-
Notifications
You must be signed in to change notification settings - Fork 18
/
Copy pathmain.py
172 lines (148 loc) · 7.97 KB
/
main.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
from __future__ import print_function
import argparse
import torch
import torch.nn.functional as F
from optimizer import PruneAdam
from model import LeNet, AlexNet
from utils import regularized_nll_loss, admm_loss, \
initialize_Z_and_U, update_X, update_Z, update_Z_l1, update_U, \
print_convergence, print_prune, apply_prune, apply_l1_prune
from torchvision import datasets, transforms
from tqdm import tqdm
def train(args, model, device, train_loader, test_loader, optimizer):
for epoch in range(args.num_pre_epochs):
print('Pre epoch: {}'.format(epoch + 1))
model.train()
for batch_idx, (data, target) in enumerate(tqdm(train_loader)):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = regularized_nll_loss(args, model, output, target)
loss.backward()
optimizer.step()
test(args, model, device, test_loader)
Z, U = initialize_Z_and_U(model)
for epoch in range(args.num_epochs):
model.train()
print('Epoch: {}'.format(epoch + 1))
for batch_idx, (data, target) in enumerate(tqdm(train_loader)):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = admm_loss(args, device, model, Z, U, output, target)
loss.backward()
optimizer.step()
X = update_X(model)
Z = update_Z_l1(X, U, args) if args.l1 else update_Z(X, U, args)
U = update_U(U, X, Z)
print_convergence(model, X, Z)
test(args, model, device, test_loader)
def test(args, model, device, test_loader):
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in test_loader:
data, target = data.to(device), target.to(device)
output = model(data)
test_loss += F.nll_loss(output, target, reduction='sum').item() # sum up batch loss
pred = output.argmax(dim=1, keepdim=True) # get the index of the max log-probability
correct += pred.eq(target.view_as(pred)).sum().item()
test_loss /= len(test_loader.dataset)
print('\nTest set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)\n'.format(
test_loss, correct, len(test_loader.dataset),
100. * correct / len(test_loader.dataset)))
def retrain(args, model, mask, device, train_loader, test_loader, optimizer):
for epoch in range(args.num_re_epochs):
print('Re epoch: {}'.format(epoch + 1))
model.train()
for batch_idx, (data, target) in enumerate(tqdm(train_loader)):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
optimizer.prune_step(mask)
test(args, model, device, test_loader)
def main():
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser.add_argument('--dataset', type=str, default="mnist", choices=["mnist", "cifar10"],
metavar='D', help='training dataset (mnist or cifar10)')
parser.add_argument('--batch-size', type=int, default=64, metavar='N',
help='input batch size for training (default: 64)')
parser.add_argument('--test-batch-size', type=int, default=1000, metavar='N',
help='input batch size for testing (default: 1000)')
parser.add_argument('--percent', type=list, default=[0.8, 0.92, 0.991, 0.93],
metavar='P', help='pruning percentage (default: 0.8)')
parser.add_argument('--alpha', type=float, default=5e-4, metavar='L',
help='l2 norm weight (default: 5e-4)')
parser.add_argument('--rho', type=float, default=1e-2, metavar='R',
help='cardinality weight (default: 1e-2)')
parser.add_argument('--l1', default=False, action='store_true',
help='prune weights with l1 regularization instead of cardinality')
parser.add_argument('--l2', default=False, action='store_true',
help='apply l2 regularization')
parser.add_argument('--num_pre_epochs', type=int, default=3, metavar='P',
help='number of epochs to pretrain (default: 3)')
parser.add_argument('--num_epochs', type=int, default=10, metavar='N',
help='number of epochs to train (default: 10)')
parser.add_argument('--num_re_epochs', type=int, default=3, metavar='R',
help='number of epochs to retrain (default: 3)')
parser.add_argument('--lr', type=float, default=1e-3, metavar='LR',
help='learning rate (default: 1e-2)')
parser.add_argument('--adam_epsilon', type=float, default=1e-8, metavar='E',
help='adam epsilon (default: 1e-8)')
parser.add_argument('--no-cuda', action='store_true', default=False,
help='disables CUDA training')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--save-model', action='store_true', default=False,
help='For Saving the current Model')
args = parser.parse_args()
use_cuda = not args.no_cuda and torch.cuda.is_available()
torch.manual_seed(args.seed)
device = torch.device("cuda" if use_cuda else "cpu")
kwargs = {'num_workers': 1, 'pin_memory': True} if use_cuda else {}
if args.dataset == "mnist":
train_loader = torch.utils.data.DataLoader(
datasets.MNIST('data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=args.batch_size, shuffle=True, **kwargs)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST('data', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=args.test_batch_size, shuffle=True, **kwargs)
else:
args.percent = [0.8, 0.92, 0.93, 0.94, 0.95, 0.99, 0.99, 0.93]
args.num_pre_epochs = 5
args.num_epochs = 20
args.num_re_epochs = 5
train_loader = torch.utils.data.DataLoader(
datasets.CIFAR10('data', train=True, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.49139968, 0.48215827, 0.44653124),
(0.24703233, 0.24348505, 0.26158768))
])), shuffle=True, batch_size=args.batch_size, **kwargs)
test_loader = torch.utils.data.DataLoader(
datasets.CIFAR10('data', train=False, download=True,
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.49139968, 0.48215827, 0.44653124),
(0.24703233, 0.24348505, 0.26158768))
])), shuffle=True, batch_size=args.test_batch_size, **kwargs)
model = LeNet().to(device) if args.dataset == "mnist" else AlexNet().to(device)
optimizer = PruneAdam(model.named_parameters(), lr=args.lr, eps=args.adam_epsilon)
train(args, model, device, train_loader, test_loader, optimizer)
mask = apply_l1_prune(model, device, args) if args.l1 else apply_prune(model, device, args)
print_prune(model)
test(args, model, device, test_loader)
retrain(args, model, mask, device, train_loader, test_loader, optimizer)
if __name__ == "__main__":
main()