 360000 (13.5413%), loss: 5.26368, 0h 15m 59s (- 1h 42m 6s) 37 | epoch: 1, iter: 370000 (13.9174%), loss: 5.24398, 0h 16m 26s (- 1h 41m 39s) 38 | epoch: 1, iter: 380000 (14.2936%), loss: 5.15462, 0h 16m 52s (- 1h 41m 13s) 39 | epoch: 1, iter: 390000 (14.6697%), loss: 5.16390, 0h 17m 18s (- 1h 40m 40s) 40 | epoch: 1, iter: 400000 (15.0459%), loss: 5.08515, 0h 17m 43s (- 1h 40m 5s) 41 | epoch: 1, iter: 410000 (15.4220%), loss: 4.99704, 0h 18m 8s (- 1h 39m 30s) 42 | epoch: 1, iter: 420000 (15.7981%), loss: 4.95028, 0h 18m 34s (- 1h 38m 58s) 43 | epoch: 1, iter: 430000 (16.1743%), loss: 5.05588, 0h 18m 59s (- 1h 38m 24s) 44 | epoch: 1, iter: 440000 (16.5504%), loss: 5.11084, 0h 19m 24s (- 1h 37m 49s) 45 | epoch: 1, iter: 450000 (16.9266%), loss: 4.83665, 0h 19m 49s (- 1h 37m 18s) 46 | epoch: 1, iter: 460000 (17.3027%), loss: 4.76196, 0h 20m 15s (- 1h 36m 50s) 47 | epoch: 1, iter: 470000 (17.6789%), loss: 4.80667, 0h 20m 41s (- 1h 36m 21s) 48 | epoch: 1, iter: 480000 (18.0550%), loss: 4.72734, 0h 21m 7s (- 1h 35m 54s) 49 | epoch: 1, iter: 490000 (18.4312%), loss: 4.68293, 0h 21m 33s (- 1h 35m 23s) 50 | epoch: 1, iter: 500000 (18.8073%), loss: 4.66898, 0h 21m 58s (- 1h 34m 52s) 51 | epoch: 1, iter: 510000 (19.1835%), loss: 4.68207, 0h 22m 24s (- 1h 34m 25s) 52 | epoch: 1, iter: 520000 (19.5596%), loss: 4.61534, 0h 22m 51s (- 1h 34m 0s) 53 | epoch: 1, iter: 530000 (19.9358%), loss: 4.59207, 0h 23m 18s (- 1h 33m 35s) 54 | epoch: 2, iter: 540000 (20.3119%), loss: 3.59004, 0h 23m 46s (- 1h 33m 14s) 55 | epoch: 2, iter: 550000 (20.6880%), loss: 4.42532, 0h 24m 12s (- 1h 32m 48s) 56 | epoch: 2, iter: 560000 (21.0642%), loss: 4.48541, 0h 24m 39s (- 1h 32m 23s) 57 | epoch: 2, iter: 570000 (21.4403%), loss: 4.38792, 0h 25m 5s (- 1h 31m 54s) 58 | epoch: 2, iter: 580000 (21.8165%), loss: 4.36692, 0h 25m 30s (- 1h 31m 24s) 59 | epoch: 2, iter: 590000 (22.1926%), loss: 4.36353, 0h 25m 55s (- 1h 30m 54s) 60 | epoch: 2, iter: 600000 (22.5688%), loss: 4.31120, 0h 26m 21s (- 1h 30m 24s) 61 | epoch: 2, iter: 610000 (22.9449%), loss: 4.41320, 0h 26m 46s (- 1h 29m 53s) 62 | epoch: 2, iter: 620000 (23.3211%), loss: 4.32036, 0h 27m 10s (- 1h 29m 22s) 63 | epoch: 2, iter: 630000 (23.6972%), loss: 4.42845, 0h 27m 35s (- 1h 28m 50s) 64 | epoch: 2, iter: 640000 (24.0734%), loss: 4.31551, 0h 27m 59s (- 1h 28m 16s) 65 | epoch: 2, iter: 650000 (24.4495%), loss: 4.21666, 0h 28m 23s (- 1h 27m 43s) 66 | epoch: 2, iter: 660000 (24.8257%), loss: 4.26595, 0h 28m 48s (- 1h 27m 13s) 67 | epoch: 2, iter: 670000 (25.2018%), loss: 4.15088, 0h 29m 12s (- 1h 26m 41s) 68 | epoch: 2, iter: 680000 (25.5779%), loss: 4.29729, 0h 29m 37s (- 1h 26m 10s) 69 | epoch: 2, iter: 690000 (25.9541%), loss: 4.23945, 0h 30m 2s (- 1h 25m 41s) 70 | epoch: 2, iter: 700000 (26.3302%), loss: 4.11713, 0h 30m 26s (- 1h 25m 10s) 71 | epoch: 2, iter: 710000 (26.7064%), loss: 4.22030, 0h 30m 50s (- 1h 24m 39s) 72 | epoch: 2, iter: 720000 (27.0825%), loss: 4.09687, 0h 31m 15s (- 1h 24m 9s) 73 | epoch: 2, iter: 730000 (27.4587%), loss: 4.15297, 0h 31m 40s (- 1h 23m 39s) 74 | epoch: 2, iter: 740000 (27.8348%), loss: 4.03311, 0h 32m 4s (- 1h 23m 10s) 75 | epoch: 2, iter: 750000 (28.2110%), loss: 3.98184, 0h 32m 29s (- 1h 22m 40s) 76 | epoch: 2, iter: 760000 (28.5871%), loss: 4.06679, 0h 32m 53s (- 1h 22m 10s) 77 | epoch: 2, iter: 770000 (28.9633%), loss: 4.01613, 0h 33m 18s (- 1h 21m 41s) 78 | epoch: 2, iter: 780000 (29.3394%), loss: 4.09542, 0h 33m 43s (- 1h 21m 13s) 79 | epoch: 2, iter: 790000 (29.7156%), loss: 3.96401, 0h 34m 8s (- 1h 20m 44s) 80 | epoch: 3, iter: 800000 (30.0917%), loss: 0.94648, 0h 34m 33s (- 1h 20m 16s) 81 | epoch: 3, iter: 810000 (30.4679%), loss: 3.96123, 0h 34m 57s (- 1h 19m 47s) 82 | epoch: 3, iter: 820000 (30.8440%), loss: 3.86490, 0h 35m 22s (- 1h 19m 19s) 83 | epoch: 3, iter: 830000 (31.2201%), loss: 3.88823, 0h 35m 47s (- 1h 18m 50s) 84 | epoch: 3, iter: 840000 (31.5963%), loss: 3.86794, 0h 36m 12s (- 1h 18m 22s) 85 | epoch: 3, iter: 850000 (31.9724%), loss: 3.94755, 0h 36m 37s (- 1h 17m 55s) 86 | epoch: 3, iter: 860000 (32.3486%), loss: 3.81933, 0h 37m 2s (- 1h 17m 28s) 87 | epoch: 3, iter: 870000 (32.7247%), loss: 3.80439, 0h 37m 27s (- 1h 17m 0s) 88 | epoch: 3, iter: 880000 (33.1009%), loss: 3.86829, 0h 37m 52s (- 1h 16m 33s) 89 | epoch: 3, iter: 890000 (33.4770%), loss: 3.75349, 0h 38m 17s (- 1h 16m 6s) 90 | epoch: 3, iter: 900000 (33.8532%), loss: 3.81364, 0h 38m 43s (- 1h 15m 39s) 91 | epoch: 3, iter: 910000 (34.2293%), loss: 3.78310, 0h 39m 8s (- 1h 15m 12s) 92 | epoch: 3, iter: 920000 (34.6055%), loss: 3.73968, 0h 39m 33s (- 1h 14m 44s) 93 | epoch: 3, iter: 930000 (34.9816%), loss: 3.78936, 0h 39m 58s (- 1h 14m 17s) 94 | epoch: 3, iter: 940000 (35.3578%), loss: 3.80942, 0h 40m 23s (- 1h 13m 50s) 95 | epoch: 3, iter: 950000 (35.7339%), loss: 3.71867, 0h 40m 48s (- 1h 13m 24s) 96 | epoch: 3, iter: 960000 (36.1100%), loss: 3.68658, 0h 41m 14s (- 1h 12m 57s) 97 | epoch: 3, iter: 970000 (36.4862%), loss: 3.63413, 0h 41m 39s (- 1h 12m 30s) 98 | epoch: 3, iter: 980000 (36.8623%), loss: 3.82311, 0h 42m 4s (- 1h 12m 3s) 99 | epoch: 3, iter: 990000 (37.2385%), loss: 3.75991, 0h 42m 31s (- 1h 11m 39s) 100 | epoch: 3, iter: 1000000 (37.6146%), loss: 3.65309, 0h 42m 58s (- 1h 11m 16s) 101 | epoch: 3, iter: 1010000 (37.9908%), loss: 3.87733, 0h 43m 24s (- 1h 10m 50s) 102 | epoch: 3, iter: 1020000 (38.3669%), loss: 3.61614, 0h 43m 50s (- 1h 10m 25s) 103 | epoch: 3, iter: 1030000 (38.7431%), loss: 3.64257, 0h 44m 15s (- 1h 9m 58s) 104 | epoch: 3, iter: 1040000 (39.1192%), loss: 3.64530, 0h 44m 40s (- 1h 9m 31s) 105 | epoch: 3, iter: 1050000 (39.4954%), loss: 3.65395, 0h 45m 5s (- 1h 9m 4s) 106 | epoch: 3, iter: 1060000 (39.8715%), loss: 3.64212, 0h 45m 30s (- 1h 8m 38s) 107 | epoch: 4, iter: 1070000 (40.2477%), loss: 2.36440, 0h 45m 57s (- 1h 8m 13s) 108 | epoch: 4, iter: 1080000 (40.6238%), loss: 3.53831, 0h 46m 25s (- 1h 7m 50s) 109 | epoch: 4, iter: 1090000 (40.9999%), loss: 3.58030, 0h 46m 51s (- 1h 7m 25s) 110 | epoch: 4, iter: 1100000 (41.3761%), loss: 3.48421, 0h 47m 19s (- 1h 7m 3s) 111 | epoch: 4, iter: 1110000 (41.7522%), loss: 3.53088, 0h 47m 45s (- 1h 6m 38s) 112 | epoch: 4, iter: 1120000 (42.1284%), loss: 3.51427, 0h 48m 11s (- 1h 6m 12s) 113 | epoch: 4, iter: 1130000 (42.5045%), loss: 3.58787, 0h 48m 37s (- 1h 5m 46s) 114 | epoch: 4, iter: 1140000 (42.8807%), loss: 3.50215, 0h 49m 3s (- 1h 5m 20s) 115 | epoch: 4, iter: 1150000 (43.2568%), loss: 3.40816, 0h 49m 28s (- 1h 4m 53s) 116 | epoch: 4, iter: 1160000 (43.6330%), loss: 3.58338, 0h 49m 53s (- 1h 4m 26s) 117 | epoch: 4, iter: 1170000 (44.0091%), loss: 3.44755, 0h 50m 18s (- 1h 3m 59s) 118 | epoch: 4, iter: 1180000 (44.3853%), loss: 3.47206, 0h 50m 43s (- 1h 3m 33s) 119 | epoch: 4, iter: 1190000 (44.7614%), loss: 3.50208, 0h 51m 8s (- 1h 3m 7s) 120 | epoch: 4, iter: 1200000 (45.1376%), loss: 3.40210, 0h 51m 34s (- 1h 2m 41s) 121 | epoch: 4, iter: 1210000 (45.5137%), loss: 3.43364, 0h 52m 0s (- 1h 2m 15s) 122 | epoch: 4, iter: 1220000 (45.8898%), loss: 3.47704, 0h 52m 26s (- 1h 1m 50s) 123 | epoch: 4, iter: 1230000 (46.2660%), loss: 3.45663, 0h 52m 52s (- 1h 1m 24s) 124 | epoch: 4, iter: 1240000 (46.6421%), loss: 3.38673, 0h 53m 19s (- 1h 0m 59s) 125 | epoch: 4, iter: 1250000 (47.0183%), loss: 3.36118, 0h 53m 45s (- 1h 0m 34s) 126 | epoch: 4, iter: 1260000 (47.3944%), loss: 3.46180, 0h 54m 12s (- 1h 0m 10s) 127 | epoch: 4, iter: 1270000 (47.7706%), loss: 3.42837, 0h 54m 38s (- 0h 59m 44s) 128 | epoch: 4, iter: 1280000 (48.1467%), loss: 3.44432, 0h 55m 4s (- 0h 59m 18s) 129 | epoch: 4, iter: 1290000 (48.5229%), loss: 3.40011, 0h 55m 31s (- 0h 58m 54s) 130 | epoch: 4, iter: 1300000 (48.8990%), loss: 3.40122, 0h 55m 58s (- 0h 58m 29s) 131 | epoch: 4, iter: 1310000 (49.2752%), loss: 3.40839, 0h 56m 24s (- 0h 58m 3s) 132 | epoch: 4, iter: 1320000 (49.6513%), loss: 3.37471, 0h 56m 51s (- 0h 57m 39s) 133 | epoch: 5, iter: 1330000 (50.0275%), loss: 0.25569, 0h 57m 18s (- 0h 57m 14s) 134 | epoch: 5, iter: 1340000 (50.4036%), loss: 3.21507, 0h 57m 42s (- 0h 56m 47s) 135 | epoch: 5, iter: 1350000 (50.7798%), loss: 3.26745, 0h 58m 7s (- 0h 56m 20s) 136 | epoch: 5, iter: 1360000 (51.1559%), loss: 3.29456, 0h 58m 34s (- 0h 55m 56s) 137 | epoch: 5, iter: 1370000 (51.5320%), loss: 3.30635, 0h 59m 9s (- 0h 55m 38s) 138 | epoch: 5, iter: 1380000 (51.9082%), loss: 3.24455, 0h 59m 50s (- 0h 55m 26s) 139 | epoch: 5, iter: 1390000 (52.2843%), loss: 3.26913, 1h 0m 26s (- 0h 55m 9s) 140 | epoch: 5, iter: 1400000 (52.6605%), loss: 3.29212, 1h 1m 0s (- 0h 54m 50s) 141 | epoch: 5, iter: 1410000 (53.0366%), loss: 3.27594, 1h 1m 34s (- 0h 54m 31s) 142 | epoch: 5, iter: 1420000 (53.4128%), loss: 3.20994, 1h 2m 7s (- 0h 54m 11s) 143 | epoch: 5, iter: 1430000 (53.7889%), loss: 3.23428, 1h 2m 41s (- 0h 53m 51s) 144 | epoch: 5, iter: 1440000 (54.1651%), loss: 3.16178, 1h 3m 14s (- 0h 53m 30s) 145 | epoch: 5, iter: 1450000 (54.5412%), loss: 3.32330, 1h 3m 48s (- 0h 53m 10s) 146 | epoch: 5, iter: 1460000 (54.9174%), loss: 3.22935, 1h 4m 22s (- 0h 52m 51s) 147 | epoch: 5, iter: 1470000 (55.2935%), loss: 3.27961, 1h 4m 56s (- 0h 52m 30s) 148 | epoch: 5, iter: 1480000 (55.6697%), loss: 3.29182, 1h 5m 29s (- 0h 52m 9s) 149 | epoch: 5, iter: 1490000 (56.0458%), loss: 3.25655, 1h 6m 2s (- 0h 51m 47s) 150 | epoch: 5, iter: 1500000 (56.4219%), loss: 3.21777, 1h 6m 36s (- 0h 51m 26s) 151 | epoch: 5, iter: 1510000 (56.7981%), loss: 3.19342, 1h 7m 10s (- 0h 51m 5s) 152 | epoch: 5, iter: 1520000 (57.1742%), loss: 3.30349, 1h 7m 43s (- 0h 50m 44s) 153 | epoch: 5, iter: 1530000 (57.5504%), loss: 3.25336, 1h 8m 17s (- 0h 50m 22s) 154 | epoch: 5, iter: 1540000 (57.9265%), loss: 3.19261, 1h 8m 51s (- 0h 50m 0s) 155 | epoch: 5, iter: 1550000 (58.3027%), loss: 3.14542, 1h 9m 24s (- 0h 49m 38s) 156 | epoch: 5, iter: 1560000 (58.6788%), loss: 3.16256, 1h 10m 1s (- 0h 49m 18s) 157 | epoch: 5, iter: 1570000 (59.0550%), loss: 3.18417, 1h 10m 28s (- 0h 48m 51s) 158 | epoch: 5, iter: 1580000 (59.4311%), loss: 3.26877, 1h 10m 53s (- 0h 48m 23s) 159 | epoch: 5, iter: 1590000 (59.8073%), loss: 3.19909, 1h 11m 17s (- 0h 47m 54s) 160 | epoch: 6, iter: 1600000 (60.1834%), loss: 1.46111, 1h 11m 42s (- 0h 47m 26s) 161 | epoch: 6, iter: 1610000 (60.5596%), loss: 3.16764, 1h 12m 7s (- 0h 46m 58s) 162 | epoch: 6, iter: 1620000 (60.9357%), loss: 3.08508, 1h 12m 32s (- 0h 46m 30s) 163 | epoch: 6, iter: 1630000 (61.3118%), loss: 3.08024, 1h 12m 59s (- 0h 46m 3s) 164 | epoch: 6, iter: 1640000 (61.6880%), loss: 3.08277, 1h 13m 25s (- 0h 45m 35s) 165 | epoch: 6, iter: 1650000 (62.0641%), loss: 3.10155, 1h 13m 50s (- 0h 45m 8s) 166 | epoch: 6, iter: 1660000 (62.4403%), loss: 3.07659, 1h 14m 17s (- 0h 44m 41s) 167 | epoch: 6, iter: 1670000 (62.8164%), loss: 3.12121, 1h 14m 42s (- 0h 44m 13s) 168 | epoch: 6, iter: 1680000 (63.1926%), loss: 3.08521, 1h 15m 9s (- 0h 43m 46s) 169 | epoch: 6, iter: 1690000 (63.5687%), loss: 3.06605, 1h 15m 37s (- 0h 43m 20s) 170 | epoch: 6, iter: 1700000 (63.9449%), loss: 2.98073, 1h 16m 4s (- 0h 42m 53s) 171 | epoch: 6, iter: 1710000 (64.3210%), loss: 3.09513, 1h 16m 30s (- 0h 42m 26s) 172 | epoch: 6, iter: 1720000 (64.6972%), loss: 3.08373, 1h 16m 55s (- 0h 41m 58s) 173 | epoch: 6, iter: 1730000 (65.0733%), loss: 3.16916, 1h 17m 22s (- 0h 41m 31s) 174 | epoch: 6, iter: 1740000 (65.4495%), loss: 3.03526, 1h 17m 48s (- 0h 41m 4s) 175 | epoch: 6, iter: 1750000 (65.8256%), loss: 3.05531, 1h 18m 15s (- 0h 40m 37s) 176 | epoch: 6, iter: 1760000 (66.2017%), loss: 3.07280, 1h 18m 41s (- 0h 40m 10s) 177 | epoch: 6, iter: 1770000 (66.5779%), loss: 3.04820, 1h 19m 6s (- 0h 39m 42s) 178 | epoch: 6, iter: 1780000 (66.9540%), loss: 3.10741, 1h 19m 33s (- 0h 39m 15s) 179 | epoch: 6, iter: 1790000 (67.3302%), loss: 2.99784, 1h 19m 59s (- 0h 38m 48s) 180 | epoch: 6, iter: 1800000 (67.7063%), loss: 3.10673, 1h 20m 25s (- 0h 38m 21s) 181 | epoch: 6, iter: 1810000 (68.0825%), loss: 3.07649, 1h 20m 51s (- 0h 37m 54s) 182 | epoch: 6, iter: 1820000 (68.4586%), loss: 3.07239, 1h 21m 17s (- 0h 37m 27s) 183 | epoch: 6, iter: 1830000 (68.8348%), loss: 2.97465, 1h 21m 42s (- 0h 36m 59s) 184 | epoch: 6, iter: 1840000 (69.2109%), loss: 3.12573, 1h 22m 6s (- 0h 36m 31s) 185 | epoch: 6, iter: 1850000 (69.5871%), loss: 3.02472, 1h 22m 31s (- 0h 36m 3s) 186 | epoch: 6, iter: 1860000 (69.9632%), loss: 2.96794, 1h 22m 55s (- 0h 35m 36s) 187 | epoch: 7, iter: 1870000 (70.3394%), loss: 2.70732, 1h 23m 20s (- 0h 35m 8s) 188 | epoch: 7, iter: 1880000 (70.7155%), loss: 2.94222, 1h 23m 44s (- 0h 34m 40s) 189 | epoch: 7, iter: 1890000 (71.0917%), loss: 2.99878, 1h 24m 9s (- 0h 34m 13s) 190 | epoch: 7, iter: 1900000 (71.4678%), loss: 2.97443, 1h 24m 34s (- 0h 33m 46s) 191 | epoch: 7, iter: 1910000 (71.8439%), loss: 2.96662, 1h 25m 2s (- 0h 33m 19s) 192 | epoch: 7, iter: 1920000 (72.2201%), loss: 2.93785, 1h 25m 29s (- 0h 32m 53s) 193 | epoch: 7, iter: 1930000 (72.5962%), loss: 2.91697, 1h 25m 54s (- 0h 32m 25s) 194 | epoch: 7, iter: 1940000 (72.9724%), loss: 2.97648, 1h 26m 19s (- 0h 31m 58s) 195 | epoch: 7, iter: 1950000 (73.3485%), loss: 2.89172, 1h 26m 44s (- 0h 31m 31s) 196 | epoch: 7, iter: 1960000 (73.7247%), loss: 2.90661, 1h 27m 9s (- 0h 31m 3s) 197 | epoch: 7, iter: 1970000 (74.1008%), loss: 2.93943, 1h 27m 35s (- 0h 30m 36s) 198 | epoch: 7, iter: 1980000 (74.4770%), loss: 2.99296, 1h 28m 1s (- 0h 30m 9s) 199 | epoch: 7, iter: 1990000 (74.8531%), loss: 2.92378, 1h 28m 26s (- 0h 29m 42s) 200 | epoch: 7, iter: 2000000 (75.2293%), loss: 2.94073, 1h 28m 52s (- 0h 29m 15s) 201 | epoch: 7, iter: 2010000 (75.6054%), loss: 3.01182, 1h 29m 17s (- 0h 28m 48s) 202 | epoch: 7, iter: 2020000 (75.9816%), loss: 2.93459, 1h 29m 41s (- 0h 28m 21s) 203 | epoch: 7, iter: 2030000 (76.3577%), loss: 2.84046, 1h 30m 6s (- 0h 27m 53s) 204 | epoch: 7, iter: 2040000 (76.7338%), loss: 2.89216, 1h 30m 31s (- 0h 27m 26s) 205 | epoch: 7, iter: 2050000 (77.1100%), loss: 3.04648, 1h 30m 55s (- 0h 26m 59s) 206 | epoch: 7, iter: 2060000 (77.4861%), loss: 2.96871, 1h 31m 20s (- 0h 26m 32s) 207 | epoch: 7, iter: 2070000 (77.8623%), loss: 2.90001, 1h 31m 48s (- 0h 26m 6s) 208 | epoch: 7, iter: 2080000 (78.2384%), loss: 2.92839, 1h 32m 14s (- 0h 25m 39s) 209 | epoch: 7, iter: 2090000 (78.6146%), loss: 2.90026, 1h 32m 41s (- 0h 25m 12s) 210 | epoch: 7, iter: 2100000 (78.9907%), loss: 2.93536, 1h 33m 7s (- 0h 24m 46s) 211 | epoch: 7, iter: 2110000 (79.3669%), loss: 2.93882, 1h 33m 33s (- 0h 24m 19s) 212 | epoch: 7, iter: 2120000 (79.7430%), loss: 2.88577, 1h 34m 0s (- 0h 23m 52s) 213 | epoch: 8, iter: 2130000 (80.1192%), loss: 0.88244, 1h 34m 27s (- 0h 23m 26s) 214 | epoch: 8, iter: 2140000 (80.4953%), loss: 2.85178, 1h 34m 55s (- 0h 22m 59s) 215 | epoch: 8, iter: 2150000 (80.8715%), loss: 2.81177, 1h 35m 20s (- 0h 22m 33s) 216 | epoch: 8, iter: 2160000 (81.2476%), loss: 2.84692, 1h 35m 47s (- 0h 22m 6s) 217 | epoch: 8, iter: 2170000 (81.6237%), loss: 2.81680, 1h 36m 13s (- 0h 21m 39s) 218 | epoch: 8, iter: 2180000 (81.9999%), loss: 2.82378, 1h 36m 39s (- 0h 21m 13s) 219 | epoch: 8, iter: 2190000 (82.3760%), loss: 2.84853, 1h 37m 5s (- 0h 20m 46s) 220 | epoch: 8, iter: 2200000 (82.7522%), loss: 2.78139, 1h 37m 31s (- 0h 20m 19s) 221 | epoch: 8, iter: 2210000 (83.1283%), loss: 2.82786, 1h 37m 56s (- 0h 19m 52s) 222 | epoch: 8, iter: 2220000 (83.5045%), loss: 2.78604, 1h 38m 22s (- 0h 19m 26s) 223 | epoch: 8, iter: 2230000 (83.8806%), loss: 2.84531, 1h 38m 52s (- 0h 19m 0s) 224 | epoch: 8, iter: 2240000 (84.2568%), loss: 2.87955, 1h 39m 20s (- 0h 18m 33s) 225 | epoch: 8, iter: 2250000 (84.6329%), loss: 2.86468, 1h 39m 47s (- 0h 18m 7s) 226 | epoch: 8, iter: 2260000 (85.0091%), loss: 2.84353, 1h 40m 16s (- 0h 17m 40s) 227 | epoch: 8, iter: 2270000 (85.3852%), loss: 2.84273, 1h 40m 47s (- 0h 17m 15s) 228 | epoch: 8, iter: 2280000 (85.7614%), loss: 2.87725, 1h 41m 17s (- 0h 16m 49s) 229 | epoch: 8, iter: 2290000 (86.1375%), loss: 2.83547, 1h 41m 44s (- 0h 16m 22s) 230 | epoch: 8, iter: 2300000 (86.5137%), loss: 2.85483, 1h 42m 11s (- 0h 15m 55s) 231 | epoch: 8, iter: 2310000 (86.8898%), loss: 2.79206, 1h 42m 39s (- 0h 15m 29s) 232 | epoch: 8, iter: 2320000 (87.2659%), loss: 2.87458, 1h 43m 9s (- 0h 15m 3s) 233 | epoch: 8, iter: 2330000 (87.6421%), loss: 2.80382, 1h 43m 37s (- 0h 14m 36s) 234 | epoch: 8, iter: 2340000 (88.0182%), loss: 2.86297, 1h 44m 5s (- 0h 14m 10s) 235 | epoch: 8, iter: 2350000 (88.3944%), loss: 2.79714, 1h 44m 29s (- 0h 13m 43s) 236 | epoch: 8, iter: 2360000 (88.7705%), loss: 2.79348, 1h 44m 53s (- 0h 13m 16s) 237 | epoch: 8, iter: 2370000 (89.1467%), loss: 2.85355, 1h 45m 18s (- 0h 12m 49s) 238 | epoch: 8, iter: 2380000 (89.5228%), loss: 2.85424, 1h 45m 44s (- 0h 12m 22s) 239 | epoch: 8, iter: 2390000 (89.8990%), loss: 2.81557, 1h 46m 9s (- 0h 11m 55s) 240 | epoch: 9, iter: 2400000 (90.2751%), loss: 1.98737, 1h 46m 37s (- 0h 11m 29s) 241 | epoch: 9, iter: 2410000 (90.6513%), loss: 2.77624, 1h 47m 5s (- 0h 11m 2s) 242 | epoch: 9, iter: 2420000 (91.0274%), loss: 2.72111, 1h 47m 31s (- 0h 10m 35s) 243 | epoch: 9, iter: 2430000 (91.4036%), loss: 2.74775, 1h 47m 56s (- 0h 10m 9s) 244 | epoch: 9, iter: 2440000 (91.7797%), loss: 2.72274, 1h 48m 22s (- 0h 9m 42s) 245 | epoch: 9, iter: 2450000 (92.1558%), loss: 2.77082, 1h 48m 50s (- 0h 9m 15s) 246 | epoch: 9, iter: 2460000 (92.5320%), loss: 2.72169, 1h 49m 19s (- 0h 8m 49s) 247 | epoch: 9, iter: 2470000 (92.9081%), loss: 2.73264, 1h 49m 46s (- 0h 8m 22s) 248 | epoch: 9, iter: 2480000 (93.2843%), loss: 2.75245, 1h 50m 13s (- 0h 7m 56s) 249 | epoch: 9, iter: 2490000 (93.6604%), loss: 2.73358, 1h 50m 39s (- 0h 7m 29s) 250 | epoch: 9, iter: 2500000 (94.0366%), loss: 2.73293, 1h 51m 4s (- 0h 7m 2s) 251 | epoch: 9, iter: 2510000 (94.4127%), loss: 2.79031, 1h 51m 30s (- 0h 6m 35s) 252 | epoch: 9, iter: 2520000 (94.7889%), loss: 2.75548, 1h 51m 57s (- 0h 6m 9s) 253 | epoch: 9, iter: 2530000 (95.1650%), loss: 2.73723, 1h 52m 27s (- 0h 5m 42s) 254 | epoch: 9, iter: 2540000 (95.5412%), loss: 2.71113, 1h 52m 57s (- 0h 5m 16s) 255 | epoch: 9, iter: 2550000 (95.9173%), loss: 2.68406, 1h 53m 23s (- 0h 4m 49s) 256 | epoch: 9, iter: 2560000 (96.2935%), loss: 2.78313, 1h 53m 49s (- 0h 4m 22s) 257 | epoch: 9, iter: 2570000 (96.6696%), loss: 2.79222, 1h 54m 17s (- 0h 3m 56s) 258 | epoch: 9, iter: 2580000 (97.0457%), loss: 2.79014, 1h 54m 49s (- 0h 3m 29s) 259 | epoch: 9, iter: 2590000 (97.4219%), loss: 2.79996, 1h 55m 15s (- 0h 3m 3s) 260 | epoch: 9, iter: 2600000 (97.7980%), loss: 2.74891, 1h 55m 42s (- 0h 2m 36s) 261 | epoch: 9, iter: 2610000 (98.1742%), loss: 2.76762, 1h 56m 9s (- 0h 2m 9s) 262 | epoch: 9, iter: 2620000 (98.5503%), loss: 2.74303, 1h 56m 38s (- 0h 1m 42s) 263 | epoch: 9, iter: 2630000 (98.9265%), loss: 2.76841, 1h 57m 4s (- 0h 1m 16s) 264 | epoch: 9, iter: 2640000 (99.3026%), loss: 2.76362, 1h 57m 32s (- 0h 0m 49s) 265 | epoch: 9, iter: 2650000 (99.6788%), loss: 2.72746, 1h 58m 7s (- 0h 0m 22s) 266 | -------------------------------------------------------------------------------- /glove/logs/glove-50-2.log: -------------------------------------------------------------------------------- 1 | epoch: 0, iter: 10000 (0.3761%), loss: 89.00255, 0h 0m 26s (- 1h 57m 52s) 2 | epoch: 0, iter: 10000 (0.3761%), loss: 4.23990, 0h 0m 25s (- 1h 52m 9s) 3 | epoch: 0, iter: 20000 (0.7523%), loss: 3.56112, 0h 0m 50s (- 1h 50m 0s) 4 | epoch: 0, iter: 30000 (1.1284%), loss: 3.37674, 0h 1m 14s (- 1h 49m 6s) 5 | epoch: 0, iter: 40000 (1.5046%), loss: 3.24785, 0h 1m 40s (- 1h 49m 48s) 6 | epoch: 0, iter: 50000 (1.8807%), loss: 3.14348, 0h 2m 9s (- 1h 52m 25s) 7 | epoch: 0, iter: 60000 (2.2569%), loss: 3.00857, 0h 2m 35s (- 1h 52m 18s) 8 | epoch: 0, iter: 70000 (2.6330%), loss: 3.06209, 0h 3m 3s (- 1h 52m 54s) 9 | epoch: 0, iter: 80000 (3.0092%), loss: 3.00152, 0h 3m 30s (- 1h 52m 49s) 10 | epoch: 0, iter: 90000 (3.3853%), loss: 2.92013, 0h 3m 57s (- 1h 53m 0s) 11 | epoch: 0, iter: 100000 (3.7615%), loss: 2.94168, 0h 4m 24s (- 1h 52m 44s) 12 | epoch: 0, iter: 110000 (4.1376%), loss: 2.97350, 0h 4m 50s (- 1h 52m 6s) 13 | epoch: 0, iter: 120000 (4.5138%), loss: 2.99004, 0h 5m 14s (- 1h 50m 56s) 14 | epoch: 0, iter: 130000 (4.8899%), loss: 2.87887, 0h 5m 40s (- 1h 50m 22s) 15 | epoch: 0, iter: 140000 (5.2660%), loss: 2.82599, 0h 6m 6s (- 1h 49m 53s) 16 | epoch: 0, iter: 150000 (5.6422%), loss: 2.89015, 0h 6m 33s (- 1h 49m 46s) 17 | epoch: 0, iter: 160000 (6.0183%), loss: 2.89898, 0h 6m 58s (- 1h 48m 58s) 18 | epoch: 0, iter: 170000 (6.3945%), loss: 2.88210, 0h 7m 23s (- 1h 48m 12s) 19 | epoch: 0, iter: 180000 (6.7706%), loss: 2.87438, 0h 7m 48s (- 1h 47m 32s) 20 | epoch: 0, iter: 190000 (7.1468%), loss: 2.79753, 0h 8m 13s (- 1h 46m 49s) 21 | epoch: 0, iter: 200000 (7.5229%), loss: 2.85030, 0h 8m 38s (- 1h 46m 12s) 22 | epoch: 0, iter: 210000 (7.8991%), loss: 2.90540, 0h 9m 3s (- 1h 45m 32s) 23 | epoch: 0, iter: 220000 (8.2752%), loss: 2.77856, 0h 9m 38s (- 1h 46m 48s) 24 | epoch: 0, iter: 230000 (8.6514%), loss: 2.79845, 0h 10m 11s (- 1h 47m 32s) 25 | epoch: 0, iter: 240000 (9.0275%), loss: 2.79822, 0h 10m 47s (- 1h 48m 40s) 26 | epoch: 0, iter: 250000 (9.4037%), loss: 2.80284, 0h 11m 22s (- 1h 49m 35s) 27 | epoch: 0, iter: 260000 (9.7798%), loss: 2.81575, 0h 11m 58s (- 1h 50m 24s) 28 | epoch: 1, iter: 270000 (10.1560%), loss: 1.07692, 0h 12m 31s (- 1h 50m 49s) 29 | epoch: 1, iter: 280000 (10.5321%), loss: 2.50637, 0h 12m 56s (- 1h 49m 56s) 30 | epoch: 1, iter: 290000 (10.9082%), loss: 2.55343, 0h 13m 21s (- 1h 49m 2s) 31 | epoch: 1, iter: 300000 (11.2844%), loss: 2.52180, 0h 13m 46s (- 1h 48m 19s) 32 | epoch: 1, iter: 310000 (11.6605%), loss: 2.46981, 0h 14m 12s (- 1h 47m 37s) 33 | epoch: 1, iter: 320000 (12.0367%), loss: 2.53523, 0h 14m 37s (- 1h 46m 53s) 34 | epoch: 1, iter: 330000 (12.4128%), loss: 2.48009, 0h 15m 2s (- 1h 46m 10s) 35 | epoch: 1, iter: 340000 (12.7890%), loss: 2.52317, 0h 15m 28s (- 1h 45m 28s) 36 | epoch: 1, iter: 350000 (13.1651%), loss: 2.47349, 0h 15m 53s (- 1h 44m 48s) 37 | epoch: 1, iter: 360000 (13.5413%), loss: 2.52527, 0h 16m 18s (- 1h 44m 9s) 38 | epoch: 1, iter: 370000 (13.9174%), loss: 2.56186, 0h 16m 44s (- 1h 43m 34s) 39 | epoch: 1, iter: 380000 (14.2936%), loss: 2.56055, 0h 17m 9s (- 1h 42m 55s) 40 | epoch: 1, iter: 390000 (14.6697%), loss: 2.53532, 0h 17m 35s (- 1h 42m 20s) 41 | epoch: 1, iter: 400000 (15.0459%), loss: 2.46117, 0h 18m 1s (- 1h 41m 43s) 42 | epoch: 1, iter: 410000 (15.4220%), loss: 2.54861, 0h 18m 26s (- 1h 41m 7s) 43 | epoch: 1, iter: 420000 (15.7981%), loss: 2.53792, 0h 18m 51s (- 1h 40m 31s) 44 | epoch: 1, iter: 430000 (16.1743%), loss: 2.52409, 0h 19m 20s (- 1h 40m 13s) 45 | epoch: 1, iter: 440000 (16.5504%), loss: 2.49265, 0h 19m 49s (- 1h 39m 58s) 46 | epoch: 1, iter: 450000 (16.9266%), loss: 2.58585, 0h 20m 17s (- 1h 39m 36s) 47 | epoch: 1, iter: 460000 (17.3027%), loss: 2.52512, 0h 20m 45s (- 1h 39m 15s) 48 | epoch: 1, iter: 470000 (17.6789%), loss: 2.54264, 0h 21m 14s (- 1h 38m 55s) 49 | epoch: 1, iter: 480000 (18.0550%), loss: 2.52345, 0h 21m 45s (- 1h 38m 46s) 50 | epoch: 1, iter: 490000 (18.4312%), loss: 2.55483, 0h 22m 20s (- 1h 38m 51s) 51 | epoch: 1, iter: 500000 (18.8073%), loss: 2.50006, 0h 22m 53s (- 1h 38m 50s) 52 | epoch: 1, iter: 510000 (19.1835%), loss: 2.50646, 0h 23m 27s (- 1h 38m 51s) 53 | epoch: 1, iter: 520000 (19.5596%), loss: 2.55317, 0h 24m 0s (- 1h 38m 42s) 54 | epoch: 1, iter: 530000 (19.9358%), loss: 2.52338, 0h 24m 32s (- 1h 38m 32s) 55 | epoch: 2, iter: 540000 (20.3119%), loss: 1.96104, 0h 25m 8s (- 1h 38m 38s) 56 | epoch: 2, iter: 550000 (20.6880%), loss: 2.37426, 0h 25m 40s (- 1h 38m 25s) 57 | epoch: 2, iter: 560000 (21.0642%), loss: 2.34239, 0h 26m 13s (- 1h 38m 15s) 58 | epoch: 2, iter: 570000 (21.4403%), loss: 2.41956, 0h 26m 48s (- 1h 38m 12s) 59 | epoch: 2, iter: 580000 (21.8165%), loss: 2.40809, 0h 27m 21s (- 1h 38m 4s) 60 | epoch: 2, iter: 590000 (22.1926%), loss: 2.40283, 0h 27m 55s (- 1h 37m 53s) 61 | epoch: 2, iter: 600000 (22.5688%), loss: 2.41015, 0h 28m 29s (- 1h 37m 43s) 62 | epoch: 2, iter: 610000 (22.9449%), loss: 2.36017, 0h 29m 2s (- 1h 37m 33s) 63 | epoch: 2, iter: 620000 (23.3211%), loss: 2.39561, 0h 29m 37s (- 1h 37m 22s) 64 | epoch: 2, iter: 630000 (23.6972%), loss: 2.38860, 0h 30m 10s (- 1h 37m 8s) 65 | epoch: 2, iter: 640000 (24.0734%), loss: 2.37138, 0h 30m 42s (- 1h 36m 51s) 66 | epoch: 2, iter: 650000 (24.4495%), loss: 2.39914, 0h 31m 21s (- 1h 36m 54s) 67 | epoch: 2, iter: 660000 (24.8257%), loss: 2.42098, 0h 32m 0s (- 1h 36m 55s) 68 | epoch: 2, iter: 670000 (25.2018%), loss: 2.40523, 0h 32m 39s (- 1h 36m 55s) 69 | epoch: 2, iter: 680000 (25.5779%), loss: 2.45009, 0h 33m 16s (- 1h 36m 48s) 70 | epoch: 2, iter: 690000 (25.9541%), loss: 2.43312, 0h 33m 52s (- 1h 36m 38s) 71 | epoch: 2, iter: 700000 (26.3302%), loss: 2.40913, 0h 34m 29s (- 1h 36m 30s) 72 | epoch: 2, iter: 710000 (26.7064%), loss: 2.39561, 0h 35m 6s (- 1h 36m 20s) 73 | epoch: 2, iter: 720000 (27.0825%), loss: 2.38998, 0h 35m 43s (- 1h 36m 11s) 74 | epoch: 2, iter: 730000 (27.4587%), loss: 2.41761, 0h 36m 20s (- 1h 35m 59s) 75 | epoch: 2, iter: 740000 (27.8348%), loss: 2.46604, 0h 36m 56s (- 1h 35m 46s) 76 | epoch: 2, iter: 750000 (28.2110%), loss: 2.41224, 0h 37m 32s (- 1h 35m 32s) 77 | epoch: 2, iter: 760000 (28.5871%), loss: 2.38202, 0h 38m 8s (- 1h 35m 16s) 78 | epoch: 2, iter: 770000 (28.9633%), loss: 2.39928, 0h 38m 43s (- 1h 34m 57s) 79 | epoch: 2, iter: 780000 (29.3394%), loss: 2.40417, 0h 39m 17s (- 1h 34m 38s) 80 | epoch: 2, iter: 790000 (29.7156%), loss: 2.47527, 0h 39m 54s (- 1h 34m 23s) 81 | epoch: 3, iter: 800000 (30.0917%), loss: 0.57014, 0h 40m 36s (- 1h 34m 20s) 82 | epoch: 3, iter: 810000 (30.4679%), loss: 2.30265, 0h 41m 14s (- 1h 34m 6s) 83 | epoch: 3, iter: 820000 (30.8440%), loss: 2.30050, 0h 41m 45s (- 1h 33m 37s) 84 | epoch: 3, iter: 830000 (31.2201%), loss: 2.33637, 0h 42m 14s (- 1h 33m 4s) 85 | epoch: 3, iter: 840000 (31.5963%), loss: 2.31170, 0h 42m 48s (- 1h 32m 41s) 86 | epoch: 3, iter: 850000 (31.9724%), loss: 2.29906, 0h 43m 19s (- 1h 32m 10s) 87 | epoch: 3, iter: 860000 (32.3486%), loss: 2.28992, 0h 43m 44s (- 1h 31m 29s) 88 | epoch: 3, iter: 870000 (32.7247%), loss: 2.29410, 0h 44m 10s (- 1h 30m 48s) 89 | epoch: 3, iter: 880000 (33.1009%), loss: 2.28426, 0h 44m 35s (- 1h 30m 7s) 90 | epoch: 3, iter: 890000 (33.4770%), loss: 2.31735, 0h 45m 1s (- 1h 29m 27s) 91 | epoch: 3, iter: 900000 (33.8532%), loss: 2.30471, 0h 45m 26s (- 