-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathrunner_base.py
348 lines (308 loc) · 15 KB
/
runner_base.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
import os.path
import torch
from tqdm import tqdm
import logging
import pickle
import numpy as np
import datetime
class ExperimentRunnerBase(torch.nn.Module):
def __init__(self, args):
super(ExperimentRunnerBase, self).__init__()
data = args.data
self.name = args.name
self.task = args.task
self.batch_size = args.batch_size
self.epochs = args.epochs
self.loss_weighting = args.loss_weighting
self.TYPE = torch.LongTensor
self.TYPEF = torch.FloatTensor
self.use_cuda = torch.cuda.is_available()
self.out_file = args.out_file
self.from_which_enable = args.from_which
self.small_data = args.small
if self.use_cuda:
self.TYPE = torch.cuda.LongTensor
self.TYPEF = torch.cuda.FloatTensor
if data == "babi":
from data_babi import find_entities, read_dataset, TextDataset, collate_fn
data_dir = "data/dialog-bAbI-tasks"
kb_path = "dialog-babi-kb-all.txt"
data_file_prefix = self.getBabiDataNames(self.task)
elif data == "personal_context":
from data_personal_context import find_entities, read_dataset, TextDataset, collate_fn
elif data == "personal":
from data_personalized import find_entities, read_dataset, TextDataset, collate_fn
else:
raise ModuleNotFoundError()
if data.startswith("personal"):
data_dir = "data/personalized-dialog-dataset/full"
if self.small_data:
data_dir = "data/personalized-dialog-dataset/small"
kb_path = "personalized-dialog-kb-all.txt"
data_file_prefix = self.getPersonalDataNames(args.task)
self.kb_entries = find_entities(os.path.join(data_dir, "../full", kb_path))
train, self.w2i = list(read_dataset(os.path.join(data_dir, f"{data_file_prefix}-trn.txt"), self.kb_entries))
dev, _ = list(read_dataset(os.path.join(data_dir, f"{data_file_prefix}-dev.txt"), self.kb_entries))
test, _ = list(read_dataset(os.path.join(data_dir, f"{data_file_prefix}-tst.txt"), self.kb_entries))
self.data_train = TextDataset(train, self.w2i)
self.data_dev = TextDataset(dev, self.w2i)
self.data_test = TextDataset(test, self.w2i)
self.i2w = {v: k for k, v in self.w2i.items()}
self.train_data_loader = torch.utils.data.DataLoader(dataset=self.data_train,
batch_size=self.batch_size,
shuffle=True,
collate_fn=collate_fn,
pin_memory=True,
num_workers=5)
self.dev_data_loader = torch.utils.data.DataLoader(dataset=self.data_dev,
batch_size=self.batch_size,
shuffle=False,
collate_fn=collate_fn,
pin_memory=True,
num_workers=5)
self.test_data_loader = torch.utils.data.DataLoader(dataset=self.data_test,
batch_size=self.batch_size,
shuffle=False,
collate_fn=collate_fn,
pin_memory=True,
num_workers=5)
self.args = args
self.nwords = len(self.w2i)
self.cross_entropy = torch.nn.CrossEntropyLoss()
self.n = 1
self.loss = 0
self.ploss = 0
self.vloss = 0
self.acc = 0
self.avg_best = 0
self.train_plot_data = {
'train': {
'batch': [],
'epoch': [],
'loss': [],
'vocab_loss': [],
'ptr_loss': [],
}}
self.val_plot_data ={
'val': {
'batch': [],
'loss': [],
'vocab_loss': [],
'ptr_loss': [],
'wer': [],
'acc': []
}}
self.test_plot_data = {
'test': {
'loss': [],
'vocab_loss': [],
'ptr_loss': [],
}}
if self.loss_weighting:
self.loss_weights = torch.tensor([1.0, 1.0], requires_grad=True)
if self.use_cuda:
self.loss_weights = self.loss_weights.cuda()
