-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathexperiment.py
708 lines (555 loc) · 38.4 KB
/
experiment.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
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
import torch
import torch.nn as nn
import torch.nn.functional as F
from torch.utils.data import DataLoader
from torch import optim
import os
import numpy as np
import time
from dataloaders import data_dict,data_set
from sklearn.metrics import confusion_matrix
import yaml
# import models
from models.model_builder import model_builder
from torch.utils.data.sampler import WeightedRandomSampler
from sklearn.metrics import accuracy_score
from sklearn.metrics import f1_score
from utils import EarlyStopping, adjust_learning_rate_class
from datetime import datetime
import matplotlib.pyplot as plt
import seaborn as sns
from dataloaders.augmentation import RandomAugment, mixup_data
import random
import os
class MixUpLoss(nn.Module):
"""
Mixup implementation heavily borrowed from https://github.com/fastai/fastai/blob/master/fastai/callbacks/mixup.py#L42
Adapt the loss function `crit` to go with mixup.
"""
def __init__(self, crit, reduction='mean'):
super().__init__()
if hasattr(crit, 'reduction'):
self.crit = crit
self.old_red = crit.reduction
setattr(self.crit, 'reduction', 'none')
self.reduction = reduction
def forward(self, output, target):
if len(target.size()) == 2:
loss1, loss2 = self.crit(output, target[:, 0].long()), self.crit(output, target[:, 1].long())
d = loss1 * target[:, 2] + loss2 * (1 - target[:, 2])
else:
d = self.crit(output, target)
if self.reduction == 'mean':
return d.mean()
elif self.reduction == 'sum':
return d.sum()
return d
def get_old(self):
if hasattr(self, 'old_crit'):
return self.old_crit
elif hasattr(self, 'old_red'):
setattr(self.crit, 'reduction', self.old_red)
return self.crit
class Exp(object):
def __init__(self, args):
self.args = args
# set the device
self.device = self.acquire_device()
self.optimizer_dict = {"Adam":optim.Adam}
self.criterion_dict = {"MSE":nn.MSELoss,"CrossEntropy":nn.CrossEntropyLoss}
# Biuld The Model
self.model = self.build_model().to(self.device)
print("Done!")
self.model_size = np.sum([para.numel() for para in self.model.parameters() if para.requires_grad])
print("Parameter :", self.model_size)
print("Set the seed as : ", self.args.seed)
def acquire_device(self):
if self.args.use_gpu:
os.environ["CUDA_VISIBLE_DEVICES"] = str(self.args.gpu) if not self.args.use_multi_gpu else self.args.devices
device = torch.device('cuda:{}'.format(self.args.gpu))
print('Use GPU: cuda:{}'.format(self.args.gpu))
else:
device = torch.device('cpu')
print('Use CPU')
return device
def build_model(self):
model = model_builder(self.args)
return model.double()
def _select_optimizer(self):
if self.args.optimizer not in self.optimizer_dict.keys():
raise NotImplementedError
model_optim = self.optimizer_dict[self.args.optimizer](self.model.parameters(), lr=self.args.learning_rate)
return model_optim
def _select_criterion(self):
if self.args.criterion not in self.criterion_dict.keys():
raise NotImplementedError
criterion = self.criterion_dict[self.args.criterion]()
return criterion
def _get_data(self, data, flag="train", weighted_sampler = False):
if flag == 'train':
shuffle_flag = True
else:
shuffle_flag = False
data = data_set(self.args,data,flag)
if flag == "train":
if self.args.mixup_probability < 1 or self.args.random_augmentation_prob<1:
random_aug =RandomAugment(self.args.random_augmentation_nr, self.args.random_augmentation_config, self.args.max_aug )
def collate_fn(batch):
batch_x1 = []
batch_x2 = []
batch_y = []
for x,y,z in batch:
if np.random.uniform(0,1,1)[0]>=self.args.random_augmentation_prob:
batch_x1.append(random_aug(x))
else:
batch_x1.append(x)
batch_x2.append(y)
batch_y.append(z)
batch_x1 = torch.tensor(np.concatenate(batch_x1, axis=0))