1h 28m 48s) 92 | epoch: 3, iter: 910000 (34.2293%), loss: 2.31350, 0h 45m 52s (- 1h 28m 8s) 93 | epoch: 3, iter: 920000 (34.6055%), loss: 2.28167, 0h 46m 17s (- 1h 27m 29s) 94 | epoch: 3, iter: 930000 (34.9816%), loss: 2.34886, 0h 46m 42s (- 1h 26m 49s) 95 | epoch: 3, iter: 940000 (35.3578%), loss: 2.33444, 0h 47m 8s (- 1h 26m 10s) 96 | epoch: 3, iter: 950000 (35.7339%), loss: 2.29114, 0h 47m 33s (- 1h 25m 32s) 97 | epoch: 3, iter: 960000 (36.1100%), loss: 2.30619, 0h 47m 58s (- 1h 24m 53s) 98 | epoch: 3, iter: 970000 (36.4862%), loss: 2.38447, 0h 48m 23s (- 1h 24m 14s) 99 | epoch: 3, iter: 980000 (36.8623%), loss: 2.34918, 0h 48m 48s (- 1h 23m 36s) 100 | epoch: 3, iter: 990000 (37.2385%), loss: 2.33532, 0h 49m 13s (- 1h 22m 58s) 101 | epoch: 3, iter: 1000000 (37.6146%), loss: 2.36332, 0h 49m 38s (- 1h 22m 20s) 102 | epoch: 3, iter: 1010000 (37.9908%), loss: 2.33649, 0h 50m 2s (- 1h 21m 41s) 103 | epoch: 3, iter: 1020000 (38.3669%), loss: 2.35362, 0h 50m 27s (- 1h 21m 2s) 104 | epoch: 3, iter: 1030000 (38.7431%), loss: 2.36210, 0h 50m 51s (- 1h 20m 24s) 105 | epoch: 3, iter: 1040000 (39.1192%), loss: 2.40225, 0h 51m 15s (- 1h 19m 46s) 106 | epoch: 3, iter: 1050000 (39.4954%), loss: 2.35172, 0h 51m 39s (- 1h 19m 8s) 107 | epoch: 3, iter: 1060000 (39.8715%), loss: 2.36974, 0h 52m 3s (- 1h 18m 31s) 108 | epoch: 4, iter: 1070000 (40.2477%), loss: 1.45886, 0h 52m 28s (- 1h 17m 54s) 109 | epoch: 4, iter: 1080000 (40.6238%), loss: 2.23856, 0h 52m 52s (- 1h 17m 16s) 110 | epoch: 4, iter: 1090000 (40.9999%), loss: 2.26468, 0h 53m 16s (- 1h 16m 40s) 111 | epoch: 4, iter: 1100000 (41.3761%), loss: 2.28752, 0h 53m 40s (- 1h 16m 3s) 112 | epoch: 4, iter: 1110000 (41.7522%), loss: 2.23307, 0h 54m 4s (- 1h 15m 26s) 113 | epoch: 4, iter: 1120000 (42.1284%), loss: 2.25352, 0h 54m 28s (- 1h 14m 50s) 114 | epoch: 4, iter: 1130000 (42.5045%), loss: 2.25366, 0h 54m 53s (- 1h 14m 14s) 115 | epoch: 4, iter: 1140000 (42.8807%), loss: 2.25482, 0h 55m 17s (- 1h 13m 38s) 116 | epoch: 4, iter: 1150000 (43.2568%), loss: 2.30038, 0h 55m 41s (- 1h 13m 3s) 117 | epoch: 4, iter: 1160000 (43.6330%), loss: 2.28757, 0h 56m 5s (- 1h 12m 28s) 118 | epoch: 4, iter: 1170000 (44.0091%), loss: 2.24441, 0h 56m 30s (- 1h 11m 52s) 119 | epoch: 4, iter: 1180000 (44.3853%), loss: 2.21708, 0h 56m 54s (- 1h 11m 17s) 120 | epoch: 4, iter: 1190000 (44.7614%), loss: 2.28030, 0h 57m 18s (- 1h 10m 43s) 121 | epoch: 4, iter: 1200000 (45.1376%), loss: 2.26103, 0h 57m 42s (- 1h 10m 8s) 122 | epoch: 4, iter: 1210000 (45.5137%), loss: 2.28927, 0h 58m 6s (- 1h 9m 33s) 123 | epoch: 4, iter: 1220000 (45.8898%), loss: 2.28697, 0h 58m 30s (- 1h 8m 59s) 124 | epoch: 4, iter: 1230000 (46.2660%), loss: 2.28894, 0h 58m 54s (- 1h 8m 25s) 125 | epoch: 4, iter: 1240000 (46.6421%), loss: 2.31984, 0h 59m 19s (- 1h 7m 51s) 126 | epoch: 4, iter: 1250000 (47.0183%), loss: 2.28814, 0h 59m 43s (- 1h 7m 17s) 127 | epoch: 4, iter: 1260000 (47.3944%), loss: 2.32861, 1h 0m 7s (- 1h 6m 44s) 128 | epoch: 4, iter: 1270000 (47.7706%), loss: 2.31273, 1h 0m 32s (- 1h 6m 11s) 129 | epoch: 4, iter: 1280000 (48.1467%), loss: 2.29261, 1h 0m 56s (- 1h 5m 38s) 130 | epoch: 4, iter: 1290000 (48.5229%), loss: 2.27582, 1h 1m 21s (- 1h 5m 6s) 131 | epoch: 4, iter: 1300000 (48.8990%), loss: 2.33096, 1h 1m 46s (- 1h 4m 32s) 132 | epoch: 4, iter: 1310000 (49.2752%), loss: 2.29106, 1h 2m 10s (- 1h 3m 59s) 133 | epoch: 4, iter: 1320000 (49.6513%), loss: 2.24239, 1h 2m 34s (- 1h 3m 26s) 134 | epoch: 5, iter: 1330000 (50.0275%), loss: 0.15822, 1h 2m 58s (- 1h 2m 54s) 135 | epoch: 5, iter: 1340000 (50.4036%), loss: 2.18791, 1h 3m 22s (- 1h 2m 21s) 136 | epoch: 5, iter: 1350000 (50.7798%), loss: 2.21895, 1h 3m 47s (- 1h 1m 50s) 137 | epoch: 5, iter: 1360000 (51.1559%), loss: 2.23810, 1h 4m 16s (- 1h 1m 21s) 138 | epoch: 5, iter: 1370000 (51.5320%), loss: 2.26210, 1h 4m 41s (- 1h 0m 50s) 139 | epoch: 5, iter: 1380000 (51.9082%), loss: 2.23711, 1h 5m 5s (- 1h 0m 18s) 140 | epoch: 5, iter: 1390000 (52.2843%), loss: 2.20307, 1h 5m 29s (- 0h 59m 46s) 141 | epoch: 5, iter: 1400000 (52.6605%), loss: 2.21629, 1h 5m 54s (- 0h 59m 14s) 142 | epoch: 5, iter: 1410000 (53.0366%), loss: 2.21685, 1h 6m 18s (- 0h 58m 43s) 143 | epoch: 5, iter: 1420000 (53.4128%), loss: 2.23507, 1h 6m 45s (- 0h 58m 13s) 144 | epoch: 5, iter: 1430000 (53.7889%), loss: 2.24258, 1h 7m 13s (- 0h 57m 45s) 145 | epoch: 5, iter: 1440000 (54.1651%), loss: 2.20575, 1h 7m 40s (- 0h 57m 16s) 146 | epoch: 5, iter: 1450000 (54.5412%), loss: 2.20946, 1h 8m 6s (- 0h 56m 46s) 147 | epoch: 5, iter: 1460000 (54.9174%), loss: 2.20850, 1h 8m 31s (- 0h 56m 15s) 148 | epoch: 5, iter: 1470000 (55.2935%), loss: 2.21064, 1h 8m 55s (- 0h 55m 44s) 149 | epoch: 5, iter: 1480000 (55.6697%), loss: 2.25551, 1h 9m 21s (- 0h 55m 13s) 150 | epoch: 5, iter: 1490000 (56.0458%), loss: 2.25426, 1h 9m 51s (- 0h 54m 47s) 151 | epoch: 5, iter: 1500000 (56.4219%), loss: 2.19820, 1h 10m 19s (- 0h 54m 19s) 152 | epoch: 5, iter: 1510000 (56.7981%), loss: 2.24299, 1h 10m 48s (- 0h 53m 51s) 153 | epoch: 5, iter: 1520000 (57.1742%), loss: 2.20730, 1h 11m 16s (- 0h 53m 23s) 154 | epoch: 5, iter: 1530000 (57.5504%), loss: 2.25956, 1h 11m 46s (- 0h 52m 56s) 155 | epoch: 5, iter: 1540000 (57.9265%), loss: 2.23702, 1h 12m 12s (- 0h 52m 26s) 156 | epoch: 5, iter: 1550000 (58.3027%), loss: 2.26998, 1h 12m 38s (- 0h 51m 57s) 157 | epoch: 5, iter: 1560000 (58.6788%), loss: 2.25470, 1h 13m 4s (- 0h 51m 27s) 158 | epoch: 5, iter: 1570000 (59.0550%), loss: 2.26432, 1h 13m 30s (- 0h 50m 58s) 159 | epoch: 5, iter: 1580000 (59.4311%), loss: 2.26547, 1h 13m 55s (- 0h 50m 27s) 160 | epoch: 5, iter: 1590000 (59.8073%), loss: 2.28580, 1h 14m 19s (- 0h 49m 56s) 161 | epoch: 6, iter: 1600000 (60.1834%), loss: 1.06640, 1h 14m 47s (- 0h 49m 28s) 162 | epoch: 6, iter: 1610000 (60.5596%), loss: 2.19394, 1h 15m 14s (- 0h 49m 0s) 163 | epoch: 6, iter: 1620000 (60.9357%), loss: 2.16942, 1h 15m 40s (- 0h 48m 30s) 164 | epoch: 6, iter: 1630000 (61.3118%), loss: 2.17930, 1h 16m 6s (- 0h 48m 1s) 165 | epoch: 6, iter: 1640000 (61.6880%), loss: 2.17686, 1h 16m 32s (- 0h 47m 32s) 166 | epoch: 6, iter: 1650000 (62.0641%), loss: 2.18260, 1h 16m 58s (- 0h 47m 2s) 167 | epoch: 6, iter: 1660000 (62.4403%), loss: 2.19942, 1h 17m 23s (- 0h 46m 33s) 168 | epoch: 6, iter: 1670000 (62.8164%), loss: 2.16456, 1h 17m 49s (- 0h 46m 3s) 169 | epoch: 6, iter: 1680000 (63.1926%), loss: 2.20212, 1h 18m 17s (- 0h 45m 35s) 170 | epoch: 6, iter: 1690000 (63.5687%), loss: 2.17432, 1h 18m 43s (- 0h 45m 6s) 171 | epoch: 6, iter: 1700000 (63.9449%), loss: 2.16437, 1h 19m 10s (- 0h 44m 38s) 172 | epoch: 6, iter: 1710000 (64.3210%), loss: 2.18602, 1h 19m 36s (- 0h 44m 9s) 173 | epoch: 6, iter: 1720000 (64.6972%), loss: 2.23061, 1h 20m 1s (- 0h 43m 40s) 174 | epoch: 6, iter: 1730000 (65.0733%), loss: 2.20753, 1h 20m 27s (- 0h 43m 11s) 175 | epoch: 6, iter: 1740000 (65.4495%), loss: 2.21474, 1h 20m 51s (- 0h 42m 41s) 176 | epoch: 6, iter: 1750000 (65.8256%), loss: 2.15269, 1h 21m 18s (- 0h 42m 12s) 177 | epoch: 6, iter: 1760000 (66.2017%), loss: 2.23088, 1h 21m 45s (- 0h 41m 44s) 178 | epoch: 6, iter: 1770000 (66.5779%), loss: 2.20283, 1h 22m 11s (- 0h 41m 15s) 179 | epoch: 6, iter: 1780000 (66.9540%), loss: 2.20501, 1h 22m 35s (- 0h 40m 45s) 180 | epoch: 6, iter: 1790000 (67.3302%), loss: 2.19465, 1h 23m 0s (- 0h 40m 16s) 181 | epoch: 6, iter: 1800000 (67.7063%), loss: 2.22396, 1h 23m 24s (- 0h 39m 46s) 182 | epoch: 6, iter: 1810000 (68.0825%), loss: 2.22853, 1h 23m 48s (- 0h 39m 17s) 183 | epoch: 6, iter: 1820000 (68.4586%), loss: 2.27065, 1h 24m 13s (- 0h 38m 48s) 184 | epoch: 6, iter: 1830000 (68.8348%), loss: 2.24704, 1h 24m 37s (- 0h 38m 19s) 185 | epoch: 6, iter: 1840000 (69.2109%), loss: 2.18483, 1h 25m 5s (- 0h 37m 51s) 186 | epoch: 6, iter: 1850000 (69.5871%), loss: 2.19509, 1h 25m 31s (- 0h 37m 22s) 187 | epoch: 6, iter: 1860000 (69.9632%), loss: 2.27627, 1h 25m 57s (- 0h 36m 54s) 188 | epoch: 7, iter: 1870000 (70.3394%), loss: 1.96092, 1h 26m 43s (- 0h 36m 34s) 189 | epoch: 7, iter: 1880000 (70.7155%), loss: 2.14027, 1h 27m 18s (- 0h 36m 9s) 190 | epoch: 7, iter: 1890000 (71.0917%), loss: 2.13711, 1h 27m 48s (- 0h 35m 42s) 191 | epoch: 7, iter: 1900000 (71.4678%), loss: 2.14337, 1h 28m 16s (- 0h 35m 14s) 192 | epoch: 7, iter: 1910000 (71.8439%), loss: 2.12662, 1h 28m 42s (- 0h 34m 46s) 193 | epoch: 7, iter: 1920000 (72.2201%), loss: 2.17027, 1h 29m 9s (- 0h 34m 17s) 194 | epoch: 7, iter: 1930000 (72.5962%), loss: 2.12584, 1h 29m 54s (- 0h 33m 56s) 195 | epoch: 7, iter: 1940000 (72.9724%), loss: 2.17203, 1h 30m 41s (- 0h 33m 35s) 196 | epoch: 7, iter: 1950000 (73.3485%), loss: 2.16768, 1h 31m 16s (- 0h 33m 10s) 197 | epoch: 7, iter: 1960000 (73.7247%), loss: 2.16004, 1h 31m 44s (- 0h 32m 41s) 198 | epoch: 7, iter: 1970000 (74.1008%), loss: 2.13675, 1h 32m 10s (- 0h 32m 12s) 199 | epoch: 7, iter: 1980000 (74.4770%), loss: 2.16717, 1h 32m 35s (- 0h 31m 43s) 200 | epoch: 7, iter: 1990000 (74.8531%), loss: 2.16852, 1h 33m 0s (- 0h 31m 14s) 201 | epoch: 7, iter: 2000000 (75.2293%), loss: 2.15482, 1h 33m 25s (- 0h 30m 45s) 202 | epoch: 7, iter: 2010000 (75.6054%), loss: 2.15962, 1h 33m 51s (- 0h 30m 17s) 203 | epoch: 7, iter: 2020000 (75.9816%), loss: 2.22663, 1h 34m 17s (- 0h 29m 48s) 204 | epoch: 7, iter: 2030000 (76.3577%), loss: 2.20139, 1h 34m 42s (- 0h 29m 19s) 205 | epoch: 7, iter: 2040000 (76.7338%), loss: 2.23373, 1h 35m 7s (- 0h 28m 50s) 206 | epoch: 7, iter: 2050000 (77.1100%), loss: 2.16034, 1h 35m 32s (- 0h 28m 21s) 207 | epoch: 7, iter: 2060000 (77.4861%), loss: 2.18742, 1h 35m 58s (- 0h 27m 53s) 208 | epoch: 7, iter: 2070000 (77.8623%), loss: 2.19489, 1h 36m 23s (- 0h 27m 24s) 209 | epoch: 7, iter: 2080000 (78.2384%), loss: 2.20431, 1h 36m 48s (- 0h 26m 55s) 210 | epoch: 7, iter: 2090000 (78.6146%), loss: 2.19879, 1h 37m 13s (- 0h 26m 26s) 211 | epoch: 7, iter: 2100000 (78.9907%), loss: 2.18698, 1h 37m 37s (- 0h 25m 57s) 212 | epoch: 7, iter: 2110000 (79.3669%), loss: 2.20181, 1h 38m 2s (- 0h 25m 29s) 213 | epoch: 7, iter: 2120000 (79.7430%), loss: 2.23161, 1h 38m 26s (- 0h 25m 0s) 214 | epoch: 8, iter: 2130000 (80.1192%), loss: 0.65337, 1h 38m 51s (- 0h 24m 31s) 215 | epoch: 8, iter: 2140000 (80.4953%), loss: 2.12894, 1h 39m 16s (- 0h 24m 3s) 216 | epoch: 8, iter: 2150000 (80.8715%), loss: 2.11727, 1h 39m 42s (- 0h 23m 34s) 217 | epoch: 8, iter: 2160000 (81.2476%), loss: 2.15135, 1h 40m 7s (- 0h 23m 6s) 218 | epoch: 8, iter: 2170000 (81.6237%), loss: 2.11057, 1h 40m 31s (- 0h 22m 37s) 219 | epoch: 8, iter: 2180000 (81.9999%), loss: 2.13452, 1h 40m 56s (- 0h 22m 9s) 220 | epoch: 8, iter: 2190000 (82.3760%), loss: 2.12348, 1h 41m 22s (- 0h 21m 41s) 221 | epoch: 8, iter: 2200000 (82.7522%), loss: 2.14145, 1h 41m 48s (- 0h 21m 13s) 222 | epoch: 8, iter: 2210000 (83.1283%), loss: 2.11348, 1h 42m 13s (- 0h 20m 44s) 223 | epoch: 8, iter: 2220000 (83.5045%), loss: 2.18577, 1h 42m 39s (- 0h 20m 16s) 224 | epoch: 8, iter: 2230000 (83.8806%), loss: 2.12986, 1h 43m 12s (- 0h 19m 50s) 225 | epoch: 8, iter: 2240000 (84.2568%), loss: 2.16181, 1h 43m 47s (- 0h 19m 23s) 226 | epoch: 8, iter: 2250000 (84.6329%), loss: 2.14731, 1h 44m 20s (- 0h 18m 56s) 227 | epoch: 8, iter: 2260000 (85.0091%), loss: 2.17851, 1h 44m 53s (- 0h 18m 29s) 228 | epoch: 8, iter: 2270000 (85.3852%), loss: 2.13049, 1h 45m 26s (- 0h 18m 2s) 229 | epoch: 8, iter: 2280000 (85.7614%), loss: 2.15400, 1h 46m 8s (- 0h 17m 37s) 230 | epoch: 8, iter: 2290000 (86.1375%), loss: 2.14661, 1h 46m 43s (- 0h 17m 10s) 231 | epoch: 8, iter: 2300000 (86.5137%), loss: 2.15712, 1h 47m 16s (- 0h 16m 43s) 232 | epoch: 8, iter: 2310000 (86.8898%), loss: 2.13982, 1h 47m 49s (- 0h 16m 16s) 233 | epoch: 8, iter: 2320000 (87.2659%), loss: 2.15644, 1h 48m 22s (- 0h 15m 48s) 234 | epoch: 8, iter: 2330000 (87.6421%), loss: 2.17828, 1h 48m 54s (- 0h 15m 21s) 235 | epoch: 8, iter: 2340000 (88.0182%), loss: 2.15768, 1h 49m 27s (- 0h 14m 54s) 236 | epoch: 8, iter: 2350000 (88.3944%), loss: 2.20500, 1h 50m 3s (- 0h 14m 27s) 237 | epoch: 8, iter: 2360000 (88.7705%), loss: 2.16378, 1h 50m 36s (- 0h 13m 59s) 238 | epoch: 8, iter: 2370000 (89.1467%), loss: 2.18628, 1h 51m 9s (- 0h 13m 31s) 239 | epoch: 8, iter: 2380000 (89.5228%), loss: 2.16947, 1h 51m 41s (- 0h 13m 4s) 240 | epoch: 8, iter: 2390000 (89.8990%), loss: 2.20324, 1h 52m 14s (- 0h 12m 36s) 241 | epoch: 9, iter: 2400000 (90.2751%), loss: 1.54456, 1h 52m 46s (- 0h 12m 8s) 242 | epoch: 9, iter: 2410000 (90.6513%), loss: 2.09595, 1h 53m 19s (- 0h 11m 41s) 243 | epoch: 9, iter: 2420000 (91.0274%), loss: 2.10083, 1h 53m 50s (- 0h 11m 13s) 244 | epoch: 9, iter: 2430000 (91.4036%), loss: 2.10994, 1h 54m 22s (- 0h 10m 45s) 245 | epoch: 9, iter: 2440000 (91.7797%), loss: 2.12289, 1h 54m 54s (- 0h 10m 17s) 246 | epoch: 9, iter: 2450000 (92.1558%), loss: 2.13542, 1h 55m 26s (- 0h 9m 49s) 247 | epoch: 9, iter: 2460000 (92.5320%), loss: 2.12480, 1h 55m 58s (- 0h 9m 21s) 248 | epoch: 9, iter: 2470000 (92.9081%), loss: 2.13357, 1h 56m 30s (- 0h 8m 53s) 249 | epoch: 9, iter: 2480000 (93.2843%), loss: 2.12404, 1h 57m 1s (- 0h 8m 25s) 250 | epoch: 9, iter: 2490000 (93.6604%), loss: 2.12222, 1h 57m 33s (- 0h 7m 57s) 251 | epoch: 9, iter: 2500000 (94.0366%), loss: 2.10488, 1h 58m 5s (- 0h 7m 29s) 252 | epoch: 9, iter: 2510000 (94.4127%), loss: 2.14049, 1h 58m 37s (- 0h 7m 1s) 253 | epoch: 9, iter: 2520000 (94.7889%), loss: 2.13937, 1h 59m 8s (- 0h 6m 32s) 254 | epoch: 9, iter: 2530000 (95.1650%), loss: 2.15016, 1h 59m 40s (- 0h 6m 4s) 255 | epoch: 9, iter: 2540000 (95.5412%), loss: 2.13877, 2h 0m 11s (- 0h 5m 36s) 256 | epoch: 9, iter: 2550000 (95.9173%), loss: 2.12774, 2h 0m 43s (- 0h 5m 8s) 257 | epoch: 9, iter: 2560000 (96.2935%), loss: 2.12512, 2h 1m 15s (- 0h 4m 40s) 258 | epoch: 9, iter: 2570000 (96.6696%), loss: 2.13565, 2h 1m 46s (- 0h 4m 11s) 259 | epoch: 9, iter: 2580000 (97.0457%), loss: 2.12948, 2h 2m 18s (- 0h 3m 43s) 260 | epoch: 9, iter: 2590000 (97.4219%), loss: 2.14585, 2h 2m 49s (- 0h 3m 15s) 261 | epoch: 9, iter: 2600000 (97.7980%), loss: 2.17543, 2h 3m 21s (- 0h 2m 46s) 262 | epoch: 9, iter: 2610000 (98.1742%), loss: 2.12660, 2h 3m 53s (- 0h 2m 18s) 263 | epoch: 9, iter: 2620000 (98.5503%), loss: 2.12200, 2h 4m 27s (- 0h 1m 49s) 264 | epoch: 9, iter: 2630000 (98.9265%), loss: 2.16472, 2h 5m 2s (- 0h 1m 21s) 265 | epoch: 9, iter: 2640000 (99.3026%), loss: 2.14623, 2h 5m 39s (- 0h 0m 52s) 266 | epoch: 9, iter: 2650000 (99.6788%), loss: 2.14151, 2h 6m 13s (- 0h 0m 24s) 267 | -------------------------------------------------------------------------------- /glove/logs/glove-50-3.log: -------------------------------------------------------------------------------- 1 | epoch: 0, iter: 10000 (0.3761%), loss: 3.44183, 0h 0m 25s (- 1h 50m 31s) 2 | epoch: 0, iter: 20000 (0.7523%), loss: 3.01936, 0h 0m 49s (- 1h 48m 7s) 3 | epoch: 0, iter: 30000 (1.1284%), loss: 2.84242, 0h 1m 13s (- 1h 46m 54s) 4 | epoch: 0, iter: 40000 (1.5046%), loss: 2.70479, 0h 1m 37s (- 1h 46m 49s) 5 | epoch: 0, iter: 50000 (1.8807%), loss: 2.61509, 0h 2m 2s (- 1h 46m 23s) 6 | epoch: 0, iter: 60000 (2.2569%), loss: 2.56834, 0h 2m 26s (- 1h 45m 40s) 7 | epoch: 0, iter: 70000 (2.6330%), loss: 2.56049, 0h 2m 50s (- 1h 45m 3s) 8 | epoch: 0, iter: 80000 (3.0092%), loss: 2.52846, 0h 3m 14s (- 1h 44m 29s) 9 | epoch: 0, iter: 90000 (3.3853%), loss: 2.51799, 0h 3m 40s (- 1h 44m 46s) 10 | epoch: 0, iter: 100000 (3.7615%), loss: 2.47396, 0h 4m 4s (- 1h 44m 22s) 11 | epoch: 0, iter: 110000 (4.1376%), loss: 2.49277, 0h 4m 29s (- 1h 43m 53s) 12 | epoch: 0, iter: 120000 (4.5138%), loss: 2.44337, 0h 4m 54s (- 1h 43m 43s) 13 | epoch: 0, iter: 130000 (4.8899%), loss: 2.47379, 0h 5m 18s (- 1h 43m 16s) 14 | epoch: 0, iter: 140000 (5.2660%), loss: 2.46379, 0h 5m 42s (- 1h 42m 44s) 15 | epoch: 0, iter: 150000 (5.6422%), loss: 2.50746, 0h 6m 6s (- 1h 42m 13s) 16 | epoch: 0, iter: 160000 (6.0183%), loss: 2.44395, 0h 6m 30s (- 1h 41m 43s) 17 | epoch: 0, iter: 170000 (6.3945%), loss: 2.41002, 0h 6m 54s (- 1h 41m 14s) 18 | epoch: 0, iter: 180000 (6.7706%), loss: 2.44952, 0h 7m 18s (- 1h 40m 44s) 19 | epoch: 0, iter: 190000 (7.1468%), loss: 2.45448, 0h 7m 43s (- 1h 40m 16s) 20 | epoch: 0, iter: 200000 (7.5229%), loss: 2.41002, 0h 8m 7s (- 1h 39m 50s) 21 | epoch: 0, iter: 210000 (7.8991%), loss: 2.39892, 0h 8m 31s (- 1h 39m 22s) 22 | epoch: 0, iter: 220000 (8.2752%), loss: 2.43904, 0h 8m 55s (- 1h 38m 55s) 23 | epoch: 0, iter: 230000 (8.6514%), loss: 2.43715, 0h 9m 20s (- 1h 38m 34s) 24 | epoch: 0, iter: 240000 (9.0275%), loss: 2.44379, 0h 9m 44s (- 1h 38m 9s) 25 | epoch: 0, iter: 250000 (9.4037%), loss: 2.43152, 0h 10m 8s (- 1h 37m 43s) 26 | epoch: 0, iter: 260000 (9.7798%), loss: 2.42755, 0h 10m 32s (- 1h 37m 17s) 27 | epoch: 1, iter: 270000 (10.1560%), loss: 0.87230, 0h 10m 57s (- 1h 36m 53s) 28 | epoch: 1, iter: 280000 (10.5321%), loss: 2.10191, 0h 11m 21s (- 1h 36m 25s) 29 | epoch: 1, iter: 290000 (10.9082%), loss: 2.12928, 0h 11m 45s (- 1h 36m 0s) 30 | epoch: 1, iter: 300000 (11.2844%), loss: 2.14489, 0h 12m 9s (- 1h 35m 34s) 31 | epoch: 1, iter: 310000 (11.6605%), loss: 2.14480, 0h 12m 33s (- 1h 35m 8s) 32 | epoch: 1, iter: 320000 (12.0367%), loss: 2.14743, 0h 12m 57s (- 1h 34m 40s) 33 | epoch: 1, iter: 330000 (12.4128%), loss: 2.16723, 0h 13m 21s (- 1h 34m 14s) 34 | epoch: 1, iter: 340000 (12.7890%), loss: 2.16226, 0h 13m 45s (- 1h 33m 47s) 35 | epoch: 1, iter: 350000 (13.1651%), loss: 2.19250, 0h 14m 9s (- 1h 33m 22s) 36 | epoch: 1, iter: 360000 (13.5413%), loss: 2.15121, 0h 14m 33s (- 1h 32m 56s) 37 | epoch: 1, iter: 370000 (13.9174%), loss: 2.18072, 0h 14m 57s (- 1h 32m 31s) 38 | epoch: 1, iter: 380000 (14.2936%), loss: 2.16037, 0h 15m 21s (- 1h 32m 6s) 39 | epoch: 1, iter: 390000 (14.6697%), loss: 2.20750, 0h 15m 46s (- 1h 31m 44s) 40 | epoch: 1, iter: 400000 (15.0459%), loss: 2.16055, 0h 16m 12s (- 1h 31m 29s) 41 | epoch: 1, iter: 410000 (15.4220%), loss: 2.20408, 0h 16m 37s (- 1h 31m 13s) 42 | epoch: 1, iter: 420000 (15.7981%), loss: 2.21058, 0h 17m 9s (- 1h 31m 28s) 43 | epoch: 1, iter: 430000 (16.1743%), loss: 2.17911, 0h 17m 43s (- 1h 31m 52s) 44 | epoch: 1, iter: 440000 (16.5504%), loss: 2.22190, 0h 18m 21s (- 1h 32m 35s) 45 | epoch: 1, iter: 450000 (16.9266%), loss: 2.22899, 0h 18m 56s (- 1h 32m 56s) 46 | epoch: 1, iter: 460000 (17.3027%), loss: 2.22010, 0h 19m 30s (- 1h 33m 13s) 47 | epoch: 1, iter: 470000 (17.6789%), loss: 2.22067, 0h 20m 1s (- 1h 33m 14s) 48 | epoch: 1, iter: 480000 (18.0550%), loss: 2.20816, 0h 20m 31s (- 1h 33m 9s) 49 | epoch: 1, iter: 490000 (18.4312%), loss: 2.22236, 0h 21m 1s (- 1h 33m 3s) 50 | epoch: 1, iter: 500000 (18.8073%), loss: 2.18344, 0h 21m 32s (- 1h 33m 1s) 51 | epoch: 1, iter: 510000 (19.1835%), loss: 2.21158, 0h 22m 5s (- 1h 33m 5s) 52 | epoch: 1, iter: 520000 (19.5596%), loss: 2.22212, 0h 22m 35s (- 1h 32m 56s) 53 | epoch: 1, iter: 530000 (19.9358%), loss: 2.18697, 0h 23m 6s (- 1h 32m 47s) 54 | epoch: 2, iter: 540000 (20.3119%), loss: 1.72911, 0h 23m 42s (- 1h 32m 59s) 55 | epoch: 2, iter: 550000 (20.6880%), loss: 2.06331, 0h 24m 17s (- 1h 33m 7s) 56 | epoch: 2, iter: 560000 (21.0642%), loss: 2.09845, 0h 24m 47s (- 1h 32m 55s) 57 | epoch: 2, iter: 570000 (21.4403%), loss: 2.07759, 0h 25m 22s (- 1h 33m 0s) 58 | epoch: 2, iter: 580000 (21.8165%), loss: 2.06383, 0h 25m 57s (- 1h 33m 2s) 59 | epoch: 2, iter: 590000 (22.1926%), loss: 2.07357, 0h 26m 33s (- 1h 33m 6s) 60 | epoch: 2, iter: 600000 (22.5688%), loss: 2.08772, 0h 27m 8s (- 1h 33m 8s) 61 | epoch: 2, iter: 610000 (22.9449%), loss: 2.09649, 0h 27m 42s (- 1h 33m 1s) 62 | epoch: 2, iter: 620000 (23.3211%), loss: 2.10091, 0h 28m 14s (- 1h 32m 50s) 63 | epoch: 2, iter: 630000 (23.6972%), loss: 2.10469, 0h 28m 45s (- 1h 32m 37s) 64 | epoch: 2, iter: 640000 (24.0734%), loss: 2.10093, 0h 29m 17s (- 1h 32m 21s) 65 | epoch: 2, iter: 650000 (24.4495%), loss: 2.09875, 0h 29m 48s (- 1h 32m 7s) 66 | epoch: 2, iter: 660000 (24.8257%), loss: 2.12848, 0h 30m 22s (- 1h 31m 57s) 67 | epoch: 2, iter: 670000 (25.2018%), loss: 2.11049, 0h 30m 55s (- 1h 31m 46s) 68 | epoch: 2, iter: 680000 (25.5779%), loss: 2.12384, 0h 31m 27s (- 1h 31m 30s) 69 | epoch: 2, iter: 690000 (25.9541%), loss: 2.13213, 0h 32m 0s (- 1h 31m 17s) 70 | epoch: 2, iter: 700000 (26.3302%), loss: 2.10539, 0h 32m 32s (- 1h 31m 3s) 71 | epoch: 2, iter: 710000 (26.7064%), loss: 2.13846, 0h 33m 3s (- 1h 30m 44s) 72 | epoch: 2, iter: 720000 (27.0825%), loss: 2.14516, 0h 33m 37s (- 1h 30m 30s) 73 | epoch: 2, iter: 730000 (27.4587%), loss: 2.15690, 0h 34m 15s (- 1h 30m 30s) 74 | epoch: 2, iter: 740000 (27.8348%), loss: 2.17276, 0h 34m 54s (- 1h 30m 30s) 75 | epoch: 2, iter: 750000 (28.2110%), loss: 2.14199, 0h 35m 32s (- 1h 30m 25s) 76 | epoch: 2, iter: 760000 (28.5871%), loss: 2.12951, 0h 36m 6s (- 1h 30m 11s) 77 | epoch: 2, iter: 770000 (28.9633%), loss: 2.12076, 0h 36m 38s (- 1h 29m 51s) 78 | epoch: 2, iter: 780000 (29.3394%), loss: 2.09621, 0h 37m 3s (- 1h 29m 15s) 79 | epoch: 2, iter: 790000 (29.7156%), loss: 2.15327, 0h 37m 30s (- 1h 28m 42s) 80 | epoch: 3, iter: 800000 (30.0917%), loss: 0.49388, 0h 37m 54s (- 1h 28m 4s) 81 | epoch: 3, iter: 810000 (30.4679%), loss: 1.99995, 0h 38m 18s (- 1h 27m 26s) 82 | epoch: 3, iter: 820000 (30.8440%), loss: 2.04447, 0h 38m 42s (- 1h 26m 48s) 83 | epoch: 3, iter: 830000 (31.2201%), loss: 2.02838, 0h 39m 6s (- 1h 26m 10s) 84 | epoch: 3, iter: 840000 (31.5963%), loss: 2.06405, 0h 39m 32s (- 1h 25m 37s) 85 | epoch: 3, iter: 850000 (31.9724%), loss: 2.04268, 0h 39m 58s (- 1h 25m 3s) 86 | epoch: 3, iter: 860000 (32.3486%), loss: 2.05325, 0h 40m 23s (- 1h 24m 29s) 87 | epoch: 3, iter: 870000 (32.7247%), loss: 2.01556, 0h 40m 50s (- 1h 23m 58s) 88 | epoch: 3, iter: 880000 (33.1009%), loss: 2.09432, 0h 41m 17s (- 1h 23m 26s) 89 | epoch: 3, iter: 890000 (33.4770%), loss: 2.07266, 0h 41m 41s (- 1h 22m 51s) 90 | epoch: 3, iter: 900000 (33.8532%), loss: 2.09371, 0h 42m 8s (- 1h 22m 19s) 91 | epoch: 3, iter: 910000 (34.2293%), loss: 2.05994, 0h 42m 34s (- 1h 21m 49s) 92 | epoch: 3, iter: 920000 (34.6055%), loss: 2.07405, 0h 43m 0s (- 1h 21m 15s) 93 | epoch: 3, iter: 930000 (34.9816%), loss: 2.08685, 0h 43m 25s (- 1h 20m 41s) 94 | epoch: 3, iter: 940000 (35.3578%), loss: 2.08037, 0h 43m 49s (- 1h 20m 6s) 95 | epoch: 3, iter: 950000 (35.7339%), loss: 2.08255, 0h 44m 12s (- 1h 19m 31s) 96 | epoch: 3, iter: 960000 (36.1100%), loss: 2.07661, 0h 44m 36s (- 1h 18m 56s) 97 | epoch: 3, iter: 970000 (36.4862%), loss: 2.08263, 0h 45m 1s (- 1h 18m 22s) 98 | epoch: 3, iter: 980000 (36.8623%), loss: 2.12395, 0h 45m 25s (- 1h 17m 48s) 99 | epoch: 3, iter: 990000 (37.2385%), loss: 2.07975, 0h 45m 49s (- 1h 17m 13s) 100 | epoch: 3, iter: 1000000 (37.6146%), loss: 2.09179, 0h 46m 14s (- 1h 16m 41s) 101 | epoch: 3, iter: 1010000 (37.9908%), loss: 2.09122, 0h 46m 40s (- 1h 16m 11s) 102 | epoch: 3, iter: 1020000 (38.3669%), loss: 2.11465, 0h 47m 5s (- 1h 15m 39s) 103 | epoch: 3, iter: 1030000 (38.7431%), loss: 2.10626, 0h 47m 32s (- 1h 15m 10s) 104 | epoch: 3, iter: 1040000 (39.1192%), loss: 2.07933, 0h 47m 57s (- 1h 14m 38s) 105 | epoch: 3, iter: 1050000 (39.4954%), loss: 2.10543, 0h 48m 23s (- 1h 14m 8s) 106 | epoch: 3, iter: 1060000 (39.8715%), loss: 2.10846, 0h 48m 50s (- 1h 13m 39s) 107 | epoch: 4, iter: 1070000 (40.2477%), loss: 1.28646, 0h 49m 17s (- 1h 13m 10s) 108 | epoch: 4, iter: 1080000 (40.6238%), loss: 2.02433, 0h 49m 44s (- 1h 12m 41s) 109 | epoch: 4, iter: 1090000 (40.9999%), loss: 1.98960, 0h 50m 10s (- 1h 12m 12s) 110 | epoch: 4, iter: 1100000 (41.3761%), loss: 2.02933, 0h 50m 36s (- 1h 11m 41s) 111 | epoch: 4, iter: 1110000 (41.7522%), loss: 2.02056, 0h 50m 59s (- 1h 11m 8s) 112 | epoch: 4, iter: 1120000 (42.1284%), loss: 2.00377, 0h 51m 23s (- 1h 10m 35s) 113 | epoch: 4, iter: 1130000 (42.5045%), loss: 2.02278, 0h 51m 47s (- 1h 10m 3s) 114 | epoch: 4, iter: 1140000 (42.8807%), loss: 2.03479, 0h 52m 11s (- 1h 9m 30s) 115 | epoch: 4, iter: 1150000 (43.2568%), loss: 2.02364, 0h 52m 35s (- 1h 8m 58s) 116 | epoch: 4, iter: 1160000 (43.6330%), loss: 2.06939, 0h 52m 58s (- 1h 8m 26s) 117 | epoch: 4, iter: 1170000 (44.0091%), loss: 2.09282, 0h 53m 22s (- 1h 7m 54s) 118 | epoch: 4, iter: 1180000 (44.3853%), loss: 2.06587, 0h 53m 46s (- 1h 7m 22s) 119 | epoch: 4, iter: 1190000 (44.7614%), loss: 2.01028, 0h 54m 10s (- 1h 6m 51s) 120 | epoch: 4, iter: 1200000 (45.1376%), loss: 2.02984, 0h 54m 34s (- 1h 6m 20s) 121 | epoch: 4, iter: 1210000 (45.5137%), loss: 2.05224, 0h 54m 58s (- 1h 5m 49s) 122 | epoch: 4, iter: 1220000 (45.8898%), loss: 2.07301, 0h 55m 22s (- 1h 5m 18s) 123 | epoch: 4, iter: 1230000 (46.2660%), loss: 2.08169, 0h 55m 46s (- 1h 4m 47s) 124 | epoch: 4, iter: 1240000 (46.6421%), loss: 2.03644, 0h 56m 10s (- 1h 4m 15s) 125 | epoch: 4, iter: 1250000 (47.0183%), loss: 2.07852, 0h 56m 34s (- 1h 3m 44s) 126 | epoch: 4, iter: 1260000 (47.3944%), loss: 2.08143, 0h 56m 58s (- 1h 3m 13s) 127 | epoch: 4, iter: 1270000 (47.7706%), loss: 2.08055, 0h 57m 21s (- 1h 2m 43s) 128 | epoch: 4, iter: 1280000 (48.1467%), loss: 2.08491, 0h 57m 45s (- 1h 2m 12s) 129 | epoch: 4, iter: 1290000 (48.5229%), loss: 2.06902, 0h 58m 9s (- 1h 1m 41s) 130 | epoch: 4, iter: 1300000 (48.8990%), loss: 2.07177, 0h 58m 33s (- 1h 1m 11s) 131 | epoch: 4, iter: 1310000 (49.2752%), loss: 2.07476, 0h 58m 56s (- 1h 0m 40s) 132 | epoch: 4, iter: 1320000 (49.6513%), loss: 2.08070, 0h 59m 20s (- 1h 0m 10s) 133 | epoch: 5, iter: 1330000 (50.0275%), loss: 0.13933, 0h 59m 44s (- 0h 59m 40s) 134 | epoch: 5, iter: 1340000 (50.4036%), loss: 1.98851, 1h 0m 8s (- 0h 59m 10s) 135 | epoch: 5, iter: 1350000 (50.7798%), loss: 1.97842, 1h 0m 32s (- 0h 58m 40s) 136 | epoch: 5, iter: 1360000 (51.1559%), loss: 1.99991, 1h 0m 55s (- 0h 58m 10s) 137 | epoch: 5, iter: 1370000 (51.5320%), loss: 2.01677, 1h 1m 19s (- 0h 57m 40s) 138 | epoch: 5, iter: 1380000 (51.9082%), loss: 2.00795, 1h 1m 43s (- 0h 57m 11s) 139 | epoch: 5, iter: 1390000 (52.2843%), loss: 2.01434, 1h 2m 7s (- 0h 56m 41s) 140 | epoch: 5, iter: 1400000 (52.6605%), loss: 1.97428, 1h 2m 30s (- 0h 56m 11s) 141 | epoch: 5, iter: 1410000 (53.0366%), loss: 2.00983, 1h 2m 54s (- 0h 55m 42s) 142 | epoch: 5, iter: 1420000 (53.4128%), loss: 2.03191, 1h 3m 18s (- 0h 55m 13s) 143 | epoch: 5, iter: 1430000 (53.7889%), loss: 1.99030, 1h 3m 42s (- 0h 54m 43s) 144 | epoch: 5, iter: 1440000 (54.1651%), loss: 2.01654, 1h 4m 6s (- 0h 54m 14s) 145 | epoch: 5, iter: 1450000 (54.5412%), loss: 2.03098, 1h 4m 30s (- 