-
Notifications
You must be signed in to change notification settings - Fork 1
/
training.py
536 lines (465 loc) · 22.5 KB
/
training.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.backends.cudnn as cudnn
import torch.optim as optim
# import torchvision
import torchvision.transforms as transforms
from models import MTANet
from dataset import Aff2
import myloss
import numpy as np
import argparse
import os
from torch.utils.data import DataLoader, RandomSampler
import shutil
import time
import logging
import math
import sklearn.metrics as sm
from models.metrics import ArcMarginProduct
from utils import prep_experiment, print_eval
parser = argparse.ArgumentParser(description='MTANet Training')
parser.add_argument('-a', '--arch', type=str, default='resnext', metavar='ARCH', help='Using model')
parser.add_argument('--bs', type=int, default=80, help='Batch size')
parser.add_argument('--lr', type=float, default=0.001, help='Learning rate')
parser.add_argument('--weight_decay', type=float, default=1e-4)
parser.add_argument('--epochs', type=int, default=100, help='Epoch')
parser.add_argument('--workers', type=int, default=0, help='Dataloader num_worker')
parser.add_argument('--poly_exp', type=float, default=1.0, help='Polynomial LR exponent')
parser.add_argument('--resume', type=str, default='', metavar='PATH', help='Path to checkpoint')
parser.add_argument('--start_epoch', type=int, default=0, metavar='N', help='Manual epoch number')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true', help='Evaluate mode on validation set')
parser.add_argument('-p', '--print_freq', type=int, default=10, metavar='N', help='Print frequency')
parser.add_argument('--ckpt_path', type=str, default='logs/ckpt')
parser.add_argument('--tb_path', type=str, default='logs/tb')
parser.add_argument('--exp', type=str, default='exp', help='experiment directory name')
parser.add_argument('--tb_tag', type=str, default='', help='add tag to tb dir')
parser.add_argument('--loss_type', type=str, default='learned', help='fixed or learned')
parser.add_argument('-arc', '--arcface', dest='arcface', action='store_true', help='Using acrface metric')
parser.add_argument('--optim', type=str, default='sgd', help='SGD or Adam')
parser.add_argument('-smp', '--sampler', dest='sampler', action='store_true', help='Using sampler')
args = parser.parse_args()
args.best_record = {'epoch': -1, 'val_loss': 1e10, 'best_va': 1e10, 'best_acc1': 0,
'best_au_strict': 0, 'best_expr_f1': 0, 'best_au_f1': 0}
size = (112, 112)
initial_ses = [1., 1., 1.]
cudnn.benchmark = True
device = torch.device('cuda')
net = MTANet.aff2net(initial_ses=initial_ses, arc_face=args.arcface, backbone=args.arch)
net = net.to(device)
if args.arcface:
metric_fc = ArcMarginProduct(1024, 7, s=30, m=0.5, easy_margin=False)
metric_fc.to(device)
def main():
best_acc1 = 0
best_va = 0
best_au_strict = 0
best_loss = 1e10
best_expr_f1 = 0
best_au_f1 = 0
final_cm = 0
final_mcm = 0
# setup optimizer
params = filter(lambda p: p.requires_grad, net.parameters())
if args.arcface:
params = [{'params': params}, {'params': metric_fc.parameters()}]
if args.optim == 'sgd':
# [{'params': params}, {'params': metric_fc.parameters()}]
optimizer = optim.SGD(params, lr=args.lr, momentum=0.9, weight_decay=args.weight_decay, nesterov=True)
else:
# [{'params': params}, {'params': metric_fc.parameters()}]
optimizer = optim.Adam(params, lr=args.lr, weight_decay=args.weight_decay)
poly_lambda = lambda epoch: math.pow(1 - epoch / args.epochs, args.poly_exp)
scheduler = optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=poly_lambda)
criterion = myloss.MultiTaskLoss(loss_type=args.loss_type, loss_uncertainties=net.get_loss_weights(), gamma=2)
# if args.resume:
if os.path.isfile(args.resume):
print("=> loading checkpoint '{}'".format(args.resume))
checkpoint = torch.load(args.resume)
best_acc1 = checkpoint['best_acc1']
best_loss = checkpoint['best_loss']
best_va = checkpoint['best_va']
best_au_strict = checkpoint['bse_au_strict']
best_expr_f1 = checkpoint['best_expr_f1']
best_au_f1 = checkpoint['best_au_f1']
net.load_state_dict(checkpoint['state_dict'], strict=False)