batch_x2 = torch.tensor(batch_x2)
batch_y = torch.tensor(batch_y)
if np.random.uniform(0,1,1)[0] >= self.args.mixup_probability:
batch_x1 , batch_y = mixup_data(batch_x1 , batch_y, self.args.mixup_alpha, argmax = self.args.mixup_argmax)
#print("Mixup",batch_x1.shape,batch_y.shape)
#else:
#print(batch_x1.shape,batch_y.shape)
batch_x1 = torch.unsqueeze(batch_x1,1)
return batch_x1,batch_x2,batch_y
else:
collate_fn = None
else:
collate_fn = None
if weighted_sampler and flag == 'train':
sampler = WeightedRandomSampler(
data.act_weights, len(data.act_weights)
)
data_loader = DataLoader(data,
batch_size = self.args.batch_size,
#shuffle = shuffle_flag,
num_workers = 0,
sampler=sampler,
drop_last = False,
collate_fn = collate_fn)
else:
data_loader = DataLoader(data,
batch_size = self.args.batch_size,
shuffle = shuffle_flag,
num_workers = 0,
drop_last = False,
collate_fn = collate_fn)
return data_loader
def get_setting_name(self):
if self.args.model_type in ["deepconvlstm", "deepconvlstm_attn", "mcnn", "attend", "sahar", "tinyhar"]:
setting = "model_{}_data_{}_seed_{}_differencing_{}_Seperation_{}_magnitude_{}_Mixup_{}_RandomAug_{}_Scaling_{}_Mixupargmax_{}".format(self.args.model_type,
self.args.data_name,
self.args.seed,
self.args.difference,
self.args.filtering,
self.args.magnitude,
self.args.mixup_probability,
self.args.random_augmentation_prob,
self.args.filter_scaling_factor,
self.args.mixup_argmax
)
return setting
else:
raise NotImplementedError
# if self.args.model_type == "deepconvlstm":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["deepconvlstm"]
# setting = "deepconvlstm_data_{}_seed_{}_windowsize_{}_waveFilter_{}_Fscaling_{}_cvfilter_{}_lstmfilter_{}_Regu_{}_wavelearnble_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# self.args.wavelet_filtering,
# self.args.filter_scaling_factor,
# config["nb_filters"],
# config["nb_units_lstm"],
# self.args.wavelet_filtering_regularization,
# self.args.wavelet_filtering_learnable )
# return setting
# if self.args.model_type == "deepconvlstm_attn":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["deepconvlstm_attn"]
# setting = "deepconvlstm_attn_data_{}_seed_{}_windowsize_{}_waveFilter_{}_Fscaling_{}_cvfilter_{}_lstmfilter_{}_Regu_{}_wavelearnble_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# self.args.wavelet_filtering,
# self.args.filter_scaling_factor,
# config["nb_filters"],
# config["nb_units_lstm"],
# self.args.wavelet_filtering_regularization,
# self.args.wavelet_filtering_learnable )
# return setting
# if self.args.model_type == "mcnn":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["mcnn"]
# setting = "mcnn_data_{}_seed_{}_windowsize_{}_waveFilter_{}_Fscaling_{}_cvfilter_{}_Regu_{}_wavelearnble_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# self.args.wavelet_filtering,
# self.args.filter_scaling_factor,
# config["nb_filters"],
# self.args.wavelet_filtering_regularization,
# self.args.wavelet_filtering_learnable )
# return setting
# elif self.args.model_type == "attend":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["attend"]
# setting = "attend_data_{}_seed_{}_windowsize_{}_waveFilter_{}_Fscaling_{}_cvfilter_{}_grufilter_{}_Regu_{}_wavelearnble_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# self.args.wavelet_filtering,
# self.args.filter_scaling_factor,
# config["filter_num"],
# config["hidden_dim"],
# self.args.wavelet_filtering_regularization,
# self.args.wavelet_filtering_learnable)
# return setting
# elif self.args.model_type == "sahar":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["sahar"]
# setting = "sahar_data_{}_seed_{}_windowsize_{}_waveFilter_{}_Fscaling_{}_cvfilter_{}_grufilter_{}_Regu_{}_wavelearnble_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# self.args.wavelet_filtering,
# self.args.filter_scaling_factor,
# config["nb_filters"],
# None,
# self.args.wavelet_filtering_regularization,