0h 53m 45s) 146 | epoch: 5, iter: 1460000 (54.9174%), loss: 2.04121, 1h 4m 54s (- 0h 53m 16s) 147 | epoch: 5, iter: 1470000 (55.2935%), loss: 2.04271, 1h 5m 17s (- 0h 52m 47s) 148 | epoch: 5, iter: 1480000 (55.6697%), loss: 2.04228, 1h 5m 41s (- 0h 52m 18s) 149 | epoch: 5, iter: 1490000 (56.0458%), loss: 2.05897, 1h 6m 5s (- 0h 51m 49s) 150 | epoch: 5, iter: 1500000 (56.4219%), loss: 2.05661, 1h 6m 29s (- 0h 51m 21s) 151 | epoch: 5, iter: 1510000 (56.7981%), loss: 2.06066, 1h 6m 53s (- 0h 50m 52s) 152 | epoch: 5, iter: 1520000 (57.1742%), loss: 2.04659, 1h 7m 16s (- 0h 50m 23s) 153 | epoch: 5, iter: 1530000 (57.5504%), loss: 2.03776, 1h 7m 40s (- 0h 49m 55s) 154 | epoch: 5, iter: 1540000 (57.9265%), loss: 2.05630, 1h 8m 4s (- 0h 49m 26s) 155 | epoch: 5, iter: 1550000 (58.3027%), loss: 2.04782, 1h 8m 28s (- 0h 48m 58s) 156 | epoch: 5, iter: 1560000 (58.6788%), loss: 2.03437, 1h 8m 52s (- 0h 48m 29s) 157 | epoch: 5, iter: 1570000 (59.0550%), loss: 2.06714, 1h 9m 17s (- 0h 48m 2s) 158 | epoch: 5, iter: 1580000 (59.4311%), loss: 2.04709, 1h 9m 43s (- 0h 47m 35s) 159 | epoch: 5, iter: 1590000 (59.8073%), loss: 2.04397, 1h 10m 9s (- 0h 47m 8s) 160 | epoch: 6, iter: 1600000 (60.1834%), loss: 0.94560, 1h 10m 33s (- 0h 46m 40s) 161 | epoch: 6, iter: 1610000 (60.5596%), loss: 1.96777, 1h 10m 57s (- 0h 46m 12s) 162 | epoch: 6, iter: 1620000 (60.9357%), loss: 1.97319, 1h 11m 22s (- 0h 45m 45s) 163 | epoch: 6, iter: 1630000 (61.3118%), loss: 1.97417, 1h 11m 47s (- 0h 45m 18s) 164 | epoch: 6, iter: 1640000 (61.6880%), loss: 1.96789, 1h 12m 11s (- 0h 44m 50s) 165 | epoch: 6, iter: 1650000 (62.0641%), loss: 1.96592, 1h 12m 36s (- 0h 44m 22s) 166 | epoch: 6, iter: 1660000 (62.4403%), loss: 1.99629, 1h 12m 59s (- 0h 43m 54s) 167 | epoch: 6, iter: 1670000 (62.8164%), loss: 2.00893, 1h 13m 24s (- 0h 43m 27s) 168 | epoch: 6, iter: 1680000 (63.1926%), loss: 2.02915, 1h 13m 49s (- 0h 43m 0s) 169 | epoch: 6, iter: 1690000 (63.5687%), loss: 2.01853, 1h 14m 15s (- 0h 42m 33s) 170 | epoch: 6, iter: 1700000 (63.9449%), loss: 2.01835, 1h 14m 40s (- 0h 42m 6s) 171 | epoch: 6, iter: 1710000 (64.3210%), loss: 2.03439, 1h 15m 5s (- 0h 41m 39s) 172 | epoch: 6, iter: 1720000 (64.6972%), loss: 2.04839, 1h 15m 31s (- 0h 41m 12s) 173 | epoch: 6, iter: 1730000 (65.0733%), loss: 2.01215, 1h 15m 57s (- 0h 40m 46s) 174 | epoch: 6, iter: 1740000 (65.4495%), loss: 2.02629, 1h 16m 23s (- 0h 40m 19s) 175 | epoch: 6, iter: 1750000 (65.8256%), loss: 1.99861, 1h 16m 59s (- 0h 39m 58s) 176 | epoch: 6, iter: 1760000 (66.2017%), loss: 2.02231, 1h 17m 35s (- 0h 39m 37s) 177 | epoch: 6, iter: 1770000 (66.5779%), loss: 2.01290, 1h 18m 8s (- 0h 39m 13s) 178 | epoch: 6, iter: 1780000 (66.9540%), loss: 2.02131, 1h 18m 39s (- 0h 38m 49s) 179 | epoch: 6, iter: 1790000 (67.3302%), loss: 2.01125, 1h 19m 12s (- 0h 38m 25s) 180 | epoch: 6, iter: 1800000 (67.7063%), loss: 2.01449, 1h 19m 40s (- 0h 38m 0s) 181 | epoch: 6, iter: 1810000 (68.0825%), loss: 2.01962, 1h 20m 10s (- 0h 37m 35s) 182 | epoch: 6, iter: 1820000 (68.4586%), loss: 2.02819, 1h 20m 39s (- 0h 37m 9s) 183 | epoch: 6, iter: 1830000 (68.8348%), loss: 2.06762, 1h 21m 7s (- 0h 36m 43s) 184 | epoch: 6, iter: 1840000 (69.2109%), loss: 2.03281, 1h 21m 38s (- 0h 36m 19s) 185 | epoch: 6, iter: 1850000 (69.5871%), loss: 2.06630, 1h 22m 7s (- 0h 35m 53s) 186 | epoch: 6, iter: 1860000 (69.9632%), loss: 2.03683, 1h 22m 36s (- 0h 35m 27s) 187 | epoch: 7, iter: 1870000 (70.3394%), loss: 1.75568, 1h 23m 4s (- 0h 35m 1s) 188 | epoch: 7, iter: 1880000 (70.7155%), loss: 1.94221, 1h 23m 31s (- 0h 34m 35s) 189 | epoch: 7, iter: 1890000 (71.0917%), loss: 1.97357, 1h 23m 59s (- 0h 34m 9s) 190 | epoch: 7, iter: 1900000 (71.4678%), loss: 1.94950, 1h 24m 28s (- 0h 33m 43s) 191 | epoch: 7, iter: 1910000 (71.8439%), loss: 1.98160, 1h 24m 57s (- 0h 33m 17s) 192 | epoch: 7, iter: 1920000 (72.2201%), loss: 1.98204, 1h 25m 26s (- 0h 32m 52s) 193 | epoch: 7, iter: 1930000 (72.5962%), loss: 1.98766, 1h 26m 4s (- 0h 32m 29s) 194 | epoch: 7, iter: 1940000 (72.9724%), loss: 2.02286, 1h 26m 37s (- 0h 32m 4s) 195 | epoch: 7, iter: 1950000 (73.3485%), loss: 1.98105, 1h 27m 9s (- 0h 31m 40s) 196 | epoch: 7, iter: 1960000 (73.7247%), loss: 1.97440, 1h 27m 41s (- 0h 31m 15s) 197 | epoch: 7, iter: 1970000 (74.1008%), loss: 2.02292, 1h 28m 14s (- 0h 30m 50s) 198 | epoch: 7, iter: 1980000 (74.4770%), loss: 2.00684, 1h 28m 47s (- 0h 30m 25s) 199 | epoch: 7, iter: 1990000 (74.8531%), loss: 2.00329, 1h 29m 21s (- 0h 30m 1s) 200 | epoch: 7, iter: 2000000 (75.2293%), loss: 2.02230, 1h 29m 55s (- 0h 29m 36s) 201 | epoch: 7, iter: 2010000 (75.6054%), loss: 2.03330, 1h 30m 28s (- 0h 29m 11s) 202 | epoch: 7, iter: 2020000 (75.9816%), loss: 2.00720, 1h 31m 0s (- 0h 28m 46s) 203 | epoch: 7, iter: 2030000 (76.3577%), loss: 1.99933, 1h 31m 33s (- 0h 28m 20s) 204 | epoch: 7, iter: 2040000 (76.7338%), loss: 2.02754, 1h 32m 5s (- 0h 27m 55s) 205 | epoch: 7, iter: 2050000 (77.1100%), loss: 2.02407, 1h 32m 37s (- 0h 27m 29s) 206 | epoch: 7, iter: 2060000 (77.4861%), loss: 2.00424, 1h 33m 10s (- 0h 27m 4s) 207 | epoch: 7, iter: 2070000 (77.8623%), loss: 2.03398, 1h 33m 40s (- 0h 26m 37s) 208 | epoch: 7, iter: 2080000 (78.2384%), loss: 2.00374, 1h 34m 10s (- 0h 26m 11s) 209 | epoch: 7, iter: 2090000 (78.6146%), loss: 2.02240, 1h 34m 40s (- 0h 25m 45s) 210 | epoch: 7, iter: 2100000 (78.9907%), loss: 2.00050, 1h 35m 10s (- 0h 25m 18s) 211 | epoch: 7, iter: 2110000 (79.3669%), loss: 2.03274, 1h 35m 40s (- 0h 24m 52s) 212 | epoch: 7, iter: 2120000 (79.7430%), loss: 2.00418, 1h 36m 10s (- 0h 24m 25s) 213 | epoch: 8, iter: 2130000 (80.1192%), loss: 0.60949, 1h 36m 40s (- 0h 23m 59s) 214 | epoch: 8, iter: 2140000 (80.4953%), loss: 1.96516, 1h 37m 10s (- 0h 23m 32s) 215 | epoch: 8, iter: 2150000 (80.8715%), loss: 1.93759, 1h 37m 41s (- 0h 23m 6s) 216 | epoch: 8, iter: 2160000 (81.2476%), loss: 1.95875, 1h 38m 11s (- 0h 22m 39s) 217 | epoch: 8, iter: 2170000 (81.6237%), loss: 1.97303, 1h 38m 42s (- 0h 22m 13s) 218 | epoch: 8, iter: 2180000 (81.9999%), loss: 1.97177, 1h 39m 12s (- 0h 21m 46s) 219 | epoch: 8, iter: 2190000 (82.3760%), loss: 1.95263, 1h 39m 42s (- 0h 21m 19s) 220 | epoch: 8, iter: 2200000 (82.7522%), loss: 1.95736, 1h 40m 13s (- 0h 20m 53s) 221 | epoch: 8, iter: 2210000 (83.1283%), loss: 1.97626, 1h 40m 43s (- 0h 20m 26s) 222 | epoch: 8, iter: 2220000 (83.5045%), loss: 1.98855, 1h 41m 13s (- 0h 19m 59s) 223 | epoch: 8, iter: 2230000 (83.8806%), loss: 1.99695, 1h 41m 43s (- 0h 19m 32s) 224 | epoch: 8, iter: 2240000 (84.2568%), loss: 1.97252, 1h 42m 14s (- 0h 19m 6s) 225 | epoch: 8, iter: 2250000 (84.6329%), loss: 1.99628, 1h 42m 44s (- 0h 18m 39s) 226 | epoch: 8, iter: 2260000 (85.0091%), loss: 1.97783, 1h 43m 15s (- 0h 18m 12s) 227 | epoch: 8, iter: 2270000 (85.3852%), loss: 1.98183, 1h 43m 45s (- 0h 17m 45s) 228 | epoch: 8, iter: 2280000 (85.7614%), loss: 1.98071, 1h 44m 15s (- 0h 17m 18s) 229 | epoch: 8, iter: 2290000 (86.1375%), loss: 2.02002, 1h 44m 42s (- 0h 16m 51s) 230 | epoch: 8, iter: 2300000 (86.5137%), loss: 1.98167, 1h 45m 8s (- 0h 16m 23s) 231 | epoch: 8, iter: 2310000 (86.8898%), loss: 2.01519, 1h 45m 34s (- 0h 15m 55s) 232 | epoch: 8, iter: 2320000 (87.2659%), loss: 2.04242, 1h 45m 59s (- 0h 15m 28s) 233 | epoch: 8, iter: 2330000 (87.6421%), loss: 2.01019, 1h 46m 24s (- 0h 15m 0s) 234 | epoch: 8, iter: 2340000 (88.0182%), loss: 2.00114, 1h 46m 49s (- 0h 14m 32s) 235 | epoch: 8, iter: 2350000 (88.3944%), loss: 2.02041, 1h 47m 14s (- 0h 14m 4s) 236 | epoch: 8, iter: 2360000 (88.7705%), loss: 2.01906, 1h 47m 39s (- 0h 13m 37s) 237 | epoch: 8, iter: 2370000 (89.1467%), loss: 1.99272, 1h 48m 4s (- 0h 13m 9s) 238 | epoch: 8, iter: 2380000 (89.5228%), loss: 2.01103, 1h 48m 28s (- 0h 12m 41s) 239 | epoch: 8, iter: 2390000 (89.8990%), loss: 2.03332, 1h 48m 52s (- 0h 12m 13s) 240 | epoch: 9, iter: 2400000 (90.2751%), loss: 1.44418, 1h 49m 16s (- 0h 11m 46s) 241 | epoch: 9, iter: 2410000 (90.6513%), loss: 1.92688, 1h 49m 40s (- 0h 11m 18s) 242 | epoch: 9, iter: 2420000 (91.0274%), loss: 1.93048, 1h 50m 5s (- 0h 10m 51s) 243 | epoch: 9, iter: 2430000 (91.4036%), loss: 1.95862, 1h 50m 29s (- 0h 10m 23s) 244 | epoch: 9, iter: 2440000 (91.7797%), loss: 1.95498, 1h 50m 53s (- 0h 9m 55s) 245 | epoch: 9, iter: 2450000 (92.1558%), loss: 1.94435, 1h 51m 17s (- 0h 9m 28s) 246 | epoch: 9, iter: 2460000 (92.5320%), loss: 1.95451, 1h 51m 41s (- 0h 9m 0s) 247 | epoch: 9, iter: 2470000 (92.9081%), loss: 1.97202, 1h 52m 5s (- 0h 8m 33s) 248 | epoch: 9, iter: 2480000 (93.2843%), loss: 1.99874, 1h 52m 29s (- 0h 8m 5s) 249 | epoch: 9, iter: 2490000 (93.6604%), loss: 1.97803, 1h 52m 54s (- 0h 7m 38s) 250 | epoch: 9, iter: 2500000 (94.0366%), loss: 1.96782, 1h 53m 18s (- 0h 7m 11s) 251 | epoch: 9, iter: 2510000 (94.4127%), loss: 1.99530, 1h 53m 43s (- 0h 6m 43s) 252 | epoch: 9, iter: 2520000 (94.7889%), loss: 1.96846, 1h 54m 6s (- 0h 6m 16s) 253 | epoch: 9, iter: 2530000 (95.1650%), loss: 2.00842, 1h 54m 30s (- 0h 5m 49s) 254 | epoch: 9, iter: 2540000 (95.5412%), loss: 1.96119, 1h 54m 55s (- 0h 5m 21s) 255 | epoch: 9, iter: 2550000 (95.9173%), loss: 2.00466, 1h 55m 19s (- 0h 4m 54s) 256 | epoch: 9, iter: 2560000 (96.2935%), loss: 2.02193, 1h 55m 45s (- 0h 4m 27s) 257 | epoch: 9, iter: 2570000 (96.6696%), loss: 1.98834, 1h 56m 10s (- 0h 4m 0s) 258 | epoch: 9, iter: 2580000 (97.0457%), loss: 1.99262, 1h 56m 34s (- 0h 3m 32s) 259 | epoch: 9, iter: 2590000 (97.4219%), loss: 1.96559, 1h 56m 59s (- 0h 3m 5s) 260 | epoch: 9, iter: 2600000 (97.7980%), loss: 2.00044, 1h 57m 23s (- 0h 2m 38s) 261 | epoch: 9, iter: 2610000 (98.1742%), loss: 2.00297, 1h 57m 49s (- 0h 2m 11s) 262 | epoch: 9, iter: 2620000 (98.5503%), loss: 1.99540, 1h 58m 13s (- 0h 1m 44s) 263 | epoch: 9, iter: 2630000 (98.9265%), loss: 2.00239, 1h 58m 38s (- 0h 1m 17s) 264 | epoch: 9, iter: 2640000 (99.3026%), loss: 1.99627, 1h 59m 2s (- 0h 0m 50s) 265 | epoch: 9, iter: 2650000 (99.6788%), loss: 1.98041, 1h 59m 26s (- 0h 0m 23s) 266 | -------------------------------------------------------------------------------- /glove/logs/glove-50.log: -------------------------------------------------------------------------------- 1 | epoch: 0, iter: 10000 (0.2837%), loss: 3582.88399, 0h 2m 10s (- 12h 41m 49s) 2 | epoch: 0, iter: 20000 (0.5674%), loss: 1026.73824, 0h 2m 44s (- 7h 59m 13s) 3 | epoch: 0, iter: 30000 (0.8510%), loss: 713.78422, 0h 3m 18s (- 6h 25m 6s) 4 | epoch: 0, iter: 10000 (0.2837%), loss: 3562.17056, 0h 1m 7s (- 6h 36m 37s) 5 | epoch: 0, iter: 20000 (0.5674%), loss: 998.42624, 0h 2m 15s (- 6h 36m 53s) 6 | epoch: 0, iter: 30000 (0.8510%), loss: 696.92690, 0h 3m 21s (- 6h 30m 46s) 7 | epoch: 1, iter: 40000 (1.1347%), loss: 260.98812, 0h 4m 31s (- 6h 34m 11s) 8 | epoch: 1, iter: 50000 (1.4184%), loss: 512.52003, 0h 5m 38s (- 6h 32m 17s) 9 | epoch: 1, iter: 60000 (1.7021%), loss: 465.81203, 0h 6m 41s (- 6h 26m 46s) 10 | epoch: 1, iter: 70000 (1.9858%), loss: 432.56304, 0h 7m 44s (- 6h 22m 27s) 11 | epoch: 2, iter: 80000 (2.2694%), loss: 381.46248, 0h 8m 49s (- 6h 19m 42s) 12 | epoch: 2, iter: 90000 (2.5531%), loss: 380.94147, 0h 9m 54s (- 6h 18m 8s) 13 | epoch: 2, iter: 100000 (2.8368%), loss: 366.49043, 0h 10m 59s (- 6h 16m 23s) 14 | epoch: 3, iter: 110000 (3.1205%), loss: 147.19145, 0h 12m 4s (- 6h 14m 51s) 15 | epoch: 3, iter: 120000 (3.4042%), loss: 341.42547, 0h 13m 8s (- 6h 12m 58s) 16 | epoch: 3, iter: 130000 (3.6878%), loss: 334.46817, 0h 14m 11s (- 6h 10m 33s) 17 | epoch: 3, iter: 140000 (3.9715%), loss: 325.10545, 0h 15m 11s (- 6h 7m 26s) 18 | epoch: 4, iter: 150000 (4.2552%), loss: 281.61939, 0h 16m 11s (- 6h 4m 24s) 19 | epoch: 4, iter: 160000 (4.5389%), loss: 306.69933, 0h 17m 12s (- 6h 1m 47s) 20 | epoch: 4, iter: 170000 (4.8226%), loss: 307.50975, 0h 18m 11s (- 5h 59m 7s) 21 | epoch: 5, iter: 180000 (5.1062%), loss: 109.63375, 0h 19m 12s (- 5h 57m 3s) 22 | epoch: 5, iter: 190000 (5.3899%), loss: 293.00476, 0h 20m 13s (- 5h 55m 9s) 23 | epoch: 5, iter: 200000 (5.6736%), loss: 289.99045, 0h 21m 14s (- 5h 53m 12s) 24 | epoch: 5, iter: 210000 (5.9573%), loss: 285.68641, 0h 22m 15s (- 5h 51m 27s) 25 | epoch: 6, iter: 220000 (6.2410%), loss: 235.27522, 0h 23m 16s (- 5h 49m 34s) 26 | epoch: 6, iter: 230000 (6.5246%), loss: 276.59855, 0h 24m 16s (- 5h 47m 42s) 27 | epoch: 6, iter: 240000 (6.8083%), loss: 275.16386, 0h 25m 16s (- 5h 45m 53s) 28 | epoch: 7, iter: 250000 (7.0920%), loss: 86.25059, 0h 26m 20s (- 5h 45m 7s) 29 | epoch: 7, iter: 260000 (7.3757%), loss: 266.46127, 0h 27m 17s (- 5h 42m 46s) 30 | epoch: 7, iter: 270000 (7.6594%), loss: 266.14367, 0h 28m 17s (- 5h 41m 10s) 31 | epoch: 7, iter: 280000 (7.9430%), loss: 264.08965, 0h 29m 22s (- 5h 40m 22s) 32 | epoch: 8, iter: 290000 (8.2267%), loss: 204.13761, 0h 30m 24s (- 5h 39m 13s) 33 | epoch: 8, iter: 300000 (8.5104%), loss: 259.30568, 0h 31m 26s (- 5h 38m 0s) 34 | epoch: 8, iter: 310000 (8.7941%), loss: 256.38581, 0h 32m 28s (- 5h 36m 44s) 35 | epoch: 9, iter: 320000 (9.0778%), loss: 68.41626, 0h 33m 30s (- 5h 35m 37s) 36 | epoch: 9, iter: 330000 (9.3614%), loss: 250.75895, 0h 34m 32s (- 5h 34m 25s) 37 | epoch: 9, iter: 340000 (9.6451%), loss: 249.10208, 0h 35m 33s (- 5h 33m 4s) 38 | epoch: 9, iter: 350000 (9.9288%), loss: 