-
Notifications
You must be signed in to change notification settings - Fork 1
/
utils.py
424 lines (320 loc) · 13.9 KB
/
utils.py
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
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
import tqdm
import numpy as np
import pandas as pd
from nltk.tokenize import word_tokenize
from nltk.stem import WordNetLemmatizer
from transformers import BertTokenizer
from scipy.special import softmax
import torch
import pickle
from torch.utils.data import TensorDataset, DataLoader
from transformers import BertForSequenceClassification, AdamW, BertConfig
from transformers import get_linear_schedule_with_warmup
from sklearn.model_selection import train_test_split
with open('models/discourse_model.pickle', 'rb') as f:
model = pickle.load(f)
model.eval()
import pickle
import torch
# If there's a GPU available...
if torch.cuda.is_available():
# Tell PyTorch to use the GPU.
device = torch.device("cuda")
print('There are %d GPU(s) available.' % torch.cuda.device_count())
print('We will use the GPU:', torch.cuda.get_device_name(0))
# If not...
else:
print('No GPU available, using the CPU instead.')
device = torch.device("cpu")
tqdm.tqdm.pandas()
lemmatizer = WordNetLemmatizer()
def flatten_thread_r(thread_id, comments):
extracted_comments = []
if len(comments) == 0:
return extracted_comments
else:
for comment in comments:
comment_dict = {
'id': comment['id'],
'parent_id': comment['parent_id'] or thread_id,
'controversiality': comment['controversiality'],
'body': comment['body'],
'created_utc': comment['created_utc'],
'author': comment['author']
}
extracted_comments += [comment_dict] + flatten_thread_r(thread_id, comment['children'])
return extracted_comments
def flatten_thread(thread):
flattened_comments = flatten_thread_r(thread['id'], thread['children'])
return flattened_comments
def label(threads):
def percent_upvoted(x):
if x.ups + x.downs == 0:
return 0
return x.ups / (x.ups + x.downs)
threads['percent_upvoted'] = threads.apply(percent_upvoted, axis=1)
bottom_quartile = np.percentile(threads.percent_upvoted.values, 25)
top_quartile = np.percentile(threads.percent_upvoted.values, 75)
print('Bottom quartile: ', bottom_quartile)
print('Top quartile: ', top_quartile)
def controversiality(x):
if x <= bottom_quartile:
return int(True)
elif x >= top_quartile:
return int(False)
else:
return -1
threads['label'] = threads.percent_upvoted.apply(controversiality)
threads = threads[threads.label != -1]
return threads
def prep_data(threads):
threads = label(threads)
threads['comments'] = threads.progress_apply(flatten_thread, axis=1)
return threads
def get_corpus(threads):
corpus = []
def _get_corpus(thread):
corpus.append(' '.join([lemmatizer.lemmatize(word.lower()) for word in word_tokenize(thread.selftext)]))
for comment in thread.comments:
corpus.append(' '.join([lemmatizer.lemmatize(word.lower()) for word in word_tokenize(comment['body'])]))
threads.progress_apply(_get_corpus, axis=1)
return corpus
def get_ids(threads):
ids = {}
def _get_ids(thread):
ids[thread.id] = [comment['id'] for comment in thread.comments]
threads.progress_apply(_get_ids, axis=1)
return ids
def _text_to_bert(threads):
def tokenize_fn(x):
tokenized = ['[CLS]'] + bert_tokenizer.tokenize(x)[:510] + ['[SEP]']
if len(tokenized) < 512:
tokenized += ['[PAD]'] * (512 - len(tokenized))
tokenized = bert_tokenizer.convert_tokens_to_ids(tokenized)
return tokenized
def tokenize_comments(x):
return [tokenize_fn(comment['body']) for comment in x]
title_body = threads.title + ' ' + threads.selftext
title_body = title_body.apply(tokenize_fn)
comments = df.comments.progress_apply(tokenize_comments)
return (title_body, comments)
def _create_tb_mask(x):
return [token_id > 0 for token_id in x]
def _create_comment_mask(x):
c = []
for comment in x:
c.append([token_id > 0 for token_id in comment])
return torch.LongTensor(np.stack(c)) if len(c) > 0 else torch.LongTensor([[0] * 512])
def get_discourse_acts(threads):
tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
with open('discourse_model.pickle', 'rb') as f:
discourse_model = pickle.load(f)
title_body, comments = _text_to_bert(threads)
title_body_mask = title_body.apply(_create_tb_mask)
comment_mask = comments.apply(_create_comment_mask)
