File size: 15,494 Bytes
27b305f
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np\n",
    "\n",
    "train_data=pd.read_csv(\"data/train.csv\", engine=\"python\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "attributes=[\"toxic\",\"severe_toxic\",\"obscene\",\"threat\",\"insult\",\"identity_hate\"]\n",
    "#train_data[attributes].sum().plot.bar()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch.utils.data import Dataset\n",
    "import torch\n",
    "\n",
    "class toxicity_dataset(Dataset):\n",
    "    def __init__(self,data_path,tokenizer,attributes,max_token_len= 128,sample = 5000):\n",
    "        self.data_path=data_path\n",
    "        self.tokenizer=tokenizer\n",
    "        self.attributes=attributes\n",
    "        self.max_token_len=max_token_len\n",
    "        self.sample=sample\n",
    "        self._prepare_data()\n",
    "    def _prepare_data(self):\n",
    "        data=pd.read_csv(self.data_path)\n",
    "        if self.sample is not None:\n",
    "            self.data=data.sample(self.sample,random_state=7)\n",
    "        else:\n",
    "            self.data=data\n",
    "    def __len__(self):\n",
    "        return(len(self.data))\n",
    "    def __getitem__(self,index):\n",
    "        item = self.data.iloc[index]\n",
    "        comment = str(item.comment_text)\n",
    "        attributes = torch.FloatTensor(item[self.attributes])\n",
    "        tokens = self.tokenizer.encode_plus(comment,add_special_tokens=True,return_tensors=\"pt\",truncation=True,max_length=self.max_token_len,padding=\"max_length\",return_attention_mask=True)\n",
    "        return{'input_ids':tokens.input_ids.flatten(),\"attention_mask\":tokens.attention_mask.flatten(),\"labels\":attributes}\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "c:\\Users\\jozef\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\tqdm\\auto.py:21: TqdmWarning: IProgress not found. Please update jupyter and ipywidgets. See https://ipywidgets.readthedocs.io/en/stable/user_install.html\n",
      "  from .autonotebook import tqdm as notebook_tqdm\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "{'input_ids': tensor([    0,   113, 43292,   487,  1073,  6619, 16519,  4261,  1012, 28845,\n",
       "         43292, 50118, 30086,     6,    38,   206,     5,  7729,  6619, 16519,\n",
       "          4261,  1012,  6717,  4867,    11,  2370,    32,    45, 41039,  7140,\n",
       "           250, 48149,    53,   888,    95, 41762,    30, 40823, 34740,  2071,\n",
       "             4,   407,    51,   197,   213,    11, 29617,   101, 25046, 12467,\n",
       "           381, 11742,   646, 19065,  6026,   742,  1195,    87,   634, 14530,\n",
       "           250,     4,  1437,  1437,    22,     2,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1,     1,     1,\n",
       "             1,     1,     1,     1,     1,     1,     1,     1]),\n",
       " 'attention_mask': tensor([1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,\n",
       "         1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 0, 0, 0, 0, 0, 0,\n",
       "         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",
       "         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",
       "         0, 0, 0, 0, 0, 0, 0, 0]),\n",
       " 'labels': tensor([0., 0., 0., 0., 0., 0.])}"
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from transformers import AutoTokenizer\n",
    "model_name=\"roberta-base\"\n",
    "tokenizer=AutoTokenizer.from_pretrained(model_name)\n",
    "toxic_comments_dataset=toxicity_dataset(\"data/train.csv\",tokenizer,attributes)\n",
    "toxic_comments_dataset.__getitem__(0)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pytorch_lightning as pl\n",
    "from torch.utils.data import DataLoader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Toxcity_Data_Module(pl.LightningDataModule):\n",
    "    def __init__(self,train_path,test_path,attributes,batch_size = 16, max_token_len = 128, model_name=\"roberta-base\"):\n",
    "        super().__init__()\n",
    "        self.train_path=train_path\n",
    "        self.test_path=test_path\n",
    "        self.attributes=attributes\n",
    "        self.batch_size=batch_size\n",
