-
Notifications
You must be signed in to change notification settings - Fork 6
/
train_bg.py
285 lines (246 loc) · 11.7 KB
/
train_bg.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
import argparse
import os
import time
import random
import logging
import itertools
import numpy as np
import torch
import torch.nn as nn
import torch.backends.cudnn as cudnn
import torch.nn.functional as F
from models.resnet import load_model
from utils import AverageMeter, save_checkpoint, accuracy
from datasets.color_mnist import get_biased_mnist_dataloader
from datasets.cub_dataset import get_waterbird_dataloader
from datasets.celebA_dataset import get_celebA_dataloader
import math
parser = argparse.ArgumentParser(description=' use resnet (pretrained)')
parser.add_argument('--in-dataset', default="celebA", type=str, choices = ['celebA', 'color_mnist', 'waterbird'], help='in-distribution dataset e.g. IN-9')
parser.add_argument('--model-arch', default='resnet18', type=str, help='model architecture e.g. resnet50')
parser.add_argument('--domain-num', default=4, type=int,
help='the number of environments for model training')
parser.add_argument('--method', default='erm', type=str, help='method used for model training')
parser.add_argument('--save-epoch', default=5, type=int,
help='save the model every save_epoch, default = 10')
parser.add_argument('--print-freq', '-p', default=10, type=int,
help='print frequency (default: 10)')
# ID train & val batch size
parser.add_argument('-b', '--batch-size', default=256, type=int,
help='mini-batch size (default: 64) used for training')
# training schedule
parser.add_argument('--start-epoch', default=0, type=int,
help='manual epoch number (useful on restarts)')
parser.add_argument('--epochs', default=30, type=int,
help='number of total epochs to run, default = 30')
parser.add_argument('--lr', '--learning-rate', default=1e-4, type=float,
help='initial learning rate')
parser.add_argument('--num-classes', default=2, type=int,
help='number of classes for model training')
parser.add_argument('--momentum', default=0.9, type=float, help='momentum')
parser.add_argument('--weight-decay', '--wd', default=0.005, type=float,
help='weight decay (default: 0.0001)')
parser.add_argument('--data_label_correlation', default=0.9, type=float,
help='data_label_correlation')
# saving, naming and logging
parser.add_argument('--exp-name', default = 'erm_new_0.9', type=str,
help='help identify checkpoint')
parser.add_argument('--name', default="erm_rebuttal", type=str,
help='name of experiment')
parser.add_argument('--log_name', type = str, default = "info.log",
help='Name of the Log File')
# Device options
parser.add_argument('--gpu-ids', default='6', type=str,
help='id(s) for CUDA_VISIBLE_DEVICES')
parser.add_argument('--local_rank', default=-1, type=int,
help='rank for the current node')
# Miscs
parser.add_argument('--manualSeed', type=int, help='manual seed')
parser.add_argument('--n-g-nets', default=1, type=int,
help="the number of networks for g_model in ReBias")
parser.add_argument('--penalty-multiplier', default=1.1, type=float,
help="the penalty multiplier used in IRM training")
parser.add_argument('--cosine', action='store_false',
help='using cosine annealing')
parser.add_argument('--lr_decay_epochs', type=str, default='15,25',
help=' 15, 25, 40 for waterbibrds; 10, 15 ,20 for color_mnist')
parser.add_argument('--lr_decay_rate', type=float, default=0.1,
help='decay rate for learning rate')
args = parser.parse_args()
state = {k: v for k, v in args._get_kwargs()}
directory = "checkpoints/{in_dataset}/{name}/{exp}/".format(in_dataset=args.in_dataset,
name=args.name, exp=args.exp_name)
os.makedirs(directory, exist_ok=True)
save_state_file = os.path.join(directory, 'args.txt')
fw = open(save_state_file, 'w')
print(state, file=fw)
fw.close()
# CUDA Specification
os.environ['CUDA_VISIBLE_DEVICES'] = args.gpu_ids
if torch.cuda.is_available():
torch.cuda.set_device(args.local_rank)
device = torch.device(f"cuda" if torch.cuda.is_available() else "cpu")
# Set random seed
def set_random_seed(seed):
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
if args.manualSeed is None:
args.manualSeed = random.randint(1, 10000)
set_random_seed(args.manualSeed)
def flatten(list_of_lists):
return itertools.chain.from_iterable(list_of_lists)
def train(model, train_loaders, criterion, optimizer, epoch, log):
"""Train for one epoch on the training set"""
batch_time = AverageMeter()
nat_losses = AverageMeter()
nat_top1 = AverageMeter()
# switch to train mode
model.train()
end = time.time()
batch_idx = 0
train_loaders = [iter(x) for x in train_loaders]
len_dataloader = 0
for x in train_loaders:
len_dataloader += len(x)
while True:
for loader in train_loaders:
input, target, _ = next(loader, (None, None, None))
if input is None:
return
input = input.cuda()
target = target.cuda()
_, nat_output = model(input)