optimizer.load_state_dict(checkpoint['optimizer'])
print("=> loaded checkpoint '{}' (epoch {})".format(args.resume, checkpoint['epoch']))
else:
print("=> no checkpoint found at '{}'".format(args.resume))
# Data loading
transform_train = transforms.Compose([
transforms.Resize(size), # follow VGGface's input 224*224 3C RGB
# transforms.RandomHorizontalFlip(),
transforms.ToTensor(), # range [0, 255] -> [0.0, 1.0]
transforms.Normalize([0.5, 0.5, 0.5],
[0.5, 0.5, 0.5]), # range [0.0, 1.0] -> [-1.0, 1.0]
])
transform_val = transforms.Compose([
transforms.Resize(size),
transforms.ToTensor(),
transforms.Normalize([0.5, 0.5, 0.5],
[0.5, 0.5, 0.5]), # range [0.0, 1.0] -> [-1.0, 1.0]
])
train_set = Aff2(transform=transform_train, flag="train")
val_set = Aff2(transform=transform_val, flag="val")
if args.sampler:
train_sampler = RandomSampler(train_set, True, int(1e5))
val_sampler = RandomSampler(val_set, True, int(2e4))
shuffle = False
else:
train_sampler = None
val_sampler = None
shuffle = True
train_loader = DataLoader(train_set, batch_size=args.bs, shuffle=shuffle, sampler=train_sampler,
num_workers=args.workers, pin_memory=True)
val_loader = DataLoader(val_set, batch_size=args.bs, shuffle=shuffle, sampler=val_sampler,
num_workers=args.workers, pin_memory=True)
writer = prep_experiment(args)
if args.evaluate:
loss, loss_va, top1, au_strict, cm, mcm, expr_f1, au_f1 = validate(val_loader, net, criterion, 0, args, writer)
args.best_record['val_loss'] = loss.avg
args.best_record['best_acc1'] = top1.avg
args.best_record['best_va'] = loss_va.avg
args.best_record['best_au_strict'] = au_strict.avg
args.best_record['best_expr_f1'] = expr_f1.avg
args.best_record['best_au_f1'] = au_f1.avg
print_eval(args)
np.save(os.path.join(args.exp_path, "CM.npy"), np.array(final_cm))
np.save(os.path.join(args.exp_path, "MCM.npy"), np.array(final_mcm))
return
for epoch in range(args.start_epoch, args.epochs):
train(train_loader, net, criterion, optimizer, epoch, args, writer)
# evaluate on validation set
loss, loss_va, top1, au_strict, cm, mcm, expr_f1, au_f1 = validate(val_loader, net, criterion, epoch, args, writer)
scheduler.step(epoch)
# remember best acc@1 and save checkpoint
is_best = ((loss_va.avg > best_va) & (loss.avg < best_loss)) | \
((au_f1.avg > best_au_f1) & (expr_f1.avg > best_expr_f1))
best_loss = min(loss.avg, best_loss)
best_acc1 = max(top1.avg, best_acc1)
best_va = max(loss_va.avg, best_va)
best_au_strict = max(au_strict.avg, best_au_strict)
best_expr_f1 = max(expr_f1.avg, best_expr_f1)
best_au_f1 = max(au_f1.avg, best_au_f1)
final_cm += cm
final_mcm += mcm
if is_best:
logging.info("Got best model for now. Saving model...")
args.best_record['epoch'] = epoch + 1
args.best_record['val_loss'] = best_loss
args.best_record['best_acc1'] = best_acc1
args.best_record['best_va'] = best_va
args.best_record['best_au_strict'] = best_au_strict
args.best_record['best_expr_f1'] = best_expr_f1
args.best_record['best_au_f1'] = best_au_f1
print_eval(args)
np.save(os.path.join(args.exp_path, "CM.npy"), np.array(final_cm))
np.save(os.path.join(args.exp_path, "MCM.npy"), np.array(final_mcm))
save_checkpoint({
'epoch': epoch + 1,
'arch': args.arch,
'state_dict': net.state_dict(),
'best_loss': best_loss,
'best_acc1': best_acc1,
'best_va': best_va,
'bse_au_strict': best_au_strict,
'best_expr_f1': best_expr_f1,
'best_au_f1': best_au_f1,
'optimizer': optimizer.state_dict(),
}, is_best, epoch + 1)
def train(train_loader, model, criterion, optimizer, epoch, args, writer):
batch_time = AverageMeter('Time', ':6.3f')
data_time = AverageMeter('Data', ':6.3f')
losses = AverageMeter('Total Loss', ':.4f')
loss_va = AverageMeter('VA Loss', ':.4f')
loss_au = AverageMeter('AU Loss', ':.4f')
loss_expr = AverageMeter('Expr Loss', ':.4f')
top1 = AverageMeter('Expr Acc@1', ':6.2f')
v_ccc = AverageMeter('V CCC', ':6.2f')
a_ccc = AverageMeter('A CCC', ':6.2f')
au_strict = AverageMeter('AU Strict Acc', ':6.2f')
f1_expr = AverageMeter('Expr F1-score', ':6.2f')
f1_aus = AverageMeter('AU F1-score', ':6.2f')
progress = ProgressMeter(
len(train_loader),