# self.args.wavelet_filtering_learnable)
# return setting
# elif self.args.model_type == "tinyhar":
# config_file = open('../../configs/model.yaml', mode='r')
# config = yaml.load(config_file, Loader=yaml.FullLoader)["tinyhar"]
# setting = "tinyhar_data_{}_seed_{}_windowsize_{}_cvfilter_{}_CI_{}_CA_{}_TI_{}_TA_{}".format(self.args.data_name,
# self.args.seed,
# self.args.windowsize,
# config["filter_num"],
# self.args.cross_channel_interaction_type,
# self.args.cross_channel_aggregation_type,
# self.args.temporal_info_interaction_type,
# self.args.temporal_info_aggregation_type )
# return setting
# else:
# raise NotImplementedError
def update_gamma(self ):
for n, parameter in self.model.named_parameters():
if "gamma" in n:
parameter.grad.data.add_(self.args.regulatization_tradeoff*torch.sign(parameter.data)) # L1
def train(self):
setting = self.get_setting_name()
path = os.path.join(self.args.to_save_path,'logs/'+setting)
self.path = path
if not os.path.exists(path):
os.makedirs(path)
score_log_file_name = os.path.join(self.path, "score.txt")
torch.manual_seed(self.args.seed)
torch.cuda.manual_seed(self.args.seed)
torch.cuda.manual_seed_all(self.args.seed)
torch.backends.cudnn.deterministic = True
random.seed(self.args.seed)
np.random.seed(self.args.seed)
# load the data
dataset = data_dict[self.args.data_name](self.args)
print("================ {} Mode ====================".format(dataset.exp_mode))
print("================ {} CV ======================".format(dataset.num_of_cv))
num_of_cv = dataset.num_of_cv
for iter in range(num_of_cv):
torch.manual_seed(self.args.seed)
torch.cuda.manual_seed(self.args.seed)
torch.cuda.manual_seed_all(self.args.seed)
torch.backends.cudnn.deterministic = True
random.seed(self.args.seed)
np.random.seed(self.args.seed)
g = torch.Generator()
g.manual_seed(self.args.seed)
torch.backends.cudnn.benchmark = False
os.environ["CUDA_LAUNCH_BLOCKING"] = "1"
os.environ["CUBLAS_WORKSPACE_CONFIG"] = ":16:8"
print("================ the {} th CV Experiment ================ ".format(iter))
dataset.update_train_val_test_keys()
cv_path = os.path.join(self.path,"cv_{}".format(iter))
# get the loader of train val test
train_loader = self._get_data(dataset, flag = 'train', weighted_sampler = self.args.weighted_sampler )
val_loader = self._get_data(dataset, flag = 'vali', weighted_sampler = self.args.weighted_sampler)
test_loader = self._get_data(dataset, flag = 'test', weighted_sampler = self.args.weighted_sampler)
#class_weights=torch.tensor(dataset.act_weights,dtype=torch.double).to(self.device)
train_steps = len(train_loader)
if not os.path.exists(cv_path):
os.makedirs(cv_path)
skip_train = False
skip_finetuning = False
else:
file_in_folder = os.listdir(cv_path)
if 'final_best_vali.pth' in file_in_folder:
skip_train = True
else:
skip_train = False
if 'final_finetuned_best_vali.pth' in file_in_folder:
skip_finetuning = True
else:
skip_finetuning = False
epoch_log_file_name = os.path.join(cv_path, "epoch_log.txt")
if skip_train:
print("================Skip the {} CV Experiment================".format(iter))
else:
if os.path.exists(epoch_log_file_name):
os.remove(epoch_log_file_name)
epoch_log = open(epoch_log_file_name, "a")
score_log = open(score_log_file_name, "a")
print("================ Build the model ================ ")
self.model = self.build_model().to(self.device)
early_stopping = EarlyStopping(patience=self.args.early_stop_patience, verbose=True)
learning_rate_adapter = adjust_learning_rate_class(self.args,True)
model_optim = self._select_optimizer()
#if self.args.weighted == True:
# criterion = nn.CrossEntropyLoss(reduction="mean",weight=class_weights).to(self.device)#self._select_criterion()
#else:
# criterion = nn.CrossEntropyLoss(reduction="mean").to(self.device)#self._select_criterion()
criterion = nn.CrossEntropyLoss(reduction="mean").to(self.device)