    "        self.max_token_len=max_token_len\n",
    "        self.model_name=model_name\n",
    "        self.tokenizer = AutoTokenizer.from_pretrained(model_name)\n",
    "    def setup(self, stage = None):\n",
    "        if stage in (None, \"fit\"):\n",
    "            self.train_dataset=toxicity_dataset(self.train_path,self.tokenizer,self.attributes)\n",
    "            self.test_dataset=toxicity_dataset(self.test_path,self.tokenizer,self.attributes, sample=None)\n",
    "        if stage == \"predict\":\n",
    "            self.val_dataset=toxicity_dataset(self.test_path,self.tokenizer,self.attributes)\n",
    "    def train_dataloader(self):\n",
    "        return DataLoader(self.train_dataset,batch_size=self.batch_size,num_workers=4,shuffle=True)\n",
    "    def val_dataloader(self):\n",
    "        return DataLoader(self.test_dataset,batch_size=self.batch_size,num_workers=4,shuffle=False)\n",
    "    def predict_dataloader(self):\n",
    "        return DataLoader(self.test_dataset,batch_size=self.batch_size,num_workers=4,shuffle=False)\n",
    "    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<torch.utils.data.dataloader.DataLoader at 0x2241d16e5d0>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "toxicity_data_module=Toxcity_Data_Module(\"data/train.csv\",\"data/test.csv\",attributes)\n",
    "\n",
    "toxicity_data_module.setup()\n",
    "dataloader=toxicity_data_module.train_dataloader()\n",
    "dataloader"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoModel, AdamW, get_cosine_schedule_with_warmup\n",
    "import torch.nn as nn\n",
    "import math\n",
    "from torchmetrics.functional.classification import auroc\n",
    "import torch.nn.functional as F"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Toxic_Comment_Classifier(pl.LightningModule):\n",
    "    def __init__(self,config:dict):\n",
    "        super().__init__()\n",
    "        self.config=config\n",
    "        self.pretrained_model = AutoModel.from_pretrained(config['model_name'],return_dict=True)\n",
    "        self.hidden = nn.Linear(self.pretrained_model.config.hidden_size, self.pretrained_model.config.hidden_size)\n",
    "        self.classifier = nn.Linear(self.pretrained_model.config.hidden_size, self.config['n_labels'])\n",
    "        torch.nn.init.xavier_uniform_(self.hidden.weight)\n",
    "        torch.nn.init.xavier_uniform_(self.classifier.weight)\n",
    "        self.loss_func=nn.BCEWithLogitsLoss(reduction='mean')\n",
    "        self.dropout = nn.Dropout()\n",
    "\n",
    "    def forward(self, input_ids,attention_mask,labels=None):\n",
    "        output = self.pretrained_model(input_ids=input_ids,attention_mask=attention_mask)\n",
    "        pooled_output=torch.mean(output.last_hidden_state, 1)\n",
    "        #nn classification\n",
    "        pooled_output=self.hidden(pooled_output)\n",
    "        pooled_output=self.dropout(pooled_output)\n",
    "        pooled_output=F.relu(pooled_output)\n",
    "        logits=self.classifier(pooled_output)\n",
    "        #loss\n",
    "        loss = 0\n",
    "        if labels is not None:\n",
    "            loss = self.loss_func(logits.view(-1,self.config['n_labels']), labels.view(-1,self.config['n_labels']))\n",
    "        return loss, logits\n",
    "    \n",
    "    def training_step(self, batch, batch_index):\n",
    "        loss, logits = self(**batch)\n",
    "        self.log(\"train loss\", loss, prog_bar=True, logger=True)\n",
    "        return{'loss':loss,'predictions':logits,'labels':batch['labels']}\n",
    "    def validation_step(self, batch, batch_index):\n",
    "        loss, logits = self(**batch)\n",
    "        self.log(\"validation loss\", loss, prog_bar=True, logger=True)\n",
    "        return{'val_loss':loss,'predictions':logits,'labels':batch['labels']}\n",
    "    def prediction_step(self, batch, batch_index):\n",
    "        logits = self(**batch)\n",
    "        return logits\n",
    "    \n",
    "    def configure_optimizers(self):\n",
    "        optimizer = AdamW(self.parameters(), lr=self.config['lr'], weight_decay=self.config['w_decay'])\n",
    "        total_steps = self.config['train_size']/self.config['bs']\n",