self.loss_weights = torch.nn.Parameter(self.loss_weights)
def trainer(self):
with open(f"log-{str(datetime.datetime.now())}-{self.name}", 'w') as log_file:
try:
for epoch in range(self.epochs):
pbar = tqdm(enumerate(self.train_data_loader), total=len(self.train_data_loader))
for i, batch in pbar:
if self.use_cuda:
batch[0] = batch[0].to('cuda', non_blocking=True)
batch[1] = batch[1].to('cuda', non_blocking=True)
batch[2] = batch[2].to('cuda', non_blocking=True)
batch[3] = batch[3].to('cuda', non_blocking=True)
if self.__class__.__name__.startswith("Split"):
batch[9] = batch[9].to('cuda', non_blocking=True)
self.train()
loss, vloss, ploss = self.train_batch_wrapper(batch, i == 0, 8)
pbar.set_description(self.print_loss())
if self.args.log:
print(f"epoch {epoch}: {self.print_loss()}", file=log_file)
self.train_plot_data['train']['batch'].append(i)
self.train_plot_data['train']['epoch'].append(epoch)
self.train_plot_data['train']['loss'].append(loss)
self.train_plot_data['train']['vocab_loss'].append(vloss)
self.train_plot_data['train']['ptr_loss'].append(ploss)
if epoch % self.args.val == 0:
os.makedirs('checkpoints/ckpt-' + str(self.name) + '-' + str(epoch), exist_ok=True)
self.save_models('checkpoints/ckpt-' + str(self.name) + '-' + str(epoch))
self.eval()
self.acc = self.evaluate(self.dev_data_loader, self.avg_best, self.kb_entries, self.i2w, epoch)
self.scheduler.step(self.acc)
# if 'Mem2Seq' in args['decoder']:
# model.scheduler.step(acc)
if self.acc is None or self.avg_best is None:
continue
if (self.acc >= self.avg_best):
self.avg_best = self.acc
cnt = 0
else:
cnt += 1
if cnt == 5: break
if self.acc == 1.0: break
except KeyboardInterrupt as e:
out_file = open(f"plot-data-{self.name}.pkl", 'wb')
out_file.write(pickle.dumps((self.train_plot_data, self.val_plot_data)))
out_file.close()
print(e)
exit(0)
out_file = open(f"plot-data-{self.name}.pkl", 'wb')
out_file.write(pickle.dumps((self.train_plot_data, self.val_plot_data)))
out_file.close()
def getPersonalDataNames(self, tasknum):
mapping = {"1": "personalized-dialog-task1-API-calls",
"2": "personalized-dialog-task2-API-refine",
"3": "personalized-dialog-task3-options",
"4": "personalized-dialog-task4-info",
"5": "personalized-dialog-task5-full-dialogs"}
return mapping[str(tasknum)]
def getBabiDataNames(self, tasknum):
mapping = {"1" : "dialog-babi-task1-API-calls",
"2" : "dialog-babi-task2-API-refine",
"3" : "dialog-babi-task3-options",
"4" : "dialog-babi-task4-phone-address",
"5" : "dialog-babi-task5-full-dialogs",
"6" : "dialog-babi-task6-dstc2"}
return mapping[str(tasknum)]
def losses(self):
return self.loss / self.n, self.ploss / self.n, self.vloss / self.n
def _optimize(self, predicted_answers, true_answers):
"""Implement this in subclass"""
raise NotImplementedError()
def print_loss(self):
print_loss_avg = self.loss / self.n
print_ploss = self.ploss / self.n
print_vloss = self.vloss / self.n
self.n += 1
p_str = 'L:{:.5f}, VL:{:.5f}, PL:{:.5f}'.format(print_loss_avg, print_vloss, print_ploss)
if self.loss_weighting:
p_str += ', LW: {:.3f} {:.3f}'.format(self.loss_weights[0], self.loss_weights[1])
return p_str
def evaluate(self, dev, avg_best, kb_entries, i2w, epoch):
self.loss = 0
self.ploss = 0
self.vloss = 0
self.n = 1
self.incorrect_sentinel = 0
self.i2w = i2w
def wer(r, h):
"""
This is a function that calculate the word error rate in ASR.