[batch_time, data_time, losses, loss_va, loss_au, loss_expr, v_ccc, a_ccc, top1, f1_expr, au_strict, f1_aus], # , au_category
prefix="Train Epoch: [{}]".format(epoch + 1)
)
# switch to training mode
curr_iter = epoch * len(train_loader)
model.train()
end = time.time()
for batch_index, (images, labels) in enumerate(train_loader):
# measure data loading time
data_time.update(time.time() - end)
images = images.to(device) # , torch.tensor(labels).to(device)
labels = labels.to(device)
v_labels = labels[:, 0]
a_labels = labels[:, 1]
aus_labels = labels[:, 2:-1]
expr_labels = labels[:, -1]
# compute output
outputs = model(images)
va_output, aus_output, expr_output = outputs
if args.arcface:
expr_output, expr_labels = metric_fc(expr_output, expr_labels)
outputs = (va_output, aus_output, expr_output)
loss, va_bs, au_bs, (va_loss, v_loss, a_loss, au_loss, expr_loss) = \
criterion(outputs, v_labels, a_labels, aus_labels, expr_labels)
# compute gradient and do SGD step
optimizer.zero_grad()
loss.backward()
optimizer.step()
# measure accuracy and record loss
batch_size = images.size(0)
(acc1, acc3), expr_bs, expr_f1 = expr_accuracy(expr_output, expr_labels, topk=(1, 3))
losses.update(loss.item(), batch_size)
loss_va.update(va_loss.item(), va_bs)
loss_au.update(au_loss.item(), au_bs)
loss_expr.update(expr_loss.item(), expr_bs)
top1.update(acc1[0], expr_bs)
v_ccc.update(v_loss, va_bs)
a_ccc.update(a_loss, va_bs)
strict_acc, au_f1 = aus_accuracy(aus_output, aus_labels, threshold=0.5)
au_strict.update(strict_acc, au_bs)
f1_expr.update(expr_f1, expr_bs)
f1_aus.update(au_f1, au_bs)
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
curr_iter += 1
lr = optimizer.state_dict()["param_groups"][-1]['lr']
w1, w2, w3 = model.get_loss_weights()
if batch_index % args.print_freq == 0 or batch_index == len(train_loader) - 1:
weight_info = "\tw1={:0.4f}\tw2={:0.4f}\tw3={:0.4f}".format(w1.item(), w2.item(), w3.item())
info = progress.info(batch_index + 1) + "\tlr={}".format(lr) + weight_info
logging.info(info)
# Log tensorboard metrics for each iteration of the training phase
writer.add_scalar('training/loss', losses.val, curr_iter) # (losses.val)
writer.add_scalar('training/lr', lr, curr_iter) # param_groups[-1]['lr']
writer.add_scalar('training/va_loss', loss_va.val, curr_iter)
writer.add_scalar('training/au_loss', loss_au.val, curr_iter)
writer.add_scalar('training/expr_loss', loss_expr.val, curr_iter)
writer.add_scalar('training/top1', top1.val, curr_iter)
writer.add_scalar('training/au_strict', au_strict.val, curr_iter)
def validate(val_loader, model, criterion, epoch, args, writer):
batch_time = AverageMeter('Time', ':6.3f')
losses = AverageMeter('Loss', ':.4f')
loss_va = AverageMeter('VA CCC', ':.4f')
loss_au = AverageMeter('AU Loss', ':.4f')
loss_expr = AverageMeter('Expr Loss', ':.4f')
top1 = AverageMeter('Acc@1', ':6.2f')
v_ccc = AverageMeter('V CCC', ':6.2f')
a_ccc = AverageMeter('A CCC', ':6.2f')
au_strict = AverageMeter('AU Strict Acc', ':6.2f')
f1_expr = AverageMeter('Expr F1-score', ':6.2f')
f1_aus = AverageMeter('AU F1-score', ':6.2f')
progress = ProgressMeter(
len(val_loader),
[batch_time, losses, loss_va, v_ccc, a_ccc, top1, f1_expr, au_strict, f1_aus], # , au_category
prefix="Validation Epoch: [{}]".format(epoch + 1)
)
curr_iter = epoch * len(val_loader)
cm = 0
mcm = 0
model.eval()
with torch.no_grad():
end = time.time()
for batch_index, (images, labels) in enumerate(val_loader):
images = images.to(device)
v_labels = labels[:, 0].to(device)
a_labels = labels[:, 1].to(device)
aus_labels = labels[:, 2:-1].to(device)
expr_labels = labels[:, -1].to(device)
# compute output
outputs = model(images)
va_output, aus_output, expr_output = outputs
if args.arcface:
expr_output, expr_labels = metric_fc(expr_output, expr_labels)
outputs = (va_output, aus_output, expr_output)
loss, va_bs, au_bs, (va_loss, v_loss, a_loss, au_loss, expr_loss) = \
criterion(outputs, v_labels, a_labels, aus_labels, expr_labels)
# measure accuracy and record loss
batch_size = images.size(0)
(acc1, acc3), expr_bs, expr_cm, expr_prcn, expr_rcl, expr_f1 = \
expr_accuracy(expr_output, expr_labels, topk=(1, 3), flag="val")