    "        warmup_steps = math.floor(total_steps*self.config['warmup'])\n",
    "        scheduler = get_cosine_schedule_with_warmup(optimizer,warmup_steps,total_steps)\n",
    "        return [optimizer],[scheduler]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at distilroberta-base were not used when initializing RobertaModel: ['lm_head.dense.weight', 'lm_head.layer_norm.weight', 'lm_head.layer_norm.bias', 'lm_head.dense.bias', 'lm_head.decoder.weight', 'lm_head.bias']\n",
      "- This IS expected if you are initializing RobertaModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing RobertaModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n"
     ]
    }
   ],
   "source": [
    "config = {\n",
    "    'model_name':\"distilroberta-base\",\n",
    "    'n_labels':len(attributes),\n",
    "    'bs':128,\n",
    "    'lr':1.5e-6,\n",
    "    'warmup':0.2,\n",
    "    \"train_size\":len(toxicity_data_module.train_dataloader()),\n",
    "    'w_decay':0.001,\n",
    "    'n_epochs':1\n",
    "}\n",
    "\n",
    "model = Toxic_Comment_Classifier(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "idx=0\n",
    "input_ids = toxic_comments_dataset.__getitem__(idx)[\"input_ids\"]\n",
    "attention_mask = toxic_comments_dataset.__getitem__(idx)[\"attention_mask\"]\n",
    "labels = toxic_comments_dataset.__getitem__(idx)[\"labels\"]\n",
    "loss,output = model(input_ids.unsqueeze(dim=0), attention_mask.unsqueeze(dim=0),labels.unsqueeze(dim=0))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Some weights of the model checkpoint at distilroberta-base were not used when initializing RobertaModel: ['lm_head.dense.weight', 'lm_head.layer_norm.weight', 'lm_head.layer_norm.bias', 'lm_head.dense.bias', 'lm_head.decoder.weight', 'lm_head.bias']\n",
      "- This IS expected if you are initializing RobertaModel from the checkpoint of a model trained on another task or with another architecture (e.g. initializing a BertForSequenceClassification model from a BertForPreTraining model).\n",
      "- This IS NOT expected if you are initializing RobertaModel from the checkpoint of a model that you expect to be exactly identical (initializing a BertForSequenceClassification model from a BertForSequenceClassification model).\n",
      "GPU available: False, used: False\n",
      "TPU available: False, using: 0 TPU cores\n",
      "IPU available: False, using: 0 IPUs\n",
      "HPU available: False, using: 0 HPUs\n",
      "c:\\Users\\jozef\\AppData\\Local\\Programs\\Python\\Python311\\Lib\\site-packages\\pytorch_lightning\\trainer\\connectors\\logger_connector\\logger_connector.py:67: UserWarning: Starting from v1.9.0, `tensorboardX` has been removed as a dependency of the `pytorch_lightning` package, due to potential conflicts with other packages in the ML ecosystem. For this reason, `logger=True` will use `CSVLogger` as the default logger, unless the `tensorboard` or `tensorboardX` packages are found. Please `pip install lightning[extra]` or one of them to enable TensorBoard support by default\n",
      "  warning_cache.warn(\n"
     ]
    }
   ],
   "source": [
    "toxicity_data_module=Toxcity_Data_Module(\"data/train.csv\",\"data/test.csv\",attributes,batch_size=config['bs'])\n",
    "toxicity_data_module.setup()\n",
    "model = Toxic_Comment_Classifier(config)\n",
    "\n",
    "trainer = pl.Trainer(max_epochs=config['n_epochs'],num_sanity_val_steps=50)\n",
    "#device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")\n",
    "#print(torch.cuda.get_device_name())\n",
    "#trainer.fit(model,toxicity_data_module)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = torch.load(\"ToxicityClassificationModel.pt\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def predict_raw_comments(model, dm):\n",
    "    predictions = trainer.predict(model,datamodule=dm)\n",
    "    flattened_predictions = np.stack([torch.sigmoid(torch.Tensor(p)) for batch in predictions for p in batch])\n",
    "    return flattened_predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictions = predict_raw_comments(model=model,dm=toxicity_data_module)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.11.3"
  },
  "orig_nbformat": 4
 },
 "nbformat": 4,
 "nbformat_minor": 2
}