-
Notifications
You must be signed in to change notification settings - Fork 3
/
train.py
276 lines (245 loc) · 11.3 KB
/
train.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
from __future__ import print_function
import argparse
import torch
import pdb
import time
import yaml
import os
import os.path as osp
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
from lib.utils.net_utils import weights_normal_init, adjust_learning_rate
from configs import cfg
from lib import *
from ptflops import get_model_complexity_info
from thop import profile
def compute_accuracy(output, target, topk=(1,)):
"""Computes the accuracy over the k top predictions for the specified values of k"""
with torch.no_grad():
maxk = max(topk)
batch_size = target.size(0)
_, pred = output.topk(maxk, 1, True, True)
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)
res.append(correct_k.mul_(100.0 / batch_size))
return res
def graph_node_loss(graphs, target):
loss = 0
for graph in graphs:
node = graph[0]
B, N = node.shape
mean_node = torch.mm(node, node.transpose(0, 1).contiguous())
loss += (mean_node.sum() - torch.diag(mean_node).sum()) / N / N
return loss
def graph_edge_loss(graphs, target):
loss = 0
for graph in graphs:
edge = graph[1]
B, N, _ = edge.shape
edge = F.relu(edge)
loss += (edge.mean(0).sum() - torch.diag(edge.mean(0)).sum()) / N / N
return loss
def train(args, model, device, train_loader, optimizer, epoch):
model.train()
batch_time = AverageMeter('Time', ':3.3f')
data_time = AverageMeter('Data', ':3.3f')
mem_cost = AverageMeter('Mem', ':3.3f')
losses = AverageMeter('Loss', ':.3f')
top1 = AverageMeter('Acc@1', ':3.2f')
top5 = AverageMeter('Acc@5', ':3.2f')
layer1_bf = AverageMeter('Corr@bf', ':3.2f')
layer1_af = AverageMeter('Corr@af', ':3.2f')
layer2_bf = AverageMeter('Corr@bf', ':3.2f')
layer2_af = AverageMeter('Corr@af', ':3.2f')
layer3_bf = AverageMeter('Corr@bf', ':3.2f')
layer3_af = AverageMeter('Corr@af', ':3.2f')
progress = ProgressMeter(len(train_loader), mem_cost, batch_time, data_time, losses, top1,
top5, layer1_bf, layer1_af, layer2_bf, layer2_af, layer3_bf, layer3_af,
prefix="Epoch: [{}]".format(epoch))
end = time.time()
for batch_idx, (data, target) in enumerate(train_loader):
data_time.update(time.time() - end)
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
mem = torch.cuda.max_memory_allocated()
mem_cost.update(mem / 1024 / 1024 / 1024)
if model.net.name == "invcnn":
loss = 0
for out in output:
out = F.log_softmax(out, dim=1)
loss += F.nll_loss(out, target)
loss /= len(output)
else:
output = F.log_softmax(output, dim=1)
loss = F.nll_loss(output, target)
acc1, acc5 = compute_accuracy(output[-1] if model.net.name == "invcnn" else output, target, topk=(1, 5))
losses.update(loss.item(), data.size(0))
top1.update(acc1[0], data.size(0))
top5.update(acc5[0], data.size(0))
layer1_bf.update(model.net.layer1.cn.corr_bf.item())
layer1_af.update(model.net.layer1.cn.corr_af.item())
layer2_bf.update(model.net.layer2.cn.corr_bf.item())
layer2_af.update(model.net.layer2.cn.corr_af.item())
layer3_bf.update(model.net.layer3.cn.corr_bf.item())
layer3_af.update(model.net.layer3.cn.corr_af.item())
# import pdb; pdb.set_trace()
# node_loss = graph_node_loss(graphs, target)
# edge_loss = graph_edge_loss(graphs, target)
loss.backward()
optimizer.step()
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
if batch_idx % args.log.print_interval == 0:
progress.print(batch_idx)
# print('Train Epoch: {} [{}/{} ({:.0f}%)]\tmem: {:.3f}m\tLoss: {:.6f}'.format(
# epoch, batch_idx * len(data), len(train_loader.dataset), mem / 1024 / 1024,
# 100. * batch_idx / len(train_loader), loss.item()))
def test(args, model, device, test_loader):
batch_time = AverageMeter('Time', ':3.3f')
losses = AverageMeter('Loss', ':.3f')
mem_cost = AverageMeter('Mem', ':3.3f')
top1 = AverageMeter('Acc@1', ':3.2f')
top5 = AverageMeter('Acc@5', ':3.2f')
progress = ProgressMeter(len(test_loader), mem_cost, batch_time, losses, top1, top5,
prefix='Test: ')
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
end = time.time()
for batch_idx, (data, target) in enumerate(test_loader):
data, target = data.to(device), target.to(device)
output = model(data)
if model.net.name == "invcnn":
output = torch.stack(output, 1).max(1)[0]
mem = torch.cuda.max_memory_allocated()
mem_cost.update(mem / 1024 / 1024 / 1024)
# measure accuracy and record loss
acc1, acc5 = compute_accuracy(output, target, topk=(1, 5))
top1.update(acc1[0], data.size(0))
top5.update(acc5[0], data.size(0))
# measure elapsed time
batch_time.update(time.time() - end)
end = time.time()
# compute val loss
output = F.log_softmax(output, dim=1)
loss = F.nll_loss(output, target).item() # sum up batch loss
test_loss += loss