criterion = MixUpLoss(criterion)
print("-----------------")
for epoch in range(self.args.train_epochs):
print(" IF you see this, there is no Training")
break
train_loss = []
self.model.train()
epoch_time = time.time()
for i, (batch_x1,batch_x2,batch_y) in enumerate(train_loader):
#if "cross" in self.args.model_type:
# batch_x1 = batch_x1.double().to(self.device)
# batch_x2 = batch_x2.double().to(self.device)
# batch_y = batch_y.long().to(self.device)
# # model prediction
# if self.args.output_attention:
# outputs = self.model(batch_x1,batch_x2)[0]
# else:
# outputs = self.model(batch_x1,batch_x2)
#else:
batch_x1 = batch_x1.double().to(self.device) #--
batch_y = batch_y.long().to(self.device) #--
# if self.args.mixup:
# batch_x1, batch_y = mixup_data(batch_x1, batch_y, self.args.alpha)
# # model prediction
# if self.args.output_attention:
# outputs = self.model(batch_x1)[0]
# else:
outputs = self.model(batch_x1) #--
#print("outputs ", outputs.shape)
#if self.args.mixup:
# criterion = MixUpLoss(criterion)
# loss = criterion(outputs, batch_y)
#else:
loss = criterion(outputs, batch_y) #--
if self.args.wavelet_filtering and self.args.wavelet_filtering_regularization:
reg_loss = 0
for name,parameter in self.model.named_parameters():
if "gamma" in name:
reg_loss += torch.sum(torch.abs(parameter))
loss = loss + self.args.regulatization_tradeoff*reg_loss
train_loss.append(loss.item())
model_optim.zero_grad()
loss.backward()
model_optim.step()
print("Epoch: {} cost time: {}".format(epoch+1, time.time()-epoch_time))
epoch_log.write("Epoch: {} cost time: {}".format(epoch+1, time.time()-epoch_time))
epoch_log.write("\n")
train_loss = np.average(train_loss)
vali_loss , vali_acc, vali_f_w, vali_f_macro, vali_f_micro = self.validation(self.model, val_loader, criterion)
print("VALI: Epoch: {0}, Steps: {1} | Train Loss: {2:.7f} Vali Loss: {3:.7f} Vali Accuracy: {4:.7f} Vali weighted F1: {5:.7f} Vali macro F1 {6:.7f} ".format(
epoch + 1, train_steps, train_loss, vali_loss, vali_acc, vali_f_w, vali_f_macro))
epoch_log.write("VALI: Epoch: {0}, Steps: {1} | Train Loss: {2:.7f} Vali Loss: {3:.7f} Vali Accuracy: {4:.7f} Vali weighted F1: {5:.7f} Vali macro F1 {6:.7f} \n".format(
epoch + 1, train_steps, train_loss, vali_loss, vali_acc, vali_f_w, vali_f_macro))
early_stopping(vali_loss, self.model, cv_path, vali_f_macro, vali_f_w, epoch_log)
if early_stopping.early_stop:
print("Early stopping")
break
epoch_log.write("----------------------------------------------------------------------------------------\n")
epoch_log.flush()
learning_rate_adapter(model_optim,vali_loss)
# rename the best_vali to final_best_vali
os.rename(cv_path+'/'+'best_vali.pth', cv_path+'/'+'final_best_vali.pth')
print("Loading the best validation model!")
self.model.load_state_dict(torch.load(cv_path+'/'+'final_best_vali.pth'))
#model.eval()
test_loss , test_acc, test_f_w, test_f_macro, test_f_micro = self.validation(self.model, test_loader, criterion, iter+1)
print("Final Test Performance : Test Accuracy: {0:.7f} Test weighted F1: {1:.7f} Test macro F1 {2:.7f} ".format (test_acc, test_f_w, test_f_macro))
epoch_log.write("Final Test Performance : Test weighted F1: {0:.7f} Test macro F1 {1:.7f}\n\n\n\n\n\n\n\n".format(test_f_w, test_f_macro))
epoch_log.flush()
score_log.write("Test weighted F1: {0:.7f} Test macro F1 {1:.7f}\n".format(test_f_w, test_f_macro))
score_log.flush()
epoch_log.close()
score_log.close()
# ------------------------------ code for regularization and fine tuning -----------------------------------------------------------------
if self.args.wavelet_filtering_finetuning:
finetuned_score_log_file_name = os.path.join(self.path, "finetuned_score.txt")
if skip_finetuning:
print("================Skip the {} CV Experiment Fine Tuning================".format(iter))
else:
# thre_index : selected number
epoch_log = open(epoch_log_file_name, "a")