# using total batch size as losses' batch size could result lower displayed loss.
losses.update(loss.item(), batch_size)
loss_va.update(1 - va_loss.item(), va_bs)
loss_au.update(au_loss.item(), au_bs)
loss_expr.update(expr_loss.item(), expr_bs)
top1.update(acc1[0], expr_bs)
v_ccc.update(v_loss, va_bs)
a_ccc.update(a_loss, va_bs)
f1_expr.update(expr_f1, expr_bs)
strict_acc, au_mcm, au_prcn, au_rcl, au_f1, cgr_acc = \
aus_accuracy(aus_output, aus_labels, threshold=0.5, flag="val")
au_strict.update(strict_acc, au_bs)
f1_aus.update(au_f1, au_bs)
cm += expr_cm
mcm += au_mcm
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
curr_iter += 1
if batch_index % args.print_freq == 0 or batch_index == len(val_loader) - 1:
info = progress.info(batch_index + 1)
logging.info(info)
writer.add_scalar('validating/loss', losses.val, curr_iter) # (losses.val)
writer.add_scalar('validating/va_ccc', loss_va.val, curr_iter)
writer.add_scalar('validating/au_loss', loss_au.val, curr_iter)
writer.add_scalar('validating/expr_loss', loss_expr.val, curr_iter)
writer.add_scalar('validating/top1', top1.val, curr_iter)
writer.add_scalar('validating/au_strict', au_strict.val, curr_iter)
if expr_prcn != -1 and expr_rcl != -1 and expr_f1 != -1:
writer.add_scalar('validating/expr_prcn', expr_prcn, curr_iter)
writer.add_scalar('validating/expr_rcl', expr_rcl, curr_iter)
writer.add_scalar('validating/expr_f1', expr_f1, curr_iter)
if au_prcn != -1 and au_rcl != -1 and au_f1 != -1:
writer.add_scalar('validating/au_prcn', au_prcn, curr_iter)
writer.add_scalar('validating/au_rcl', au_rcl, curr_iter)
writer.add_scalar('validating/au_f1', au_f1, curr_iter)
return losses, loss_va, top1, au_strict, cm, mcm, f1_expr, f1_aus
class AverageMeter(object):
"""Computes and stores the average and current value"""
def __init__(self, name, fmt=':f'):
self.name = name
self.fmt = fmt
self.reset()
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
if self.count == 0: self.count = 1e-10
self.avg = self.sum / self.count
def __str__(self):
fmtstr = '{name}: {val' + self.fmt + '} ({avg' + self.fmt + '})'
return fmtstr.format(**self.__dict__)
class ProgressMeter(object):
def __init__(self, num_batches, meters, prefix=""):
self.batch_fmtstr = self._get_batch_fmtstr(num_batches)
self.meters = meters
self.prefix = prefix
def info(self, batch):
entries = [self.prefix + self.batch_fmtstr.format(batch)]
entries += [str(meter) for meter in self.meters]
return '\t'.join(entries)
def display(self, batch):
entries = [self.prefix + self.batch_fmtstr.format(batch)]
entries += [str(meter) for meter in self.meters]
print('\t'.join(entries))
def _get_batch_fmtstr(self, num_batches):
num_digits = len(str(num_batches // 1))
fmt = '{:' + str(num_digits) + 'd}'
return '[' + fmt + '/' + fmt.format(num_batches) + ']'
def adjust_learning_rate(optimizer, epoch, args):
"""Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
lr = args.lr * (0.1 ** (epoch // 30))
optimizer.state_dict()["param_groups"][-1]['lr'] = lr
def expr_accuracy(output, target, topk=(1,), flag="train"):
"""Computes the accuracy over the k top predictions for the specified values of k"""
with torch.no_grad():
maxk = max(topk)
# remove ignore_index in target and output
ignore_index = -1
shape = output.shape
mask = target != ignore_index
target = target[mask]
output = output[mask].reshape(-1, shape[1])
assert target.size(0) == output.size(0)
batch_size = output.size(0)
_, pred = output.topk(maxk, 1, True, True)
top1 = pred[:, 0].cpu().numpy()