251.56643, 0h 36m 33s (- 5h 31m 42s) 39 | epoch: 10, iter: 360000 (10.2125%), loss: 181.30389, 0h 37m 37s (- 5h 30m 46s) 40 | epoch: 10, iter: 370000 (10.4962%), loss: 244.77551, 0h 38m 40s (- 5h 29m 43s) 41 | epoch: 10, iter: 380000 (10.7798%), loss: 244.02124, 0h 39m 44s (- 5h 28m 51s) 42 | epoch: 11, iter: 390000 (11.0635%), loss: 53.46864, 0h 40m 47s (- 5h 27m 53s) 43 | epoch: 11, iter: 400000 (11.3472%), loss: 240.09029, 0h 41m 52s (- 5h 27m 12s) 44 | epoch: 11, iter: 410000 (11.6309%), loss: 240.33604, 0h 42m 56s (- 5h 26m 14s) 45 | epoch: 11, iter: 420000 (11.9146%), loss: 238.76899, 0h 44m 2s (- 5h 25m 33s) 46 | epoch: 12, iter: 430000 (12.1982%), loss: 164.52532, 0h 45m 4s (- 5h 24m 27s) 47 | epoch: 12, iter: 440000 (12.4819%), loss: 235.60874, 0h 46m 15s (- 5h 24m 20s) 48 | epoch: 12, iter: 450000 (12.7656%), loss: 236.03032, 0h 47m 28s (- 5h 24m 25s) 49 | epoch: 13, iter: 460000 (13.0493%), loss: 40.60788, 0h 48m 29s (- 5h 23m 5s) 50 | epoch: 13, iter: 470000 (13.3330%), loss: 231.00218, 0h 49m 27s (- 5h 21m 29s) 51 | epoch: 13, iter: 480000 (13.6166%), loss: 232.39673, 0h 50m 27s (- 5h 20m 5s) 52 | epoch: 13, iter: 490000 (13.9003%), loss: 231.26124, 0h 51m 27s (- 5h 18m 42s) 53 | epoch: 14, iter: 500000 (14.1840%), loss: 147.38845, 0h 52m 27s (- 5h 17m 23s) 54 | epoch: 14, iter: 510000 (14.4677%), loss: 228.44521, 0h 53m 27s (- 5h 16m 1s) 55 | epoch: 14, iter: 520000 (14.7514%), loss: 228.46959, 0h 54m 29s (- 5h 14m 53s) 56 | epoch: 15, iter: 530000 (15.0350%), loss: 27.49651, 0h 55m 29s (- 5h 13m 33s) 57 | epoch: 15, iter: 540000 (15.3187%), loss: 225.34432, 0h 56m 29s (- 5h 12m 18s) 58 | epoch: 15, iter: 550000 (15.6024%), loss: 225.72398, 0h 57m 32s (- 5h 11m 16s) 59 | epoch: 15, iter: 560000 (15.8861%), loss: 225.67404, 0h 58m 32s (- 5h 9m 59s) 60 | epoch: 16, iter: 570000 (16.1698%), loss: 132.21409, 0h 59m 34s (- 5h 8m 50s) 61 | epoch: 16, iter: 580000 (16.4534%), loss: 222.80537, 1h 0m 34s (- 5h 7m 37s) 62 | epoch: 16, iter: 590000 (16.7371%), loss: 223.86467, 1h 1m 36s (- 5h 6m 30s) 63 | epoch: 17, iter: 600000 (17.0208%), loss: 16.66648, 1h 2m 38s (- 5h 5m 25s) 64 | epoch: 17, iter: 610000 (17.3045%), loss: 218.38828, 1h 3m 39s (- 5h 4m 12s) 65 | epoch: 17, iter: 620000 (17.5882%), loss: 220.96036, 1h 4m 42s (- 5h 3m 13s) 66 | epoch: 17, iter: 630000 (17.8718%), loss: 221.90386, 1h 5m 46s (- 5h 2m 14s) 67 | epoch: 18, iter: 640000 (18.1555%), loss: 118.85014, 1h 6m 47s (- 5h 1m 6s) 68 | epoch: 18, iter: 650000 (18.4392%), loss: 218.49976, 1h 7m 49s (- 5h 0m 1s) 69 | epoch: 18, iter: 660000 (18.7229%), loss: 219.08425, 1h 8m 54s (- 4h 59m 9s) 70 | epoch: 19, iter: 670000 (19.0066%), loss: 4.87612, 1h 9m 57s (- 4h 58m 8s) 71 | epoch: 19, iter: 680000 (19.2902%), loss: 215.69654, 1h 11m 1s (- 4h 57m 8s) 72 | epoch: 19, iter: 690000 (19.5739%), loss: 217.48698, 1h 12m 5s (- 4h 56m 14s) 73 | epoch: 19, iter: 700000 (19.8576%), loss: 217.26550, 1h 13m 7s (- 4h 55m 8s) 74 | epoch: 20, iter: 710000 (20.1413%), loss: 106.15122, 1h 14m 8s (- 4h 53m 57s) 75 | epoch: 20, iter: 720000 (20.4250%), loss: 215.69722, 1h 15m 10s (- 4h 52m 54s) 76 | epoch: 20, iter: 730000 (20.7086%), loss: 214.81781, 1h 16m 14s (- 4h 51m 53s) 77 | epoch: 20, iter: 740000 (20.9923%), loss: 215.54455, 1h 17m 15s (- 4h 50m 46s) 78 | epoch: 21, iter: 750000 (21.2760%), loss: 205.78288, 1h 18m 17s (- 4h 49m 40s) 79 | epoch: 21, iter: 760000 (21.5597%), loss: 212.94497, 1h 19m 18s (- 4h 48m 34s) 80 | epoch: 21, iter: 770000 (21.8434%), loss: 215.04291, 1h 20m 20s (- 4h 47m 27s) 81 | epoch: 22, iter: 780000 (22.1270%), loss: 94.40948, 1h 21m 22s (- 4h 46m 23s) 82 | epoch: 22, iter: 790000 (22.4107%), loss: 211.39550, 1h 22m 25s (- 4h 45m 20s) 83 | epoch: 22, iter: 800000 (22.6944%), loss: 211.91150, 1h 23m 25s (- 4h 44m 11s) 84 | epoch: 22, iter: 810000 (22.9781%), loss: 212.92126, 1h 24m 29s (- 4h 43m 12s) 85 | epoch: 23, iter: 820000 (23.2618%), loss: 193.75593, 1h 25m 41s (- 4h 42m 41s) 86 | epoch: 23, iter: 830000 (23.5454%), loss: 210.32490, 1h 26m 45s (- 4h 41m 41s) 87 | epoch: 23, iter: 840000 (23.8291%), loss: 210.28689, 1h 27m 51s (- 4h 40m 50s) 88 | epoch: 24, iter: 850000 (24.1128%), loss: 82.83581, 1h 28m 58s (- 4h 40m 1s) 89 | epoch: 24, iter: 860000 (24.3965%), loss: 208.82878, 1h 30m 11s (- 4h 39m 29s) 90 | epoch: 24, iter: 870000 (24.6802%), loss: 209.46116, 1h 31m 17s (- 4h 38m 36s) 91 | epoch: 24, iter: 880000 (24.9638%), loss: 209.11254, 1h 32m 24s (- 4h 37m 46s) 92 | epoch: 25, iter: 890000 (25.2475%), loss: 179.79879, 1h 33m 30s (- 4h 36m 50s) 93 | epoch: 25, iter: 900000 (25.5312%), loss: 208.10891, 1h 34m 34s (- 4h 35m 50s) 94 | epoch: 25, iter: 910000 (25.8149%), loss: 208.86848, 1h 35m 40s (- 4h 34m 55s) 95 | epoch: 26, iter: 920000 (26.0986%), loss: 71.40685, 1h 36m 45s (- 4h 34m 0s) 96 | epoch: 26, iter: 930000 (26.3822%), loss: 205.39371, 1h 37m 48s (- 4h 32m 56s) 97 | epoch: 26, iter: 940000 (26.6659%), loss: 206.74949, 1h 38m 53s (- 4h 31m 58s) 98 | epoch: 26, iter: 950000 (26.9496%), loss: 208.45557, 1h 39m 56s (- 4h 30m 53s) 99 | epoch: 27, iter: 960000 (27.2333%), loss: 168.39441, 1h 40m 58s (- 4h 29m 48s) 100 | epoch: 27, iter: 970000 (27.5169%), loss: 205.84451, 1h 42m 1s (- 4h 28m 45s) 101 | epoch: 27, iter: 980000 (27.8006%), loss: 206.20411, 1h 43m 5s (- 4h 27m 43s) 102 | epoch: 28, iter: 990000 (28.0843%), loss: 60.06616, 1h 44m 10s (- 4h 26m 45s) 103 | epoch: 28, iter: 1000000 (28.3680%), loss: 203.89503, 1h 45m 19s (- 4h 25m 58s) 104 | epoch: 28, iter: 1010000 (28.6517%), loss: 204.76428, 1h 46m 28s (- 4h 25m 8s) 105 | epoch: 28, iter: 1020000 (28.9353%), loss: 206.33285, 1h 47m 42s (- 4h 24m 31s) 106 | epoch: 29, iter: 1030000 (29.2190%), loss: 156.33513, 1h 49m 25s (- 4h 25m 3s) 107 | epoch: 29, iter: 1040000 (29.5027%), loss: 202.96980, 1h 50m 39s (- 4h 24m 24s) 108 | epoch: 29, iter: 1050000 (29.7864%), loss: 204.89602, 1h 51m 43s (- 4h 23m 22s) 109 | epoch: 30, iter: 1060000 (30.0701%), loss: 49.88182, 1h 52m 48s (- 4h 22m 20s) 110 | epoch: 30, iter: 1070000 (30.3537%), loss: 201.73471, 1h 53m 51s (- 4h 21m 15s) 111 | epoch: 30, iter: 1080000 (30.6374%), loss: 203.28619, 1h 54m 57s (- 4h 20m 14s) 112 | epoch: 30, iter: 1090000 (30.9211%), loss: 203.50570, 1h 56m 0s (- 4h 19m 9s) 113 | epoch: 31, iter: 1100000 (31.2048%), loss: 145.25405, 1h 57m 2s (- 4h 18m 2s) 114 | epoch: 31, iter: 1110000 (31.4885%), loss: 202.30928, 1h 58m 4s (- 4h 16m 54s) 115 | epoch: 31, iter: 1120000 (31.7721%), loss: 202.39322, 1h 59m 5s (- 4h 15m 45s) 116 | epoch: 32, iter: 1130000 (32.0558%), loss: 38.98377, 2h 0m 8s (- 4h 14m 39s) 117 | epoch: 32, iter: 1140000 (32.3395%), loss: 200.49852, 2h 1m 11s (- 4h 13m 32s) 118 | epoch: 32, iter: 1150000 (32.6232%), loss: 201.25514, 2h 2m 13s (- 4h 12m 25s) 119 | epoch: 32, iter: 1160000 (32.9069%), loss: 202.17546, 2h 3m 17s (- 4h 11m 22s) 120 | epoch: 33, iter: 1170000 (33.1905%), loss: 134.48769, 2h 4m 22s (- 4h 10m 20s) 121 | epoch: 33, iter: 1180000 (33.4742%), loss: 199.38152, 2h 5m 24s (- 4h 9m 14s) 122 | epoch: 33, iter: 1190000 (33.7579%), loss: 201.35127, 2h 6m 26s (- 4h 8m 7s) 123 | epoch: 34, iter: 1200000 (34.0416%), loss: 29.18051, 2h 7m 28s (- 4h 7m 0s) 124 | epoch: 34, iter: 1210000 (34.3253%), loss: 198.56807, 2h 8m 31s (- 4h 5m 53s) 125 | epoch: 34, iter: 1220000 (34.6089%), loss: 200.19857, 2h 9m 32s (- 4h 4m 46s) 126 | epoch: 34, iter: 1230000 (34.8926%), loss: 200.78890, 2h 10m 34s (- 4h 3m 37s) 127 | epoch: 35, iter: 1240000 (35.1763%), loss: 123.22386, 2h 11m 35s (- 4h 2m 30s) 128 | epoch: 35, iter: 1250000 (35.4600%), loss: 198.95047, 2h 12m 37s (- 4h 1m 23s) 129 | epoch: 35, iter: 1260000 (35.7437%), loss: 199.50285, 2h 13m 39s (- 4h 0m 16s) 130 | epoch: 36, iter: 1270000 (36.0273%), loss: 19.00897, 2h 14m 42s (- 3h 59m 11s) 131 | epoch: 36, iter: 1280000 (36.3110%), loss: 197.41577, 2h 15m 44s (- 3h 58m 4s) 132 | epoch: 36, iter: 1290000 (36.5947%), loss: 199.19375, 2h 16m 47s (- 3h 57m 1s) 133 | epoch: 36, iter: 1300000 (36.8784%), loss: 198.93374, 2h 17m 52s (- 3h 56m 0s) 134 | epoch: 37, iter: 1310000 (37.1621%), loss: 111.87777, 2h 18m 57s (- 3h 54m 57s) 135 | epoch: 37, iter: 1320000 (37.4457%), loss: 198.36039, 2h 20m 0s (- 3h 53m 52s) 136 | epoch: 37, iter: 1330000 (37.7294%), loss: 198.38075, 2h 21m 4s (- 3h 52m 49s) 137 | epoch: 38, iter: 1340000 (38.0131%), loss: 9.10398, 2h 22m 5s (- 3h 51m 41s) 138 | epoch: 38, iter: 1350000 (38.2968%), loss: 197.34802, 2h 23m 8s (- 3h 50m 37s) 139 | epoch: 38, iter: 1360000 (38.5805%), loss: 197.36508, 2h 24m 9s (- 3h 49m 29s) 140 | epoch: 38, iter: 1370000 (38.8641%), loss: 197.52020, 2h 25m 13s (- 3h 48m 26s) 141 | epoch: 39, iter: 1380000 (39.1478%), loss: 102.50339, 2h 26m 14s (- 3h 47m 19s) 142 | epoch: 39, iter: 1390000 (39.4315%), loss: 195.94070, 2h 27m 15s (- 3h 46m 11s) 143 | epoch: 39, iter: 1400000 (39.7152%), loss: 197.69650, 2h 28m 35s (- 3h 45m 32s) 144 | epoch: 39, iter: 1410000 (39.9989%), loss: 197.20569, 2h 29m 57s (- 3h 44m 56s) 145 | epoch: 40, iter: 1420000 (40.2825%), loss: 194.08130, 2h 31m 22s (- 3h 44m 23s) 146 | epoch: 40, iter: 1430000 (40.5662%), loss: 196.66872, 2h 32m 44s (- 3h 43m 46s) 147 | epoch: 40, iter: 1440000 (40.8499%), loss: 197.52420, 2h 34m 7s (- 3h 43m 10s) 148 | epoch: 41, iter: 1450000 (41.1336%), loss: 91.28728, 2h 35m 31s (- 3h 42m 34s) 149 | epoch: 41, iter: 1460000 (41.4173%), loss: 195.71225, 2h 36m 55s (- 3h 41m 57s) 150 | epoch: 41, iter: 1470000 (41.7009%), loss: 196.67277, 2h 38m 19s (- 3h 41m 21s) 151 | epoch: 41, iter: 1480000 (41.9846%), loss: 196.16795, 2h 39m 39s (- 3h 40m 37s) 152 | epoch: 42, iter: 1490000 (42.2683%), loss: 183.53463, 2h 41m 3s (- 3h 39m 58s) 153 | epoch: 42, iter: 1500000 (42.5520%), loss: 195.69018, 2h 42m 24s (- 3h 39m 15s) 154 | epoch: 42, iter: 1510000 (42.8357%), loss: 195.52721, 2h 43m 45s (- 3h 38m 32s) 155 | epoch: 43, iter: 1520000 (43.1193%), loss: 81.73509, 2h 45m 6s (- 3h 37m 48s) 156 | epoch: 43, iter: 1530000 (43.4030%), loss: 194.57625, 2h 46m 26s (- 3h 37m 2s) 157 | epoch: 43, iter: 1540000 (43.6867%), loss: 194.90987, 2h 47m 51s (- 3h 36m 21s) 158 | epoch: 43, iter: 1550000 (43.9704%), loss: 195.82993, 2h 49m 12s (- 3h 35m 37s) 159 | epoch: 44, iter: 1560000 (44.2541%), loss: 173.04376, 2h 50m 34s (- 3h 34m 52s) 160 | epoch: 44, iter: 1570000 (44.5377%), loss: 194.38690, 2h 51m 56s (- 3h 34m 6s) 161 | epoch: 44, iter: 1580000 (44.8214%), loss: 194.51451, 2h 53m 16s (- 3h 33m 18s) 162 | epoch: 45, iter: 1590000 (45.1051%), loss: 71.48242, 2h 54m 49s (- 3h 32m 46s) 163 | epoch: 45, iter: 1600000 (45.3888%), loss: 193.77686, 2h 56m 29s (- 3h 32m 21s) 164 | epoch: 45, iter: 1610000 (45.6725%), loss: 193.85211, 2h 58m 9s (- 3h 31m 54s) 165 | epoch: 45, iter: 1620000 (45.9561%), loss: 195.10095, 2h 59m 50s (- 3h 31m 29s) 166 | epoch: 46, iter: 1630000 (46.2398%), loss: 163.05437, 3h 1m 29s (- 3h 31m 0s) 167 | epoch: 46, iter: 1640000 (46.5235%), loss: 193.99109, 3h 3m 4s (- 3h 30m 26s) 168 | epoch: 46, iter: 1650000 (46.8072%), loss: 193.76603, 3h 4m 43s (- 3h 29m 55s) 169 | epoch: 47, iter: 1660000 (47.0909%), loss: 60.89032, 3h 6m 20s (- 3h 29m 21s) 170 | epoch: 47, iter: 1670000 (47.3745%), loss: 193.86451, 3h 7m 56s (- 3h 28m 45s) 171 | epoch: 47, iter: 1680000 (47.6582%), loss: 193.51411, 3h 9m 32s (- 3h 28m 10s) 172 | epoch: 47, iter: 1690000 (47.9419%), loss: 193.61232, 3h 11m 10s (- 3h 27m 35s) 173 | epoch: 48, iter: 1700000 (48.2256%), loss: 152.46293, 3h 13m 2s (- 3h 27m 15s) 174 | epoch: 48, iter: 1710000 (48.5093%), loss: 192.63297, 3h 14m 57s (- 3h 26m 56s) 175 | epoch: 48, iter: 1720000 (48.7929%), loss: 193.65949, 3h 16m 49s (- 3h 26m 33s) 176 | epoch: 49, iter: 1730000 (49.0766%), loss: 51.62913, 3h 18m 24s (- 3h 25m 52s) 177 | epoch: 49, iter: 1740000 (49.3603%), loss: 192.14340, 3h 19m 58s (- 3h 25m 9s) 178 | epoch: 49, iter: 1750000 (49.6440%), loss: 192.66957, 3h 21m 59s (- 3h 24m 53s) 179 | epoch: 49, iter: 1760000 (49.9277%), loss: 193.33952, 3h 24m 2s (- 3h 24m 37s) 180 | epoch: 50, iter: 1770000 (50.2113%), loss: 142.94142, 3h 26m 6s (- 3h 24m 22s) 181 | epoch: 50, iter: 1780000 (50.4950%), loss: 191.93783, 3h 28m 9s (- 3h 24m 4s) 182 | epoch: 50, iter: 1790000 (50.7787%), loss: 192.62905, 3h 29m 46s (- 3h 23m 20s) 183 | epoch: 51, iter: 1800000 (51.0624%), loss: 41.93212, 3h 31m 11s (- 3h 22m 24s) 184 | epoch: 51, iter: 1810000 (51.3461%), loss: 191.50055, 3h 32m 36s (- 3h 21m 27s) 185 | epoch: 51, iter: 1820000 (51.6297%), loss: 191.93131, 3h 34m 2s (- 3h 20m 32s) 186 | epoch: 51, iter: 1830000 (51.9134%), loss: 192.73529, 3h 35m 27s (- 3h 19m 34s) 187 | epoch: 52, iter: 1840000 (52.1971%), loss: 132.04742, 3h 36m 51s (- 3h 18m 36s) 188 | epoch: 52, iter: 1850000 (52.4808%), loss: 191.65766, 3h 38m 16s (- 3h 17m 38s) 189 | epoch: 52, iter: 1860000 (52.7645%), loss: 192.32697, 3h 39m 40s (- 3h 16m 39s) 190 | epoch: 53, iter: 1870000 (53.0481%), loss: 31.68704, 3h 41m 1s (- 3h 15m 37s) 191 | epoch: 53, iter: 1880000 (53.3318%), loss: 190.24711, 3h 42m 16s (- 3h 14m 29s) 192 | epoch: 53, iter: 1890000 (53.6155%), loss: 191.68136, 3h 43m 41s (- 3h 13m 31s) 193 | epoch: 53, iter: 1900000 (53.8992%), loss: 191.96693, 3h 45m 0s (- 3h 12m 26s) 194 | epoch: 54, iter: 