epoch_log.write("----------------------------------------------------------------------------------------\n")
epoch_log.write("--------------------------------------Fine Tuning-----------------------------------------\n")
epoch_log.write("----------------------------------------------------------------------------------------\n")
self.model = self.build_model().to(self.device)
self.model.load_state_dict(torch.load(cv_path+'/'+'final_best_vali.pth'))
finetuned_score_log = open(finetuned_score_log_file_name, "a")
thre_index = int(self.args.f_in * self.args.wavelet_filtering_finetuning_percent)-1
gamma_weight = self.model.gamma.squeeze().abs().clone()
sorted_gamma_weight, i = torch.sort(gamma_weight,descending=True)
threshold = sorted_gamma_weight[thre_index]
mask = gamma_weight.data.gt(threshold).float().to(self.device)
idx0 = np.squeeze(np.argwhere(np.asarray(mask.cpu().numpy())))
# build the new model
new_model = model_builder(self.args, input_f_channel = thre_index).to(self.device)
print("------------Fine Tuning : ", self.args.f_in-thre_index," will be pruned -----------------------------------------")
print("old model Parameter :", self.model_size)
print("pruned model Parameter :", np.sum([para.numel() for para in new_model.parameters()]))
print("----------------------------------------------------------------------------------------")
# copy the weights
flag_channel_selection = False
for n,p in new_model.named_parameters():
if "wavelet_conv" in n:
p.data = self.model.state_dict()[n].data[idx0.tolist(), :,:,:].clone()
elif n == "gamma":
flag_channel_selection = True
p.data = self.model.state_dict()[n].data[:, idx0.tolist(),:,:].clone()
elif flag_channel_selection and "conv" in n:
p.data = self.model.state_dict()[n].data[:, idx0.tolist(),:,:].clone()
flag_channel_selection = False
else:
p.data = self.model.state_dict()[n].data.clone()
early_stopping = EarlyStopping(patience=15, verbose=True)
learning_rate_adapter = adjust_learning_rate_class(self.args,True)
model_optim = optim.Adam(new_model.parameters(), lr=0.0001)
criterion = nn.CrossEntropyLoss(reduction="mean").to(self.device)
for epoch in range(self.args.train_epochs):
train_loss = []
new_model.train()
epoch_time = time.time()
for i, (batch_x1,batch_x2,batch_y) in enumerate(train_loader):
batch_x1 = batch_x1.double().to(self.device)
batch_y = batch_y.long().to(self.device)
outputs = new_model(batch_x1)
loss = criterion(outputs, batch_y)
train_loss.append(loss.item())
model_optim.zero_grad()
loss.backward()
model_optim.step()
print("Fine Tuning Epoch: {} cost time: {}".format(epoch+1, time.time()-epoch_time))
epoch_log.write("Fine Tuning Epoch: {} cost time: {}".format(epoch+1, time.time()-epoch_time))
epoch_log.write("\n")
train_loss = np.average(train_loss)
vali_loss , vali_acc, vali_f_w, vali_f_macro, vali_f_micro = self.validation(new_model, val_loader, criterion)
print("Fine Tuning VALI: Epoch: {0}, Steps: {1} | Train Loss: {2:.7f} Vali Loss: {3:.7f} Vali Accuracy: {4:.7f} Vali weighted F1: {5:.7f} Vali macro F1 {6:.7f} ".format(
epoch + 1, train_steps, train_loss, vali_loss, vali_acc, vali_f_w, vali_f_macro))
epoch_log.write("Fine Tuning VALI: Epoch: {0}, Steps: {1} | Train Loss: {2:.7f} Vali Loss: {3:.7f} Vali Accuracy: {4:.7f} Vali weighted F1: {5:.7f} Vali macro F1 {6:.7f} \n".format(
epoch + 1, train_steps, train_loss, vali_loss, vali_acc, vali_f_w, vali_f_macro))
early_stopping(vali_loss, new_model, cv_path, vali_f_macro, vali_f_w, epoch_log)
if early_stopping.early_stop:
print("Early stopping")
break
epoch_log.write("----------------------------------------------------------------------------------------\n")
epoch_log.flush()
learning_rate_adapter(model_optim,vali_loss)
# rename the best_vali to final_best_vali
os.rename(cv_path+'/'+'best_vali.pth', cv_path+'/'+'final_finetuned_best_vali.pth')
print("Loading the best finetuned validation model!")