nat_loss = criterion(nat_output, target)
# measure accuracy and record loss
nat_prec1 = accuracy(nat_output.data, target, topk=(1,))[0]
nat_losses.update(nat_loss.data, input.size(0))
nat_top1.update(nat_prec1, input.size(0))
# compute gradient and do SGD step
loss = nat_loss
optimizer.zero_grad()
loss.backward()
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if batch_idx % 10 == 0:
log.debug('Epoch: [{0}][{1}/{2}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})'.format(
epoch, batch_idx, len_dataloader, batch_time=batch_time,
loss=nat_losses, top1=nat_top1))
batch_idx += 1
def validate(val_loader, model, criterion, epoch, log, method):
"""Perform validation on the validation set"""
batch_time = AverageMeter()
losses = AverageMeter()
top1 = AverageMeter()
# switch to evaluate mode
model.eval()
with torch.no_grad():
end = time.time()
for i, (input, target, _) in enumerate(val_loader):
input = input.cuda()
target = target.cuda()
# compute output
_, output = model(input)
loss = criterion(output, target)
# measure accuracy and record loss
prec1 = accuracy(output.data, target, topk=(1,))[0]
losses.update(loss.data, input.size(0))
top1.update(prec1, input.size(0))
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if i % args.print_freq == 0:
log.debug('Validate: [{0}/{1}]\t'
'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
'Prec@1 {top1.val:.3f} ({top1.avg:.3f})'.format(
i, len(val_loader), batch_time=batch_time, loss=losses,
top1=top1))
log.debug(' * Prec@1 {top1.avg:.3f}'.format(top1=top1))
return top1.avg
def adjust_learning_rate(args, optimizer, epoch):
lr = args.lr
if args.cosine:
eta_min = lr * (args.lr_decay_rate ** 3)
lr = eta_min + (lr - eta_min) * (
1 + math.cos(math.pi * epoch / args.epochs)) / 2
else:
steps = np.sum(epoch > np.asarray(args.lr_decay_epochs))
if steps > 0:
lr = lr * (args.lr_decay_rate ** steps)
for param_group in optimizer.param_groups:
param_group['lr'] = lr
def main():
log = logging.getLogger(__name__)
formatter = logging.Formatter('%(asctime)s : %(message)s')
fileHandler = logging.FileHandler(os.path.join(directory, args.log_name), mode='w')
fileHandler.setFormatter(formatter)
streamHandler = logging.StreamHandler()
streamHandler.setFormatter(formatter)
log.setLevel(logging.DEBUG)
log.addHandler(fileHandler)
log.addHandler(streamHandler)
if args.in_dataset == "color_mnist":
train_loader1 = get_biased_mnist_dataloader(args, root = './datasets/MNIST', batch_size=args.batch_size,
data_label_correlation= args.data_label_correlation,
n_confusing_labels= args.num_classes - 1,
train=True, partial=True, cmap = "1")
train_loader2 = get_biased_mnist_dataloader(args, root = './datasets/MNIST', batch_size=args.batch_size,
data_label_correlation= args.data_label_correlation,
n_confusing_labels= args.num_classes - 1,
train=True, partial=True, cmap = "2")
val_loader = get_biased_mnist_dataloader(args, root = './datasets/MNIST', batch_size=args.batch_size,
data_label_correlation= args.data_label_correlation,
n_confusing_labels= args.num_classes - 1,
train=False, partial=True, cmap = "1")
elif args.in_dataset == "waterbird":
train_loader = get_waterbird_dataloader(args, data_label_correlation=args.data_label_correlation, split="train")
val_loader = get_waterbird_dataloader(args, data_label_correlation=args.data_label_correlation, split="val")
elif args.in_dataset == "celebA":
train_loader = get_celebA_dataloader(args, split="train")
val_loader = get_celebA_dataloader(args, split="val")
if args.model_arch == 'resnet18':
pretrained = True
if args.in_dataset == 'color_mnist':
pretrained = False #True for celebA & waterbird ; False for Color_MNIST
base_model = load_model(pretrained)
if torch.cuda.device_count() > 1:
base_model = torch.nn.DataParallel(base_model)
if args.method == "erm":
model = base_model.cuda()
criterion = nn.CrossEntropyLoss().cuda()
optimizer = torch.optim.Adam(model.parameters(), lr=args.lr, weight_decay=args.weight_decay)
else:
assert False, 'Not supported method: {}'.format(args.method)
cudnn.benchmark = True
freeze_bn_affine = False
def freeze_bn(model, freeze_bn_affine=True):
for m in model.modules():
if isinstance(m, nn.BatchNorm2d):
m.eval()
if freeze_bn_affine:
m.weight.requires_grad = False
m.bias.requires_grad = False
freeze_bn(model, freeze_bn_affine)
if args.in_dataset == "color_mnist":
train_loaders = [train_loader1, train_loader2]
elif args.in_dataset == "waterbird" or args.in_dataset == "celebA":
train_loaders = [train_loader]
for epoch in range(args.start_epoch, args.epochs):
print(f"Start training epoch {epoch}")
adjust_learning_rate(args, optimizer, epoch)
train(model, train_loaders, criterion, optimizer, epoch, log)
prec1 = validate(val_loader, model, criterion, epoch, log, args.method)
if (epoch + 1) % args.save_epoch == 0:
save_checkpoint(args, {
'epoch': epoch + 1,
'state_dict_model': model.state_dict(),
}, epoch + 1)
if __name__ == '__main__':
main()