1910000 (54.1829%), loss: 121.97977, 3h 46m 35s (- 3h 11m 36s) 195 | epoch: 54, iter: 1920000 (54.4665%), loss: 191.36788, 3h 47m 57s (- 3h 10m 34s) 196 | epoch: 54, iter: 1930000 (54.7502%), loss: 191.69581, 3h 49m 21s (- 3h 9m 33s) 197 | epoch: 55, iter: 1940000 (55.0339%), loss: 22.47068, 3h 50m 45s (- 3h 8m 32s) 198 | epoch: 55, iter: 1950000 (55.3176%), loss: 189.54484, 3h 52m 8s (- 3h 7m 30s) 199 | epoch: 55, iter: 1960000 (55.6013%), loss: 191.35433, 3h 53m 29s (- 3h 6m 27s) 200 | epoch: 55, iter: 1970000 (55.8849%), loss: 191.31744, 3h 54m 52s (- 3h 5m 24s) 201 | epoch: 56, iter: 1980000 (56.1686%), loss: 112.48494, 3h 56m 16s (- 3h 4m 22s) 202 | epoch: 56, iter: 1990000 (56.4523%), loss: 190.35771, 3h 57m 37s (- 3h 3m 18s) 203 | epoch: 56, iter: 2000000 (56.7360%), loss: 191.57365, 3h 58m 58s (- 3h 2m 13s) 204 | epoch: 57, iter: 2010000 (57.0197%), loss: 13.13259, 4h 0m 21s (- 3h 1m 10s) 205 | epoch: 57, iter: 2020000 (57.3033%), loss: 189.06992, 4h 1m 42s (- 3h 0m 5s) 206 | epoch: 57, iter: 2030000 (57.5870%), loss: 190.45278, 4h 3m 4s (- 2h 59m 1s) 207 | epoch: 57, iter: 2040000 (57.8707%), loss: 191.29769, 4h 4m 28s (- 2h 57m 58s) 208 | epoch: 58, iter: 2050000 (58.1544%), loss: 102.97000, 4h 5m 53s (- 2h 56m 55s) 209 | epoch: 58, iter: 2060000 (58.4381%), loss: 189.96727, 4h 7m 15s (- 2h 55m 51s) 210 | epoch: 58, iter: 2070000 (58.7217%), loss: 189.94243, 4h 8m 37s (- 2h 54m 46s) 211 | epoch: 59, iter: 2080000 (59.0054%), loss: 3.58441, 4h 10m 1s (- 2h 53m 42s) 212 | epoch: 59, iter: 2090000 (59.2891%), loss: 188.07650, 4h 11m 24s (- 2h 52m 38s) 213 | epoch: 59, iter: 2100000 (59.5728%), loss: 189.51323, 4h 12m 48s (- 2h 51m 33s) 214 | epoch: 59, iter: 2110000 (59.8565%), loss: 190.90891, 4h 14m 11s (- 2h 50m 28s) 215 | epoch: 60, iter: 2120000 (60.1401%), loss: 93.07775, 4h 15m 35s (- 2h 49m 24s) 216 | epoch: 60, iter: 2130000 (60.4238%), loss: 189.24776, 4h 16m 57s (- 2h 48m 18s) 217 | epoch: 60, iter: 2140000 (60.7075%), loss: 189.21180, 4h 18m 21s (- 2h 47m 13s) 218 | epoch: 60, iter: 2150000 (60.9912%), loss: 190.79515, 4h 19m 25s (- 2h 45m 55s) 219 | epoch: 61, iter: 2160000 (61.2749%), loss: 182.08201, 4h 20m 27s (- 2h 44m 36s) 220 | epoch: 61, iter: 2170000 (61.5585%), loss: 189.82015, 4h 21m 43s (- 2h 43m 26s) 221 | epoch: 61, iter: 2180000 (61.8422%), loss: 190.10941, 4h 23m 8s (- 2h 42m 21s) 222 | epoch: 62, iter: 2190000 (62.1259%), loss: 82.96230, 4h 24m 27s (- 2h 41m 13s) 223 | epoch: 62, iter: 2200000 (62.4096%), loss: 188.30301, 4h 25m 35s (- 2h 39m 58s) 224 | epoch: 62, iter: 2210000 (62.6933%), loss: 189.52637, 4h 26m 41s (- 2h 38m 41s) 225 | epoch: 62, iter: 2220000 (62.9769%), loss: 190.43668, 4h 27m 46s (- 2h 37m 25s) 226 | epoch: 63, iter: 2230000 (63.2606%), loss: 172.52425, 4h 28m 52s (- 2h 36m 9s) 227 | epoch: 63, iter: 2240000 (63.5443%), loss: 188.84229, 4h 29m 57s (- 2h 34m 52s) 228 | epoch: 63, iter: 2250000 (63.8280%), loss: 189.59928, 4h 31m 2s (- 2h 33m 36s) 229 | epoch: 64, iter: 2260000 (64.1117%), loss: 73.24027, 4h 32m 8s (- 2h 32m 20s) 230 | epoch: 64, iter: 2270000 (64.3953%), loss: 187.56424, 4h 33m 13s (- 2h 31m 4s) 231 | epoch: 64, iter: 2280000 (64.6790%), loss: 189.16940, 4h 34m 19s (- 2h 29m 48s) 232 | epoch: 64, iter: 2290000 (64.9627%), loss: 190.08061, 4h 35m 24s (- 2h 28m 32s) 233 | epoch: 65, iter: 2300000 (65.2464%), loss: 163.61077, 4h 36m 32s (- 2h 27m 18s) 234 | epoch: 65, iter: 2310000 (65.5301%), loss: 188.06103, 4h 37m 49s (- 2h 26m 8s) 235 | epoch: 65, iter: 2320000 (65.8137%), loss: 188.57039, 4h 39m 15s (- 2h 25m 3s) 236 | epoch: 66, iter: 2330000 (66.0974%), loss: 63.89059, 4h 40m 32s (- 2h 23m 53s) 237 | epoch: 66, iter: 2340000 (66.3811%), loss: 187.96240, 4h 41m 55s (- 2h 22m 47s) 238 | epoch: 66, iter: 2350000 (66.6648%), loss: 188.44596, 4h 43m 17s (- 2h 21m 39s) 239 | epoch: 66, iter: 2360000 (66.9485%), loss: 189.03845, 4h 44m 34s (- 2h 20m 29s) 240 | epoch: 67, iter: 2370000 (67.2321%), loss: 153.15918, 4h 45m 46s (- 2h 19m 16s) 241 | epoch: 67, iter: 2380000 (67.5158%), loss: 188.08269, 4h 46m 55s (- 2h 18m 2s) 242 | epoch: 67, iter: 2390000 (67.7995%), loss: 188.58399, 4h 48m 4s (- 2h 16m 49s) 243 | epoch: 68, iter: 2400000 (68.0832%), loss: 54.74805, 4h 49m 14s (- 2h 15m 35s) 244 | epoch: 68, iter: 2410000 (68.3669%), loss: 187.11809, 4h 50m 23s (- 2h 14m 21s) 245 | epoch: 68, iter: 2420000 (68.6505%), loss: 188.76441, 4h 51m 31s (- 2h 13m 7s) 246 | epoch: 68, iter: 2430000 (68.9342%), loss: 188.09942, 4h 52m 39s (- 2h 11m 53s) 247 | epoch: 69, iter: 2440000 (69.2179%), loss: 142.46104, 4h 53m 47s (- 2h 10m 39s) 248 | epoch: 69, iter: 2450000 (69.5016%), loss: 188.07412, 4h 54m 56s (- 2h 9m 25s) 249 | epoch: 69, iter: 2460000 (69.7853%), loss: 188.46198, 4h 56m 4s (- 2h 8m 11s) 250 | epoch: 70, iter: 2470000 (70.0689%), loss: 45.00636, 4h 57m 16s (- 2h 6m 59s) 251 | epoch: 70, iter: 2480000 (70.3526%), loss: 186.25546, 4h 58m 24s (- 2h 5m 45s) 252 | epoch: 70, iter: 2490000 (70.6363%), loss: 187.25174, 4h 59m 30s (- 2h 4m 30s) 253 | epoch: 70, iter: 2500000 (70.9200%), loss: 189.34450, 5h 0m 34s (- 2h 3m 14s) 254 | epoch: 71, iter: 2510000 (71.2037%), loss: 134.05612, 5h 1m 42s (- 2h 2m 1s) 255 | epoch: 71, iter: 2520000 (71.4873%), loss: 186.95486, 5h 2m 48s (- 2h 0m 46s) 256 | epoch: 71, iter: 2530000 (71.7710%), loss: 187.59962, 5h 3m 56s (- 1h 59m 32s) 257 | epoch: 72, iter: 2540000 (72.0547%), loss: 36.19762, 5h 5m 6s (- 1h 58m 19s) 258 | epoch: 72, iter: 2550000 (72.3384%), loss: 186.68176, 5h 6m 16s (- 1h 57m 7s) 259 | epoch: 72, iter: 2560000 (72.6221%), loss: 187.22045, 5h 7m 24s (- 1h 55m 53s) 260 | epoch: 72, iter: 2570000 (72.9057%), loss: 187.31267, 5h 8m 31s (- 1h 54m 39s) 261 | epoch: 73, iter: 2580000 (73.1894%), loss: 124.06747, 5h 9m 39s (- 1h 53m 25s) 262 | epoch: 73, iter: 2590000 (73.4731%), loss: 186.53858, 5h 10m 47s (- 1h 52m 12s) 263 | epoch: 73, iter: 2600000 (73.7568%), loss: 187.53413, 5h 11m 54s (- 1h 50m 58s) 264 | epoch: 74, iter: 2610000 (74.0405%), loss: 26.42638, 5h 13m 3s (- 1h 49m 45s) 265 | epoch: 74, iter: 2620000 (74.3241%), loss: 186.32935, 5h 14m 12s (- 1h 48m 32s) 266 | epoch: 74, iter: 2630000 (74.6078%), loss: 186.82291, 5h 15m 21s (- 1h 47m 19s) 267 | epoch: 74, iter: 2640000 (74.8915%), loss: 187.14992, 5h 16m 29s (- 1h 46m 6s) 268 | epoch: 75, iter: 2650000 (75.1752%), loss: 114.61958, 5h 17m 35s (- 1h 44m 52s) 269 | epoch: 75, iter: 2660000 (75.4589%), loss: 186.03291, 5h 18m 36s (- 1h 43m 37s) 270 | epoch: 75, iter: 2670000 (75.7425%), loss: 187.27173, 5h 19m 40s (- 1h 42m 22s) 271 | epoch: 76, iter: 2680000 (76.0262%), loss: 17.14513, 5h 20m 41s (- 1h 41m 7s) 272 | epoch: 76, iter: 2690000 (76.3099%), loss: 185.77682, 5h 21m 45s (- 1h 39m 53s) 273 | epoch: 76, iter: 2700000 (76.5936%), loss: 186.77826, 5h 22m 45s (- 1h 38m 37s) 274 | epoch: 76, iter: 2710000 (76.8773%), loss: 187.23163, 5h 23m 43s (- 1h 37m 21s) 275 | epoch: 77, iter: 2720000 (77.1609%), loss: 105.15286, 5h 24m 45s (- 1h 36m 7s) 276 | epoch: 77, iter: 2730000 (77.4446%), loss: 186.09209, 5h 25m 44s (- 1h 34m 52s) 277 | epoch: 77, iter: 2740000 (77.7283%), loss: 186.81952, 5h 26m 46s (- 1h 33m 37s) 278 | epoch: 78, iter: 2750000 (78.0120%), loss: 7.82204, 5h 27m 44s (- 1h 32m 22s) 279 | epoch: 78, iter: 2760000 (78.2957%), loss: 185.20877, 5h 28m 46s (- 1h 31m 8s) 280 | epoch: 78, iter: 2770000 (78.5793%), loss: 186.83001, 5h 30m 11s (- 1h 30m 0s) 281 | epoch: 78, iter: 2780000 (78.8630%), loss: 186.80129, 5h 31m 38s (- 1h 28m 53s) 282 | epoch: 79, iter: 2790000 (79.1467%), loss: 95.57875, 5h 33m 10s (- 1h 27m 47s) 283 | epoch: 79, iter: 2800000 (79.4304%), loss: 185.83061, 5h 34m 39s (- 1h 26m 39s) 284 | epoch: 79, iter: 2810000 (79.7141%), loss: 186.88767, 5h 36m 9s (- 1h 25m 32s) 285 | epoch: 79, iter: 2820000 (79.9977%), loss: 186.91404, 5h 37m 39s (- 1h 24m 25s) 286 | epoch: 80, iter: 2830000 (80.2814%), loss: 183.73249, 5h 39m 10s (- 1h 23m 18s) 287 | epoch: 80, iter: 2840000 (80.5651%), loss: 185.67632, 5h 40m 41s (- 1h 22m 11s) 288 | epoch: 80, iter: 2850000 (80.8488%), loss: 187.01350, 5h 42m 11s (- 1h 21m 3s) 289 | epoch: 81, iter: 2860000 (81.1325%), loss: 86.17994, 5h 43m 42s (- 1h 19m 55s) 290 | epoch: 81, iter: 2870000 (81.4161%), loss: 185.69051, 5h 45m 11s (- 1h 18m 47s) 291 | epoch: 81, iter: 2880000 (81.6998%), loss: 186.18324, 5h 46m 41s (- 1h 17m 39s) 292 | epoch: 81, iter: 2890000 (81.9835%), loss: 186.92896, 5h 48m 11s (- 1h 16m 30s) 293 | epoch: 82, iter: 2900000 (82.2672%), loss: 173.45726, 5h 49m 40s (- 1h 15m 22s) 294 | epoch: 82, iter: 2910000 (82.5508%), loss: 186.54825, 5h 51m 10s (- 1h 14m 13s) 295 | epoch: 82, iter: 2920000 (82.8345%), loss: 185.77585, 5h 52m 43s (- 1h 13m 5s) 296 | epoch: 83, iter: 2930000 (83.1182%), loss: 76.42856, 5h 54m 18s (- 1h 11m 57s) 297 | epoch: 83, iter: 2940000 (83.4019%), loss: 185.42695, 5h 55m 47s (- 1h 10m 48s) 298 | epoch: 83, iter: 2950000 (83.6856%), loss: 186.75486, 5h 57m 15s (- 1h 9m 38s) 299 | epoch: 83, iter: 2960000 (83.9692%), loss: 186.05918, 5h 58m 48s (- 1h 8m 29s) 300 | epoch: 84, iter: 2970000 (84.2529%), loss: 165.03357, 6h 0m 28s (- 1h 7m 22s) 301 | epoch: 84, iter: 2980000 (84.5366%), loss: 185.36279, 6h 2m 6s (- 1h 6m 14s) 302 | epoch: 84, iter: 2990000 (84.8203%), loss: 185.68220, 6h 3m 38s (- 1h 5m 4s) 303 | epoch: 85, iter: 3000000 (85.1040%), loss: 67.56534, 6h 4m 57s (- 1h 3m 52s) 304 | epoch: 85, iter: 3010000 (85.3876%), loss: 185.08859, 6h 6m 8s (- 1h 2m 39s) 305 | epoch: 85, iter: 3020000 (85.6713%), loss: 185.98866, 6h 7m 18s (- 1h 1m 25s) 306 | epoch: 85, iter: 3030000 (85.9550%), loss: 186.04008, 6h 8m 30s (- 1h 0m 12s) 307 | epoch: 86, iter: 3040000 (86.2387%), loss: 154.25655, 6h 9m 43s (- 0h 58m 59s) 308 | epoch: 86, iter: 3050000 (86.5224%), loss: 186.40076, 6h 10m 54s (- 0h 57m 46s) 309 | epoch: 86, iter: 3060000 (86.8060%), loss: 186.47730, 6h 12m 6s (- 0h 56m 33s) 310 | epoch: 87, iter: 3070000 (87.0897%), loss: 58.51510, 6h 13m 16s (- 0h 55m 20s) 311 | epoch: 87, iter: 3080000 (87.3734%), loss: 184.87556, 6h 14m 27s (- 0h 54m 6s) 312 | epoch: 87, iter: 3090000 (87.6571%), loss: 184.98283, 6h 15m 37s (- 0h 52m 53s) 313 | epoch: 87, iter: 3100000 (87.9408%), loss: 186.32554, 6h 16m 47s (- 0h 51m 40s) 314 | epoch: 88, iter: 3110000 (88.2244%), loss: 145.56026, 6h 17m 58s (- 0h 50m 26s) 315 | epoch: 88, iter: 3120000 (88.5081%), loss: 185.61519, 6h 19m 8s (- 0h 49m 13s) 316 | epoch: 88, iter: 3130000 (88.7918%), loss: 185.73383, 6h 20m 17s (- 0h 48m 0s) 317 | epoch: 89, iter: 3140000 (89.0755%), loss: 49.11769, 6h 21m 33s (- 0h 46m 47s) 318 | epoch: 89, iter: 3150000 (89.3592%), loss: 184.29239, 6h 22m 44s (- 0h 45m 34s) 319 | epoch: 89, iter: 3160000 (89.6428%), loss: 185.69499, 6h 23m 56s (- 0h 44m 21s) 320 | epoch: 89, iter: 3170000 (89.9265%), loss: 185.43642, 6h 25m 7s (- 0h 43m 8s) 321 | epoch: 90, iter: 3180000 (90.2102%), loss: 136.36600, 6h 26m 20s (- 0h 41m 55s) 322 | epoch: 90, iter: 3190000 (90.4939%), loss: 184.45063, 6h 27m 36s (- 0h 40m 43s) 323 | epoch: 90, iter: 3200000 (90.7776%), loss: 185.33140, 6h 28m 35s (- 0h 39m 28s) 324 | epoch: 91, iter: 3210000 (91.0612%), loss: 39.55210, 6h 29m 35s (- 0h 38m 14s) 325 | epoch: 91, iter: 3220000 (91.3449%), loss: 184.63664, 6h 30m 36s (- 0h 37m 0s) 326 | epoch: 91, iter: 3230000 (91.6286%), loss: 184.75182, 6h 31m 36s (- 0h 35m 46s) 327 | epoch: 91, iter: 3240000 (91.9123%), loss: 185.16648, 6h 32m 35s (- 0h 34m 32s) 328 | epoch: 92, iter: 3250000 (92.1960%), loss: 127.44406, 6h 33m 34s (- 0h 33m 18s) 329 | epoch: 92, iter: 3260000 (92.4796%), loss: 184.26087, 6h 34m 36s (- 0h 32m 5s) 330 | epoch: 92, iter: 3270000 (92.7633%), loss: 184.61929, 6h 35m 36s (- 0h 30m 51s) 331 | epoch: 93, iter: 3280000 (93.0470%), loss: 30.61833, 6h 36m 35s (- 0h 29m 38s) 332 | epoch: 93, iter: 3290000 (93.3307%), loss: 183.67716, 6h 37m 34s (- 0h 28m 24s) 333 | epoch: 93, iter: 3300000 (93.6144%), loss: 184.60973, 6h 38m 32s (- 0h 27m 11s) 334 | epoch: 93, iter: 3310000 (93.8980%), loss: 185.73265, 6h 39m 34s (- 0h 25m 57s) 335 | epoch: 94, iter: 3320000 (94.1817%), loss: 117.35624, 6h 40m 36s (- 0h 24m 44s) 336 | epoch: 94, iter: 3330000 (94.4654%), loss: 183.38167, 6h 41m 36s (- 0h 23m 31s) 337 | epoch: 94, iter: 3340000 (94.7491%), loss: 184.96043, 6h 42m 38s (- 0h 22m 18s) 338 | epoch: 95, iter: 3350000 (95.0328%), loss: 21.26559, 6h 43m 43s (- 0h 21m 6s) 339 | epoch: 95, iter: 3360000 (95.3164%), loss: 183.66279, 6h 44m 42s (- 0h 19m 53s) 340 | epoch: 95, iter: 3370000 (95.6001%), loss: 184.92417, 6h 45m 41s (- 0h 18m 40s) 341 | epoch: 95, iter: 3380000 (95.8838%), loss: 185.02349, 6h 46m 44s (- 0h 17m 27s) 342 | epoch: 96, iter: 3390000 (96.1675%), loss: 108.06199, 6h 47m 44s (- 0h 16m 14s) 343 | epoch: 96, iter: 3400000 (96.4512%), loss: 183.88079, 6h 48m 44s (- 0h 15m 2s) 344 | epoch: 96, iter: 3410000 (96.7348%), loss: 184.99662, 6h 49m 47s (- 0h 13m 49s) 345 | epoch: 97, iter: 3420000 (97.0185%), loss: 12.03456, 6h 50m 48s (- 0h 12m 37s) 346 | epoch: 97, iter: 3430000 (97.3022%), loss: 183.98048, 6h 51m 49s (- 0h 11m 25s) 347 | epoch: 97, iter: 3440000 (97.5859%), loss: 184.29520, 6h 52m 48s (- 0h 10m 12s) 348 | epoch: 97, iter: 3450000 (97.8696%), loss: 185.03215, 6h 53m 50s (- 0h 9m 0s) 349 | epoch: 98, iter: 3460000 (98.1532%), loss: 99.17649, 6h 54m 52s (- 0h 7m 48s) 350 | epoch: 98, iter: 3470000 (98.4369%), loss: 183.97261, 6h 55m 51s (- 0h 6m 36s) 351 | epoch: 98, iter: 3480000 (98.7206%), loss: 184.93565, 6h 56m 53s (- 0h 5m 24s) 352 | epoch: 99, iter: 3490000 (99.0043%), loss: 2.76205, 6h 57m 52s (- 0h 4m 12s) 353 | epoch: 99, iter: 3500000 (99.2880%), loss: 183.49325, 6h 58m 51s (- 0h 3m 0s) 354 | epoch: 99, iter: 3510000 (99.5716%), loss: 183.66532, 6h 59m 55s (- 0h 1m 48s) 355 | epoch: 99, iter: 3520000 (99.8553%), loss: 185.75300, 7h 0m 58s (- 0h 0m 36s) 356 | -------------------------------------------------------------------------------- /glove/model/GloveModel.