losses.update(loss, data.size(0))
# compute accuracy
pred = output.max(1, keepdim=True)[1] # get the index of the max log-probability
correct += pred.eq(target.view_as(pred)).sum().item()
if batch_idx % args.log.print_interval == 0:
progress.print(batch_idx)
test_loss /= batch_idx
data_size = len(test_loader.dataset) if hasattr(test_loader, 'dataset') else len(test_loader)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.3f}%)\n'.format(
test_loss, correct, data_size,
100. * float(correct) / data_size))
accuracy = 100. * float(correct) / data_size
print(' * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f}'
.format(top1=top1, top5=top5))
return accuracy
def count_parameters(model):
return sum(p.numel() for p in model.parameters() if p.requires_grad)
def main():
# Training settings
parser = argparse.ArgumentParser(description='PyTorch Image Classification')
parser.add_argument('--dataset', type=str, default='cifar100',
help='specify training dataset')
parser.add_argument('--session', type=int, default='1',
help='training session to recoder multiple runs')
parser.add_argument('--arch', type=str, default='resnet110',
help='specify network architecture')
parser.add_argument('--bs', dest="batch_size", type=int, default=128,
help='training batch size')
parser.add_argument('--gpu0-bs', dest="gpu0_bs", type=int, default=0,
help='training batch size on gpu0')
parser.add_argument('--add-ccn', type=str, default='no',
help='add cross neruon communication')
parser.add_argument('--mgpus', type=str, default="no",
help='multi-gpu training')
parser.add_argument('--resume', dest="resume", type=int, default=0,
help='resume epoch')
args = parser.parse_args()
cfg.merge_from_file(osp.join("configs", args.dataset + ".yaml"))
cfg.dataset = args.dataset
cfg.arch = args.arch
cfg.add_cross_neuron = True if args.add_ccn == "yes" else False
use_cuda = True if torch.cuda.is_available() else False
cfg.use_cuda = use_cuda
cfg.training.batch_size = args.batch_size
cfg.mGPUs = True if args.mgpus == "yes" else False
torch.manual_seed(cfg.initialize.seed)
device = torch.device("cuda" if use_cuda else "cpu")
train_loader, test_loader = create_data_loader(cfg)
model = CrossNeuronNet(cfg)
print("parameter numer: %d" % (count_parameters(model)))
with torch.cuda.device(0):
if args.dataset == "cifar100":
flops, params = get_model_complexity_info(model, (3, 32, 32), as_strings=True, print_per_layer_stat=True)
# flops, params = profile(model, input_size=(1, 3, 32, 32))
elif args.dataset == "imagenet":
flops, params = get_model_complexity_info(model, (3, 224, 224), as_strings=True, print_per_layer_stat=True)
# flops, params = profile(model, input_size=(1, 3, 224, 224))
print('Flops: {}'.format(flops))
print('Params: {}'.format(params))
model = model.to(device)
# optimizer_policy = model.get_optim_policies()
optimizer = optim.SGD(model.parameters(), lr=cfg.optimizer.lr, momentum=cfg.optimizer.momentum, weight_decay=cfg.optimizer.weight_decay)
# optimizer = optim.Adam(model.parameters(), lr=1e-3)
if cfg.mGPUs:
if args.gpu0_bs > 0:
model = BalancedDataParallel(args.gpu0_bs, model).to(device)
else:
model = nn.DataParallel(model).to(device)
lr = cfg.optimizer.lr
checkpoint_tag = osp.join("checkponts", args.dataset, args.arch)
if not osp.exists(checkpoint_tag):
os.makedirs(checkpoint_tag)
if args.resume > 0:
ckpt_path = osp.join(checkpoint_tag,
("ccn" if cfg.add_cross_neuron else "plain") + "_{}_{}.pth".format(args.session, args.resume))
print("resume model from {}".format(ckpt_path))
ckpt = torch.load(ckpt_path)
model.load_state_dict(ckpt["model"])
print("resume model succesfully")
acc = test(cfg, model, device, test_loader)
best_acc = 0
for epoch in range(args.resume + 1, cfg.optimizer.max_epoch + 1):
if epoch in cfg.optimizer.lr_decay_schedule:
adjust_learning_rate(optimizer, cfg.optimizer.lr_decay_gamma)
lr *= cfg.optimizer.lr_decay_gamma
print('Train Epoch: {} learning rate: {}'.format(epoch, lr))
tic = time.time()
train(cfg, model, device, train_loader, optimizer, epoch)
acc = test(cfg, model, device, test_loader)
time_cost = time.time() - tic
if acc > best_acc:
best_acc = acc
print('\nModel: {} Best Accuracy-Baseline: {}\tTime Cost per Epoch: {}\n'.format(
checkpoint_tag + ("ccn" if args.add_ccn == "yes" else "plain"),
best_acc,
time_cost))
if epoch % cfg.log.checkpoint_interval == 0:
checkpoint = {"arch": cfg.arch,
"model": model.state_dict(),
"epoch": epoch,
"lr": lr,
"test_acc": acc,
"best_acc": best_acc}
torch.save(checkpoint, osp.join(checkpoint_tag,
("ccn" if cfg.add_cross_neuron else "plain") + "_{}_{}.pth".format(args.session, epoch)))
if __name__ == '__main__':
torch.manual_seed(1)
torch.cuda.manual_seed(1)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
main()