title_body_tensor = torch.LongTensor(np.stack(title_body.values))
title_body_mask = torch.LongTensor(np.stack(title_body_mask.values))
comment_tensor = comments.progress_apply(lambda x: torch.LongTensor(np.stack(x)) if len(x) > 0 else torch.LongTensor(np.stack([[0] * 512])))
comment_mask = torch.LongTensor(np.stack(comment_mask.values))
def _find_post(id, comments):
for comment in comments:
if comment['id'] == id:
return comment
return None
def _create_bigrams(thread):
bigrams = []
for comment in thread.comments:
parent = _find_post(comment['parent_id'], thread.comments)
if parent is None:
bigrams.append((thread.discourse_act, comment['discourse_act']))
else:
bigrams.append((parent['discourse_act'], comment['discourse_act']))
return bigrams
def create_bigrams(threads):
threads['bigrams'] = threads.apply(_create_bigrams, axis=1)
def logits_to_act(logits):
return np.argmax(softmax(logits, axis=-1), axis=-1)
def assign_acts(tb_acts, c_acts, threads):
acts = logits_to_act(np.stack(tb_acts))
threads['discourse_act'] = acts
i = 0
for idx, row in tqdm.tqdm(threads.iterrows(), total=len(threads)):
cs = logits_to_act(c_acts[i])
new_comments = []
for c, c_act in zip(row.comments, cs):
c['discourse_act'] = c_act
new_comments.append(c)
i += 1
assert all(['discourse_act' in c.keys() for c in new_comments]), 'Not every comment got an act!'
row.comments = new_comments
def text_to_discourse_acts(threads):
print("Vectorizing threads...")
title_body_tensor, title_body_mask, comment_tensor, comment_mask = get_tensors(threads)
print("Running BERT")
tb_acts, c_acts = get_discourse_acts(title_body_tensor, title_body_mask, comment_tensor, comment_mask)
print("Assigning acts...")
assign_acts(tb_acts, c_acts, threads)
print('Done.')
def get_discourse_acts(title_body_tensor, title_body_mask, comment_tensor, comment_mask):
tb_acts = []
c_acts = []
for tbt, tbm, ct, cm in tqdm.tqdm(zip(title_body_tensor, title_body_mask, comment_tensor, comment_mask), total=len(title_body_tensor)):
with torch.no_grad():
dataset = TensorDataset(ct, cm)
loader = DataLoader(dataset, batch_size=3)
tbt = tbt.to(device)
tbm = tbm.to(device)
tb_act = model(tbt.view((1, -1)), attention_mask = tbm.view((1, -1)))[0].cpu().numpy()
tb_acts.append(tb_act[0])
cs = []
for batch in loader:
ct, cm = batch
ct = ct.to(device)
cm = cm.to(device)
c_act = model(ct, cm)[0].cpu().numpy()
for c in c_act:
cs.append(c)
cs = np.stack(cs)
c_acts.append(cs)
return tb_acts, c_acts
bert_tokenizer = BertTokenizer.from_pretrained('bert-base-uncased')
def _text_to_bert(threads):
def tokenize_fn(x):
tokenized = ['[CLS]'] + bert_tokenizer.tokenize(x)[:510] + ['[SEP]']
if len(tokenized) < 512:
tokenized += ['[PAD]'] * (512 - len(tokenized))
tokenized = bert_tokenizer.convert_tokens_to_ids(tokenized)
return tokenized
def tokenize_comments(x):
return [tokenize_fn(comment['body']) for comment in x]
title_body = threads.title + ' ' + threads.selftext
title_body = title_body.apply(tokenize_fn)
comments = threads.comments.progress_apply(tokenize_comments)
return (title_body, comments)
def _create_tb_mask(x):
return [int(token_id > 0) for token_id in x]
def _create_comment_mask(x):
c = []
for comment in x:
c.append([int(token_id > 0) for token_id in comment])
return torch.LongTensor(np.stack(c)) if len(c) > 0 else torch.LongTensor([[0] * 512])
def get_tensors(threads):
title_body, comments = _text_to_bert(threads)
title_body_mask = title_body.apply(_create_tb_mask)
comment_mask = comments.apply(_create_comment_mask)
title_body_tensor = torch.LongTensor(np.stack(title_body.values))
title_body_mask = torch.LongTensor(np.stack(title_body_mask.values))
comment_tensor = comments.progress_apply(lambda x: torch.LongTensor(np.stack(x)) if len(x) > 0 else torch.LongTensor(np.stack([[0] * 512])))
comment_tensor = comment_tensor.values
comment_mask = comment_mask.values
return title_body_tensor, title_body_mask, comment_tensor, comment_mask
def _flat_accuracy(preds, labels):
pred_flat = np.argmax(preds, axis=1).flatten()
labels_flat = labels.flatten()
return np.sum(pred_flat == labels_flat) / len(labels_flat)
import time
import datetime
def _format_time(elapsed):
'''
Takes a time in seconds and returns a string hh:mm:ss
'''