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.nn as nn #神经网络工具箱torch.nn 3 | import torch.nn.functional as F #神经网络函数torch.nn.functional 4 | import numpy as np 5 | import sys 6 | import math 7 | 8 | class GloveModel(nn.Module): 9 | def __init__(self, vocab_size, embed_size): 10 | super().__init__() 11 | self.vocab_size = vocab_size 12 | self.embed_size = embed_size 13 | 14 | #声明v和w为Embedding向量 15 | self.v = nn.Embedding(vocab_size, embed_size) 16 | self.w = nn.Embedding(vocab_size, embed_size) 17 | self.biasv = nn.Embedding(vocab_size, 1) 18 | self.biasw = nn.Embedding(vocab_size, 1) 19 | 20 | #随机初始化参数 21 | initrange = 0.5 / self.embed_size 22 | self.v.weight.data.uniform_(-initrange, initrange) 23 | self.w.weight.data.uniform_(-initrange, initrange) 24 | 25 | def forward(self, i, j, co_occur, weight): 26 | vi = self.v(i) 27 | wj = self.w(j) 28 | bi = self.biasv(i) 29 | bj = self.biasw(j) 30 | 31 | similarity = torch.mul(vi, wj) 32 | similarity = torch.sum(similarity, dim=1) 33 | 34 | loss = similarity + bi + bj - torch.log(co_occur) 35 | loss = 0.5 * weight * loss * loss 36 | 37 | return loss.sum().mean() 38 | 39 | def gloveMatrix(self): 40 | ''' 41 | 获得词向量，这里把两个向量相加作为最后的词向量 42 | :return: 43 | ''' 44 | return self.v.weight.data.numpy() + self.w.weight.data.numpy() 45 | -------------------------------------------------------------------------------- /glove/model/GloveTrain.py: -------------------------------------------------------------------------------- 1 | import torch 2 | import torch.utils.data as tud #Pytorch读取训练集需要用到torch.utils.data类 3 | 4 | from collections import Counter 5 | from sklearn.metrics.pairwise import cosine_similarity 6 | 7 | import pandas as pd 8 | import numpy as np 9 | import scipy 10 | 11 | import time 12 | import math 13 | import random 14 | import sys 15 | import matplotlib.pyplot as plt 16 | 17 | from WordEmbeddingDataset import WordEmbeddingDataset 18 | from GloveModel import GloveModel 19 | 20 | EMBEDDING_SIZE = 50 21 | MAX_VOCAB_SIZE = 2000 22 | WINDOW_SIZE = 5 23 | 24 | NUM_EPOCHS = 10 25 | BATCH_SIZE = 10 26 | LEARNING_RATE = 0.05 27 | 28 | TEXT_SIZE = 20000000 29 | LOG_FILE = "../logs/glove-{}.log".format(EMBEDDING_SIZE) 30 | WEIGHT_FILE = "../weights/glove-{}.th".format(EMBEDDING_SIZE) 31 | 32 | def getCorpus(filetype, size): 33 | if filetype == 'dev': 34 | filepath = '../corpus/text8.dev.txt' 35 | elif filetype == 'test': 36 | filepath = '../corpus/text8.test.txt' 37 | else: 38 | filepath = '../corpus/text8.train.txt' 39 | 40 | with open(filepath, "r") as f: 41 | text = f.read() 42 | text = text.lower().split() 43 | text = text[: min(len(text), size)] 44 | vocab_dict = dict(Counter(text).most_common(MAX_VOCAB_SIZE - 1)) 45 | vocab_dict[''] = len(text) - sum(list(vocab_dict.values())) 46 | idx_to_word = list(vocab_dict.keys()) 47 | word_to_idx = {word:ind for ind, word in enumerate(idx_to_word)} 48 | word_counts = np.array(list(vocab_dict.values()), dtype=np.float32) 49 | word_freqs = word_counts / sum(word_counts) 50 | print("Words list length:{}".format(len(text))) 51 | print("Vocab size:{}".format(len(idx_to_word))) 52 | return text, idx_to_word, word_to_idx, word_counts, word_freqs 53 | 54 | def buildCooccuranceMatrix(text, word_to_idx): 55 | vocab_size = len(word_to_idx) 56 | maxlength = len(text) 57 | text_ids = [word_to_idx.get(word, word_to_idx[""]) for word in text] 58 | cooccurance_matrix = np.zeros((vocab_size, vocab_size), dtype=np.int32) 59 | print("Co-Matrix consumed mem:%.2fMB" % (sys.getsizeof(cooccurance_matrix)/(1024*1024))) 60 | for i, center_word_id in enumerate(text_ids): 61 | window_indices = list(range(i - WINDOW_SIZE, i)) + list(range(i + 1, i + WINDOW_SIZE + 1)) 62 | window_indices = [i % maxlength for i in window_indices] 63 | window_word_ids = [text_ids[index] for index in window_indices] 64 | for context_word_id in window_word_ids: 65 | cooccurance_matrix[center_word_id][context_word_id] += 1 66 | if (i+1) % 1000000 == 0: 67 | print(">>>>> Process %dth word" % (i+1)) 68 | print(">>>>> Build co-occurance matrix completed.") 69 | return cooccurance_matrix 70 | 71 | def buildWeightMatrix(co_matrix): 72 | xmax = 100.0 73 | weight_matrix = np.zeros_like(co_matrix, dtype=np.float32) 74 | print("Weight-Matrix consumed mem:%.2fMB" % (sys.getsizeof(weight_matrix) / (1024 * 1024))) 75 | for i in range(co_matrix.shape[0]): 76 | for j in range(co_matrix.shape[1]): 77 | weight_matrix[i][j] = math.pow(co_matrix[i][j] / xmax, 0.75) if co_matrix[i][j] < xmax else 1 78 | if (i+1) % 1000 == 0: 79 | print(">>>>> Process %dth weight" % (i+1)) 80 | print(">>>>> Build weight matrix completed.") 81 | return weight_matrix 82 | 83 | def find_nearest(word, embedding_weights): 84 | index = word_to_idx[word] 85 | embedding = embedding_weights[index] 86 | cos_dis = np.array([scipy.spatial.distance.cosine(e, embedding) for e in embedding_weights]) 87 | return [idx_to_word[i] for i in cos_dis.argsort()[:10]] 88 | 89 | def asMinutes(s): 90 | h = math.floor(s / 3600) 91 | s = s - h * 3600 92 | m = math.floor(s / 60) 93 | s -= m * 60 94 | return '%dh %dm %ds' % (h, m, s) 95 | 96 | def timeSince(since, percent): 97 | now = time.time() 98 | s = now - since 99 | es = s / percent 100 | rs = es - s 101 | return '%s (- %s)' % (asMinutes(s), asMinutes(rs)) 102 | 103 | def loadModel(): 104 | path = WEIGHT_FILE 105 | model = GloveModel(MAX_VOCAB_SIZE, EMBEDDING_SIZE) 106 | model.load_state_dict(torch.load(path)) 107 | return model 108 | 109 | def findRelationshipVector(word1, word2, word3): 110 | word1_idx = word_to_idx[word1] 111 | word2_idx = word_to_idx[word2] 112 | word3_idx = word_to_idx[word3] 113 | embedding = glove_matrix[word2_idx] - glove_matrix[word1_idx] + glove_matrix[word3_idx] 114 | cos_dis = np.array([scipy.spatial.distance.cosine(e, embedding) for e in glove_matrix]) 115 | for i in cos_dis.argsort()[:5]: 116 | print("{} to {} as {} to {}".format(word1, word2, word3, idx_to_word[i])) 117 | 118 | if __name__ == '__main__': 119 | text, idx_to_word, word_to_idx, word_counts, word_freqs = getCorpus('train', size=TEXT_SIZE) #加载语料及预处理 120 | co_matrix = buildCooccuranceMatrix(text, word_to_idx) #构建共现矩阵 121 | weight_matrix = buildWeightMatrix(co_matrix) #构建权重矩阵 122 | dataset = WordEmbeddingDataset(co_matrix, weight_matrix) #创建dataset 123 | dataloader = tud.DataLoader(dataset, batch_size=BATCH_SIZE, shuffle=True, num_workers=0) 124 | model = GloveModel(MAX_VOCAB_SIZE, EMBEDDING_SIZE) #创建模型 125 | #model = loadModel() 126 | optimizer = torch.optim.Adagrad(model.parameters(), lr=LEARNING_RATE) #选择Adagrad优化器 127 | 128 | 129 | print_every = 10000 130 | save_every = 50000 131 | epochs = NUM_EPOCHS 132 | iters_per_epoch = int(dataset.__len__() / BATCH_SIZE) 133 | total_iterations = iters_per_epoch * epochs 134 | print("Iterations: %d per one epoch, Total iterations: %d " % (iters_per_epoch, total_iterations)) 135 | 136 | start = time.time() 137 | for epoch in range(epochs): 138 | loss_print_avg = 0 139 | iteration = iters_per_epoch * epoch 140 | for i, j, co_occur, weight in dataloader: 141 | iteration += 1 142 | optimizer.zero_grad() #每一批样本训练前重置缓存的梯度 143 | loss = model(i, j, co_occur, weight) #前向传播 144 | loss.backward() #反向传播 145 | optimizer.step() #更新梯度 146 | loss_print_avg += loss.item() 147 | 148 | if iteration % print_every == 0: 149 | time_desc = timeSince(start, iteration / total_iterations) 150 | iter_percent = iteration / total_iterations * 100 151 | loss_avg = loss_print_avg / print_every 152 | loss_print_avg = 0 153 | with open(LOG_FILE, "a") as fout: 154 | fout.write("epoch: %d, iter: %d (%.4f%%), loss: %.5f, %s\n" % 155 | (epoch, iteration, iter_percent, loss_avg, time_desc)) 156 | print("epoch: %d, iter: %d/%d (%.4f%%), loss: %.5f, %s" % 157 | (epoch, iteration, total_iterations, iter_percent, loss_avg, time_desc)) 158 | if iteration % save_every == 0: 159 | torch.save(model.state_dict(), WEIGHT_FILE) 160 | torch.save(model.state_dict(), WEIGHT_FILE) 161 | 162 | glove_matrix = model.gloveMatrix() 163 | for word in ["good", "one", "green", "like", "america", "queen", "better", "paris", "work", "computer", "language"]: 164 | print(word, find_nearest(word, glove_matrix)) 165 | findRelationshipVector('man', 'king', 'woman') 166 | findRelationshipVector('america', 'washington', 'france') 167 | findRelationshipVector('good', 'better', 'little') 168 | 169 | #数据降维以及可视化 170 | candidate_words = ['one','two','three','four','five','six','seven','eight','night','ten','color','green','blue','red','black', 171 | 'man','woman','king','queen','wife','son','daughter','brown','zero','computer','hardware','software','system','program', 172 | 'america','china','france','washington','good','better','bad'] 173 | candidate_indexes = [word_to_idx[word] for word in candidate_words] 174 | choosen_indexes = candidate_indexes 175 | choosen_vectors = [glove_matrix[index] for index in choosen_indexes] 176 | 177 | U, S, VH = np.linalg.svd(choosen_vectors, full_matrices=False) 178 | for i in range(len(choosen_indexes)): 179 | plt.text(U[i, 0], U[i, 1], idx_to_word[choosen_indexes[i]]) 180 | 181 | coordinate = U[:, 0:2] 182 | plt.xlim((np.min(coordinate[:, 0]) - 0.1, np.max(coordinate[:, 0]) + 0.1)) 183 | plt.ylim((np.min(coordinate[:, 1]) - 0.1, np.max(coordinate[:, 1]) + 0.1)) 184 | plt.show() -------------------------------------------------------------------------------- /glove/model/WordEmbeddingDataset.py: -------------------------------------------------------------------------------- 1 | import torch.utils.data as tud 2 | import torch 3 | 4 | class WordEmbeddingDataset(tud.Dataset): 5 | def __init__(self, co_matrix, weight_matrix): 6 | self.co_matrix = co_matrix 7 | self.weight_matrix = weight_matrix 8 | self.train_set = [] 9 | 10 | for i in range(self.weight_matrix.shape[0]): 11 | for j in range(self.weight_matrix.shape[1]): 12 | if weight_matrix[i][j] != 0: 13 | # 这里对权重进行了筛选，去掉权重为0的项 14 | # 因为共现次数为0会导致log(X)变成nan 15 | self.train_set.append((i, j)) 16 | 17 | def __len__(self): 18 | ''' 19 | 必须重写的方法 20 | :return: 返回训练集的大小 21 | ''' 22 | return len(self.train_set) 23 | 24 | def __getitem__(self, index): 25 | ''' 26 | 必须重写的方法 27 | :param index:样本索引 28 | :return: 返回一个样本 29 | ''' 30 | (i, j) = self.train_set[index] 31 | return i, j, torch.tensor(self.co_matrix[i][j], dtype=torch.float), self.weight_matrix[i][j] 32 | -------------------------------------------------------------------------------- /glove/model/__pycache__/GloveModel.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/model/__pycache__/GloveModel.cpython-36.pyc -------------------------------------------------------------------------------- /glove/model/__pycache__/GloveModel.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/model/__pycache__/GloveModel.cpython-37.pyc -------------------------------------------------------------------------------- /glove/model/__pycache__/WordEmbeddingDataset.cpython-36.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/model/__pycache__/WordEmbeddingDataset.cpython-36.pyc -------------------------------------------------------------------------------- /glove/model/__pycache__/WordEmbeddingDataset.cpython-37.pyc: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/model/__pycache__/WordEmbeddingDataset.cpython-37.pyc -------------------------------------------------------------------------------- /glove/model/record-50.log: -------------------------------------------------------------------------------- 1 | nohup: ignoring input 2 | -------------------------------------------------------------------------------- /glove/weights/glove-50-1.th: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/weights/glove-50-1.th -------------------------------------------------------------------------------- /glove/weights/glove-50-2.th: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/weights/glove-50-2.th -------------------------------------------------------------------------------- /glove/weights/glove-50-3.th: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/weights/glove-50-3.th -------------------------------------------------------------------------------- /glove/weights/glove-50.th: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/chenyuAndroid/deep-learning-examples/f909239ee1482ff977fe3e4210ac0e714f7de728/glove/weights/glove-50.th --------------------------------------------------------------------------------