new_model.load_state_dict(torch.load(cv_path+'/'+'final_finetuned_best_vali.pth'))
test_loss , test_acc, test_f_w, test_f_macro, test_f_micro = self.validation(new_model, test_loader, criterion)
print("Fine Tuning Final Test Performance : Test Accuracy: {0:.7f} Test weighted F1: {1:.7f} Test macro F1 {2:.7f} ".format (test_acc, test_f_w, test_f_macro))
epoch_log.write("Final Test Performance : Test weighted F1: {0:.7f} Test macro F1 {1:.7f}\n\n\n\n\n\n\n\n".format(test_f_w, test_f_macro))
epoch_log.flush()
finetuned_score_log.write("Test weighted F1: {0:.7f} Test macro F1 {1:.7f}\n".format(test_f_w, test_f_macro))
finetuned_score_log.flush()
epoch_log.close()
finetuned_score_log.close()
def prediction_test(self):
assert self.args.exp_mode == "Given"
model = self.build_model().to(self.device)
model.load_state_dict(torch.load(os.path.join(self.path,'cv_0/best_vali.pth')))
model.eval()
dataset = data_dict[self.args.data_name](self.args)
dataset.update_train_val_test_keys()
test_loader = self._get_data(dataset, flag = 'test')
preds = []
trues = []
for i, (batch_x1,batch_x2,batch_y) in enumerate(test_loader):
if "cross" in self.args.model_type:
batch_x1 = batch_x1.double().to(self.device)
batch_x2 = batch_x2.double().to(self.device)
batch_y = batch_y.long().to(self.device)
# model prediction
if self.args.output_attention:
outputs = self.model(batch_x1,batch_x2)[0]
else:
outputs = self.model(batch_x1,batch_x2)
else:
batch_x1 = batch_x1.double().to(self.device)
batch_y = batch_y.long().to(self.device)
# model prediction
if self.args.output_attention:
outputs = self.model(batch_x1)[0]
else:
outputs = self.model(batch_x1)
preds.extend(list(np.argmax(outputs.detach().cpu().numpy(),axis=1)))
trues.extend(list(batch_y.detach().cpu().numpy()))
acc = accuracy_score(preds,trues)
f_w = f1_score(trues, preds, average='weighted')
f_macro = f1_score(trues, preds, average='macro')
f_micro = f1_score(trues, preds, average='micro')
return preds,trues
def validation(self, model, data_loader, criterion, index_of_cv=None, selected_index = None):
model.eval()
total_loss = []
preds = []
trues = []
with torch.no_grad():
for i, (batch_x1,batch_x2,batch_y) in enumerate(data_loader):
if "cross" in self.args.model_type:
batch_x1 = batch_x1.double().to(self.device)
batch_x2 = batch_x2.double().to(self.device)
batch_y = batch_y.long().to(self.device)
# model prediction
if self.args.output_attention:
outputs = model(batch_x1,batch_x2)[0]
else:
outputs = model(batch_x1,batch_x2)
else:
if selected_index is None:
batch_x1 = batch_x1.double().to(self.device)
else:
batch_x1 = batch_x1[:, selected_index.tolist(),:,:].double().to(self.device)
batch_y = batch_y.long().to(self.device)
# model prediction
if self.args.output_attention:
outputs = model(batch_x1)[0]
else:
outputs = model(batch_x1)
pred = outputs.detach()#.cpu()
true = batch_y.detach()#.cpu()
loss = criterion(pred, true)
total_loss.append(loss.cpu())
preds.extend(list(np.argmax(outputs.detach().cpu().numpy(),axis=1)))
trues.extend(list(batch_y.detach().cpu().numpy()))
total_loss = np.average(total_loss)
acc = accuracy_score(preds,trues)
#f_1 = f1_score(trues, preds)
# print("------------------11111111111111---------------")
# import pickle
# with open('preds.pickle', 'wb') as handle:
# pickle.dump(preds, handle, protocol=pickle.HIGHEST_PROTOCOL)
# with open('trues.pickle', 'wb') as handle:
# pickle.dump(trues, handle, protocol=pickle.HIGHEST_PROTOCOL)
f_w = f1_score(trues, preds, average='weighted')
f_macro = f1_score(trues, preds, average='macro')
f_micro = f1_score(trues, preds, average='micro')
if index_of_cv:
cf_matrix = confusion_matrix(trues, preds)
#with open("{}.npy".format(index_of_cv), 'wb') as f:
# np.save(f, cf_matrix)
plt.figure()
sns.heatmap(cf_matrix, annot=True)
#plt.savefig("{}.png".format(index_of_cv))
model.train()
return total_loss, acc, f_w, f_macro, f_micro#, f_1