# Round to the nearest second.
elapsed_rounded = int(round((elapsed)))
# Format as hh:mm:ss
return str(datetime.timedelta(seconds=elapsed_rounded))
def finetune_bert(title_body_tensor, title_body_mask, threads):
labels = torch.LongTensor(np.stack(threads.label.values))
print(labels.size())
train_tbt, test_tbt, train_tbm, test_tbm, train_labels, test_labels = train_test_split(
title_body_tensor, title_body_mask, labels,
test_size=0.2, random_state=42
)
training_stats = []
train_dataset = TensorDataset(train_tbt, train_tbm, train_labels)
test_dataset = TensorDataset(test_tbt, test_tbm, test_labels)
train_dataloader = DataLoader(train_dataset, batch_size=2)
test_dataloader = DataLoader(test_dataset, batch_size=2)
epochs = 4
total_steps = len(train_dataloader) * epochs
model = BertForSequenceClassification.from_pretrained(
'bert-base-uncased',
num_labels=2,
output_attentions=False,
output_hidden_states=True
)
model.cuda()
optimizer = AdamW(model.parameters(),
lr=2e-5,
eps=1e-8)
scheduler = get_linear_schedule_with_warmup(optimizer,
num_warmup_steps = 0, # Default value in run_glue.py
num_training_steps = total_steps)
total_steps = len(train_dataloader) * epochs
total_t0 = time.time()
for epoch_i in range(0, epochs):
print("")
print('======== Epoch {:} / {:} ========'.format(epoch_i + 1, epochs))
print('Training...')
model.train()
t0 = time.time()
total_train_loss = 0
for step, batch in enumerate(train_dataloader):
b_input_ids = batch[0].to(device)
b_input_mask = batch[1].to(device)
b_labels = batch[2].to(device)
model.zero_grad()
loss, logits, _ = model(b_input_ids,
token_type_ids=None,
attention_mask=b_input_mask,
labels=b_labels)
total_train_loss += loss.item()
loss.backward()
torch.nn.utils.clip_grad_norm_(model.parameters(), 1.0)
optimizer.step()
scheduler.step()
avg_train_loss = total_train_loss / len(train_dataloader)
training_time = _format_time(time.time() - t0)
print("")
print(" Average training loss: {0:.2f}".format(avg_train_loss))
print(" Training epcoh took: {:}".format(training_time))
print("")
print("Running Validation...")
t0 = time.time()
model.eval()
total_eval_accuracy = 0
total_eval_loss = 0
nb_eval_steps = 0
for batch in test_dataloader:
b_input_ids = batch[0].to(device)
b_input_mask = batch[1].to(device)
b_labels = batch[2].to(device)
with torch.no_grad():
loss, logits, _ = model(b_input_ids,
token_type_ids=None,
attention_mask=b_input_mask,
labels=b_labels)
total_eval_loss += loss.item()
logits = logits.detach().cpu().numpy()
label_ids = b_labels.cpu().numpy()
total_eval_accuracy += _flat_accuracy(logits, label_ids)
avg_val_accuracy = total_eval_accuracy / len(test_dataloader)
print(" Accuracy: {0:.2f}".format(avg_val_accuracy))
avg_val_loss = total_eval_loss / len(test_dataloader)
validation_time = _format_time(time.time() - t0)
print(" Validation Loss: {0:.2f}".format(avg_val_loss))
print(" Validation took: {:}".format(validation_time))
# Record all statistics from this epoch.
training_stats.append(
{
'epoch': epoch_i + 1,
'Training Loss': avg_train_loss,
'Valid. Loss': avg_val_loss,
'Valid. Accur.': avg_val_accuracy,
'Training Time': training_time,
'Validation Time': validation_time
}
)
print("")
print("Training complete!")
print("Total training took {:} (h:mm:ss)".format(_format_time(time.time()-total_t0)))
return model.cpu(), training_stats