-
Notifications
You must be signed in to change notification settings - Fork 1
/
train_cifar_tiny_imagenet.py
208 lines (166 loc) · 9.04 KB
/
train_cifar_tiny_imagenet.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
'''
FORCE
Copyright (c) 2020-present NAVER Corp.
MIT license
'''
import torch
import torch.nn as nn
import torch.nn.functional as F
from tensorboardX import SummaryWriter
from ignite.engine import create_supervised_evaluator
from ignite.metrics import Accuracy, Loss
from pruning.pruning_algos import iterative_pruning
from experiments.experiments import *
from pruning.mask_networks import apply_prune_mask
import os
import argparse
import random
# from IPython import embed
def parseArgs():
parser = argparse.ArgumentParser(
description="Training CIFAR / Tiny-Imagenet.",
formatter_class=argparse.ArgumentDefaultsHelpFormatter)
parser.add_argument("--pruning_factor", type=float, default=0.01, dest="pruning_factor",
help='Fraction of connections after pruning')
parser.add_argument("--prune_method", type=int, default=3, dest="prune_method",
help="""Which pruning method to use:
1->Iter SNIP
2->GRASP-It
3->FORCE (default). """)
parser.add_argument("--dataset", type=str, default='CIFAR10',
dest="dataset_name", help='Dataset to train on')
parser.add_argument("--network_name", type=str, default='resnet50', dest="network_name",
help='Model to train')
parser.add_argument("--num_steps", type=int, default=10,
help='Number of steps to use with iterative pruning')
parser.add_argument("--mode", type=str, default='exp',
help='Mode of creating the iterative pruning steps one of "linear" or "exp".')
parser.add_argument("--num_batches", type=int, default=1,
help='''Number of batches to be used when computing the gradient.
If set to -1 they will be averaged over the whole dataset.''')
parser.add_argument("--save_interval", type=int, default=50,
dest="save_interval", help="Number of epochs between model checkpoints.")
parser.add_argument("--save_loc", type=str, default='saved_models/',
dest="save_loc", help='Path where to save the model')
parser.add_argument("--opt", type=str, default='sgd',
dest="optimiser",
help='Choice of optimisation algorithm')
parser.add_argument("--saved_model_name", type=str, default="cnn.model",
dest="saved_model_name", help="Filename of the pre-trained model")
parser.add_argument("--frac-train-data", type=float, default=0.9, dest="frac_data_for_train",
help='Fraction of data used for training (only applied in CIFAR)')
parser.add_argument("--init", type=str, default='normal_kaiming',
help='Which initialization method to use')
parser.add_argument("--in_planes", type=int, default=64,
help='''Number of input planes in Resnet. Afterwards they duplicate after
each conv with stride 2 as usual.''')
return parser.parse_args()
LOG_INTERVAL = 20
REPEAT_WITH_DIFFERENT_SEED = 3 # Number of initialize-prune-train trials (minimum of 1)
torch.backends.cudnn.deterministic = True
torch.backends.cudnn.benchmark = False
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# New additions
args = parseArgs()
def train(seed):
# Set manual seed
torch.manual_seed(seed)
if 'resnet' in args.network_name:
stable_resnet = False
if 'stable' in args.network_name:
stable_resnet = True
if 'CIFAR' in args.dataset_name:
[net, optimiser, lr_scheduler,
train_loader, val_loader,
test_loader, loss, EPOCHS] = resnet_cifar_experiment(device, args.network_name,
args.dataset_name, args.optimiser,
args.frac_data_for_train,
stable_resnet, args.in_planes)
elif 'tiny_imagenet' in args.dataset_name:
[net, optimiser, lr_scheduler,
train_loader, val_loader,
test_loader, loss, EPOCHS] = resnet_tiny_imagenet_experiment(device, args.network_name,
args.dataset_name, args.in_planes)
elif 'vgg' in args.network_name or 'VGG' in args.network_name:
if 'tiny_imagenet' in args.dataset_name:
[net, optimiser, lr_scheduler,
train_loader, val_loader,
test_loader, loss, EPOCHS] = vgg_tiny_imagenet_experiment(device, args.network_name,
args.dataset_name)
else:
[net, optimiser, lr_scheduler,
train_loader, val_loader,
test_loader, loss, EPOCHS] = vgg_cifar_experiment(device, args.network_name,
args.dataset_name, args.frac_data_for_train)
# Initialize network
for layer in net.modules():
if isinstance(layer, nn.Conv2d) or isinstance(layer, nn.Linear):
if args.init == 'normal_kaiming':
nn.init.kaiming_normal_(layer.weight, nonlinearity='relu')
elif args.init == 'normal_kaiming_fout':
nn.init.kaiming_normal_(layer.weight, nonlinearity='relu', mode='fan_out')
elif args.init == 'normal_xavier':
nn.init.xavier_normal_(layer.weight)
elif args.init == 'orthogonal':
nn.init.orthogonal_(layer.weight)
else:
raise ValueError(f"Unrecognised initialisation parameter {args.init}")
############################################################################
#################### Pruning at init ########################
############################################################################
pruning_factor = args.pruning_factor
keep_masks=[]
if pruning_factor != 1:
print(f'Pruning network iteratively for {args.num_steps} steps')
keep_masks = iterative_pruning(net, train_loader, device, pruning_factor,
prune_method=args.prune_method,
num_steps=args.num_steps,
mode=args.mode, num_batches=args.num_batches)
apply_prune_mask(net, keep_masks)
filename = f'iter_prun_{args.num_steps}'
############################################################################
#################### Training ########################
############################################################################
evaluator = create_supervised_evaluator(net, {
'accuracy': Accuracy(),
'cross_entropy': Loss(loss)
}, device)
run_name = (args.network_name + '_' + args.dataset_name + '_spars' +
str(1 - pruning_factor) + '_variant' + str(args.prune_method) +
'_train-frac' + str(args.frac_data_for_train) +
f'_steps{args.num_steps}_{args.mode}' + f'_{args.init}' +
f'_batch{args.num_batches}' + f'_rseed_{seed}')
writer_name= 'runs/' + run_name
writer = SummaryWriter(writer_name)
iterations = 0
for epoch in range(0, EPOCHS):
lr_scheduler.step()
train_loss = train_cross_entropy(epoch, net, train_loader, optimiser, device,
writer, LOG_INTERVAL=20)
iterations +=len(train_loader)
# Evaluate
evaluator.run(test_loader)
metrics = evaluator.state.metrics
# Save history
avg_accuracy = metrics['accuracy']
avg_cross_entropy = metrics['cross_entropy']
writer.add_scalar("test/loss", avg_cross_entropy, iterations)
writer.add_scalar("test/accuracy", avg_accuracy, iterations)
# Save model checkpoints
if (epoch + 1) % args.save_interval == 0:
if not os.path.exists(args.save_loc):
os.makedirs(args.save_loc)
save_name = args.save_loc + run_name + '_cross_entropy_' + str(epoch + 1) + '.model'
torch.save(net.state_dict(), save_name)
elif (epoch + 1) == EPOCHS:
if not os.path.exists(args.save_loc):
os.makedirs(args.save_loc)
save_name = args.save_loc + run_name + '_cross_entropy_' + str(epoch + 1) + '.model'
torch.save(net.state_dict(), save_name)
if __name__ == '__main__':
# Randomly pick a random seed for the experiment
# Multiply the number of seeds to be sampled by 300 so there is wide range of seeds
seeds = list(range(300 * REPEAT_WITH_DIFFERENT_SEED))
random.shuffle(seeds)
for seed in seeds[:REPEAT_WITH_DIFFERENT_SEED]:
train(seed)