You can use it like this: wer("what is it".split(), "what is".split())
"""
# build the matrix
d = np.zeros((len(r) + 1) * (len(h) + 1), dtype=np.uint8).reshape((len(r) + 1, len(h) + 1))
for i in range(len(r) + 1):
for j in range(len(h) + 1):
if i == 0:
d[0][j] = j
elif j == 0:
d[i][0] = i
for i in range(1, len(r) + 1):
for j in range(1, len(h) + 1):
if r[i - 1] == h[j - 1]:
d[i][j] = d[i - 1][j - 1]
else:
substitute = d[i - 1][j - 1] + 1
insert = d[i][j - 1] + 1
delete = d[i - 1][j] + 1
d[i][j] = min(substitute, insert, delete)
result = float(d[len(r)][len(h)]) / len(r) * 100
# result = str("%.2f" % result) + "%"
return result
logging.info("STARTING EVALUATION")
acc_avg = 0.0
wer_avg = 0.0
bleu_avg = 0.0
acc_P = 0.0
acc_V = 0.0
ref = []
hyp = []
ref_s = ""
hyp_s = ""
dialog_acc_dict = {}
global_entity_list = kb_entries
pbar = tqdm(enumerate(dev), total=len(dev))
for j, data_dev in pbar:
profile_mem = None
if self.__class__.__name__.startswith("Split"):
profile_mem = data_dev[9]
if self.use_cuda:
data_dev[0] = data_dev[0].to('cuda', non_blocking=True)
data_dev[1] = data_dev[1].to('cuda', non_blocking=True)
data_dev[2] = data_dev[2].to('cuda', non_blocking=True)
data_dev[3] = data_dev[3].to('cuda', non_blocking=True)
if self.__class__.__name__.startswith("Split"):
profile_mem = profile_mem.to('cuda', non_blocking=True)
if profile_mem != None:
profile_mem = profile_mem.transpose(0,1)
words, from_whichs = self.evaluate_batch(len(data_dev[1]), data_dev[0].transpose(0, 1), data_dev[4],
data_dev[1].transpose(0, 1), data_dev[5],
data_dev[2].transpose(0, 1), data_dev[3].transpose(0, 1), data_dev[7],
profile_mem)
transposed_words = [[row[i] for row in words] for i in range(len(words[0]))]
if self.from_which_enable:
transposed_fromwhich = [[row[i] for row in from_whichs] for i in range(len(from_whichs[0]))]
self.out_file = self.name if self.out_file == '' else self.out_file
with open(self.out_file, 'a') as f:
if j==0:
f.write("Epoch {}\n".format(epoch))
f.write('------------truth---------------\n\n')
[f.write(w + '\n') for w in data_dev[8]]
f.write('------------response-------------\n\n')
[f.write(' '.join(w) + '\n') for w in transposed_words]
f.write('\n')
if self.from_which_enable:
[f.write(" ".join(w) + '\n') for w in transposed_fromwhich]
f.write("\n")
acc = 0
w = 0
temp_gen = []
for i, row in enumerate(np.transpose(words)):
st = ''
for e in row:
if e == '<eos>':
break
else:
st += e + ' '
temp_gen.append(st)
correct = data_dev[8][i]
### compute F1 SCORE
st = st.lstrip().rstrip()
correct = correct.lstrip().rstrip()
if data_dev[6][i] not in dialog_acc_dict.keys():
dialog_acc_dict[data_dev[6][i].item()] = []
if (correct == st):
acc += 1
dialog_acc_dict[data_dev[6][i].item()].append(1)
else:
dialog_acc_dict[data_dev[6][i].item()].append(0)
w += wer(correct, st)
ref.append(str(correct))
hyp.append(str(st))
ref_s += str(correct) + "\n"
hyp_s += str(st) + "\n"
acc_avg += acc / float(len(data_dev[1]))
wer_avg += w / float(len(data_dev[1]))
pbar.set_description("R:{:.4f},W:{:.4f},I:{:.4f}".format(acc_avg / float(len(dev)),
wer_avg / float(len(dev)),
self.incorrect_sentinel / float(len(dev))))
self.val_plot_data['val']['batch'].append(j)
self.val_plot_data['val']['acc'].append(acc_avg / float(len(dev)))
self.val_plot_data['val']['wer'].append(wer_avg / float(len(dev)))
self.val_plot_data['val']['loss'].append(self.losses()[0])
self.val_plot_data['val']['vocab_loss'].append(self.losses()[1])
self.val_plot_data['val']['ptr_loss'].append(self.losses()[2])
# dialog accuracy
dia_acc = 0
for k in dialog_acc_dict.keys():
if len(dialog_acc_dict[k]) == sum(dialog_acc_dict[k]):
dia_acc += 1
logging.info("Dialog Accuracy:\t" + str(dia_acc * 1.0 / len(dialog_acc_dict.keys())))
acc_avg = acc_avg / float(len(dev))
if (acc_avg >= avg_best):
return acc_avg
return avg_best