-
Notifications
You must be signed in to change notification settings - Fork 5
/
Copy pathtrain_mnist_rigl.py
185 lines (150 loc) · 7.5 KB
/
train_mnist_rigl.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
"""
Note: This is the exact same script as found here: https://github.com/pytorch/examples/blob/0f0c9131ca5c79d1332dce1f4c06fe942fbdc665/mnist/main.py#L1
The only difference is there are a few parser arguments added and the mandatory rigl-torch code (creating the prune scheduler).
"""
from __future__ import print_function
import os
import argparse
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torch.utils.tensorboard import SummaryWriter
from torchvision import datasets, transforms
from torch.optim.lr_scheduler import StepLR
from rigl_torch.RigL import RigLScheduler
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 32, 3, 1)
self.conv2 = nn.Conv2d(32, 64, 3, 1)
self.dropout1 = nn.Dropout(0.25)
self.dropout2 = nn.Dropout(0.5)
self.fc1 = nn.Linear(9216, 128)
self.fc2 = nn.Linear(128, 10)
def forward(self, x):
x = self.conv1(x)
x = F.relu(x)
x = self.conv2(x)
x = F.relu(x)
x = F.max_pool2d(x, 2)
x = self.dropout1(x)
x = torch.flatten(x, 1)
x = self.fc1(x)
x = F.relu(x)
x = self.dropout2(x)
x = self.fc2(x)
output = F.log_softmax(x, dim=1)
return output
def train(args, model, device, train_loader, optimizer, epoch, pruner):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device)
optimizer.zero_grad()
output = model(data)
loss = F.nll_loss(output, target)
loss.backward()
if pruner():
optimizer.step()
if batch_idx % args.log_interval == 0:
print('Train Epoch: {} [{}/{} ({:.0f}%)]\tLoss: {:.6f}'.format(
epoch, batch_idx * len(data), len(train_loader.dataset),
100. * batch_idx / len(train_loader), loss.item()))
if args.dry_run:
break
def test(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)))
return test_loss, correct / len(test_loader.dataset)
def ed(param_name, default=None):
return os.environ.get(param_name, default)
def main():
# Training settings
parser = argparse.ArgumentParser(description='PyTorch MNIST Example')
parser.add_argument('--dense-allocation', default=ed('DENSE_ALLOCATION'), type=float,
help='percentage of dense parameters allowed. if None, pruning will not be used. must be on the interval (0, 1]')
parser.add_argument('--delta', default=ed('DELTA', 100), type=int,
help='delta param for pruning')
parser.add_argument('--grad-accumulation-n', default=ed('GRAD_ACCUMULATION_N', 1), type=int,
help='number of gradients to accumulate before scoring for rigl')
parser.add_argument('--alpha', default=ed('ALPHA', 0.3), type=float,
help='alpha param for pruning')
parser.add_argument('--static-topo', default=ed('STATIC_TOPO', 0), type=int, help='if 1, use random sparsity topo and remain static')
parser.add_argument('--batch-size', type=int, default=ed('BATCH_SIZE', 64), metavar='N',
help='input batch size for training (default: 64)')
parser.add_argument('--test-batch-size', type=int, default=ed('TEST_BATCH_SIZE', 1000), metavar='N',
help='input batch size for testing (default: 1000)')
parser.add_argument('--epochs', type=int, default=ed('EPOCHS', 14), metavar='N',
help='number of epochs to train (default: 14)')
parser.add_argument('--lr', type=float, default=ed('LR', 1), metavar='LR',
help='learning rate (default: 1.0)')
parser.add_argument('--gamma', type=float, default=ed('GAMMA', 0.7), metavar='M',
help='Learning rate step gamma (default: 0.7)')
parser.add_argument('--no-cuda', action='store_true', default=False,
help='disables CUDA training')
parser.add_argument('--dry-run', action='store_true', default=False,
help='quickly check a single pass')
parser.add_argument('--seed', type=int, default=1, metavar='S',
help='random seed (default: 1)')
parser.add_argument('--log-interval', type=int, default=10, metavar='N',
help='how many batches to wait before logging training status')
parser.add_argument('--save-model', default=1, type=bool,
help='For Saving the current Model')
args = parser.parse_args()
if args.dense_allocation is None:
print('-------------------------------------------------------------------')
print('heads up, RigL will not be used unless `--dense-allocation` is set!')
print('-------------------------------------------------------------------')
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")
train_kwargs = {'batch_size': args.batch_size}
test_kwargs = {'batch_size': args.test_batch_size}
if use_cuda:
cuda_kwargs = {'num_workers': 1,
'pin_memory': True,
'shuffle': True}
train_kwargs.update(cuda_kwargs)
test_kwargs.update(cuda_kwargs)
transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])
dataset1 = datasets.MNIST('../data', train=True, download=True,
transform=transform)
dataset2 = datasets.MNIST('../data', train=False,
transform=transform)
train_loader = torch.utils.data.DataLoader(dataset1,**train_kwargs)
test_loader = torch.utils.data.DataLoader(dataset2, **test_kwargs)
model = Net().to(device)
optimizer = optim.Adadelta(model.parameters(), lr=args.lr)
scheduler = StepLR(optimizer, step_size=1, gamma=args.gamma)
pruner = lambda: True
if args.dense_allocation is not None:
T_end = int(0.75 * args.epochs * len(train_loader))
pruner = RigLScheduler(model, optimizer, dense_allocation=args.dense_allocation, alpha=args.alpha, delta=args.delta, static_topo=args.static_topo, T_end=T_end, ignore_linear_layers=False, grad_accumulation_n=args.grad_accumulation_n)
writer = SummaryWriter(log_dir='./graphs')
print(model)
for epoch in range(1, args.epochs + 1):
print(pruner)
train(args, model, device, train_loader, optimizer, epoch, pruner=pruner)
loss, acc = test(model, device, test_loader)
scheduler.step()
writer.add_scalar('loss', loss, epoch)
writer.add_scalar('accuracy', acc, epoch)
if args.save_model:
torch.save(model.state_dict(), "/artifacts/mnist_cnn.pt")
if __name__ == '__main__':
main()