target_np = target.view(-1).cpu().numpy()
pred = pred.t()
correct = pred.eq(target.view(1, -1).expand_as(pred))
res = []
for k in topk:
correct_k = correct[:k].view(-1).float().sum(0, keepdim=True)
if batch_size == 0: batch_size = 1e-10
res.append(correct_k.mul_(100.0 / batch_size))
if flag == "val":
if top1.size == 0 and target_np.size == 0:
cm = 0
precision, recall, F1_score = -1, -1, -1
else:
cm = sm.confusion_matrix(target_np, top1, labels=range(7))
precision, recall, F1_score = statistic(target_np, top1)
return res, batch_size, cm, precision, recall, F1_score
else:
if top1.size == 0 and target_np.size == 0:
return res, batch_size, 0
F1_score = sm.f1_score(target_np, top1, average="macro", zero_division=1)
return res, batch_size, F1_score
def aus_accuracy(output, target, threshold=0.5, flag="train"):
with torch.no_grad():
# remove ignore_index in target and output
ignore_index = -1
shape = output.shape
mask = target != ignore_index
target = target[mask].reshape(-1, shape[1])
output = output[mask].reshape(-1, shape[1])
batch_size = output.size(0)
if batch_size == 0: batch_size = 1e10
mhot_output = torch.sigmoid(output) > threshold
# synchronize the matrix type with target
mhot_output = mhot_output.type_as(target)
# strict: must match in every class
mhot_output_np = mhot_output.cpu().numpy()
target_np = target.cpu().numpy()
tp = (mhot_output_np == target_np).sum(0)
ctg_acc = np.nan_to_num(tp / batch_size)
strict_correct = sum(list(map(lambda x, y: torch.equal(x, y), mhot_output, target))) / batch_size
"""
# soft: logical_and divide logical_or = accuracy
logical_and = np.sum(np.logical_and(mhot_output_np, target_np), axis=1)
logical_or = np.sum(np.logical_or(mhot_output_np, target_np), axis=1)
acc_and_or = np.nan_to_num(logical_and / logical_or)
soft_correct = np.nan_to_num(acc_and_or.mean())
"""
if flag == "val":
if mhot_output_np.size == 0 and target_np.size == 0:
mcm = 0
precision, recall, F1_score = -1, -1, -1
else:
mcm = sm.multilabel_confusion_matrix(target_np, mhot_output_np)
precision, recall, F1_score = statistic(target_np, mhot_output_np)
return strict_correct * 100, mcm, precision, recall, F1_score, ctg_acc
else:
if mhot_output_np.size == 0 and target_np.size == 0:
return strict_correct * 100, 0
F1_score = sm.f1_score(target_np, mhot_output_np, average="macro", zero_division=1)
return strict_correct * 100, F1_score
def statistic(target, predict):
precision = sm.precision_score(target, predict, average="macro", zero_division=1)
recall = sm.recall_score(target, predict, average="macro", zero_division=1)
F1_score = sm.f1_score(target, predict, average="macro", zero_division=1)
return precision, recall, F1_score
def fast_hist(pred, true, num_classes):
mask = (true >= 0) & (true < num_classes)
hist = np.bincount(num_classes * true[mask].astype(int) + pred[mask].astype(int),
minlength=num_classes ** 2).reshape(num_classes, num_classes)
tp = np.diag(hist)
fp = hist.sum(axis=1) - tp # fp axis=1 or 0?
fn = hist.sum(axis=0) - tp
precision = tp / (tp + fp)
recall = tp / (tp + fn)
f1 = 2 * precision * recall / (precision + recall)
return hist, precision, recall, f1
def save_checkpoint(state, is_best, epoch):
filename = "ckpt@"+str(epoch)+".pth.tar"
torch.save(state, os.path.join(args.exp_path, filename))
if is_best:
shutil.copyfile(os.path.join(args.exp_path, filename),
os.path.join(args.exp_path, 'model_best@' + str(epoch) + '.pth.tar'))
if __name__ == '__main__':
# os.environ['CUDA_LAUNCH_BLOCKING'] = "1"
# torch.multiprocessing.set_start_method('spawn')
main()