-
Notifications
You must be signed in to change notification settings - Fork 0
/
main_minist.py
208 lines (164 loc) · 8.76 KB
/
main_minist.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
import torch
import torch.nn as nn
import torchvision
from torchvision import transforms
from tensorboardX import SummaryWriter
import sys, os
sys.path.append(os.path.join(os.getcwd(), '../optimizer'))
import adampp as adampp
import adam_test1 as adam_test1
from torchvision import datasets, transforms
import torch.nn.functional as F
import argparse
import Sadam as Sadam
#Architecture
class NeuralNet(nn.Module):
def __init__(self):
super(NeuralNet, self).__init__()
#self.conv1 = nn.Conv2d(1, 20, 5)
#self.conv2 = nn.Conv2d(20, 50, 5)
#self.fc1 = nn.Linear(4*4*50, 500)
#self.fc2 = nn.Linear(500, 10)
self.conv1 = nn.Conv2d(1, 6, 5)
self.conv2 = nn.Conv2d(6, 16, 5)
#self.fc1 = nn.Linear(16*5*5, 120)
self.fc1 = nn.Linear(16*4*4, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = F.relu(self.conv1(x))
x = F.max_pool2d(x, 2, 2)
x = F.relu(self.conv2(x))
x = F.max_pool2d(x, 2, 2)
#x = x.view(-1, 4*4*50)
x = x.view(-1, 16*4*4)
x = F.relu(self.fc1(x))
x= F.relu(self.fc2(x))
#x = self.fc2(x)
x = self.fc3(x)
return F.log_softmax(x, dim=1)
net = NeuralNet()
save_net = net
# Test
def test(model, device, test_loader,loss_function):
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 += loss_function(output, target)*len( data) # 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)
test_accuracy = 100. * correct / len(test_loader.dataset)
return test_loss, test_accuracy
def main(opt = 'sgd', lr = 0.05,eps=1e-8, momentum =0, r= 0, partial=1/4, weight_decay=0.0001, amsgrad=True, transformer='linear', test_type='Padam', hist=True, grad_transf = 'square'):
# Training settings
epochs =40
log_interval=100
folder = 'mnist'
#torch.manual_seed(args.seed)
# Device configuration
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
# MNIST dataset
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=100, shuffle=True)
test_loader = torch.utils.data.DataLoader(
datasets.MNIST('../data', train=False, transform=transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])),
batch_size=1000, shuffle=True)
net = NeuralNet().to(device)
# Loss and optimizer
loss_function = nn.CrossEntropyLoss()
if opt == "Padam":
optimizer = Sadam.Sadam(net.parameters(), lr=lr, eps=eps, partial=partial, weight_decay=weight_decay, amsgrad=amsgrad, transformer= transformer, test_type=test_type, hist=hist, grad_transf = grad_transf)
if transformer == 'Padam':
folder_name = '../logs/'+folder+"/"+str(test_type )+"_lr"+str(lr)+'_eps_'+str( eps)+'_weight_decay_'+str( weight_decay)+'_partial_'+str( partial)+'_amsgrad_'+str( amsgrad)+'_denom_transformer_'+str( transformer)+'_grad_transf_'+str( grad_transf)+'_'+str( r )
file_name = "../logs/"+folder+"/"+str(test_type )+"_lr"+str(lr)+'_eps_'+str( eps)+'_weight_decay_'+str( weight_decay)+'_partial_'+str( partial)+'_amsgrad_'+str( amsgrad) +'_denom_transformer_'+str( transformer)+'_grad_transf_'+str( grad_transf)+'_'+str( r )
else:
folder_name = '../logs/'+folder+"/"+str(test_type )+"_lr"+str(lr)+'_eps_'+str( eps)+'_weight_decay_'+str( weight_decay)+'_amsgrad_'+str( amsgrad)+'_denom_transformer_'+str( transformer)+'_grad_transf_'+str( grad_transf)+'_'+str( r )
file_name = "../logs/"+folder+"/"+str(test_type )+"_lr"+str(lr)+'_eps_'+str( eps)+'_weight_decay_'+str( weight_decay)+'_amsgrad_'+str( amsgrad) +'_denom_transformer_'+str( transformer)+'_grad_transf_'+str( grad_transf)+'_'+str( r )
elif opt == "sgd":
optimizer = torch.optim.SGD(net.parameters(), lr = lr, momentum=momentum, weight_decay=weight_decay)
folder_name = '../logs/'+folder+"/sgd_lr"+str(lr)+"_mom"+str( momentum)+'_'+str( r )+'_weight_decay_'+str( weight_decay)
file_name = "../logs/"+folder+"sgd_lr"+str(lr)+"_mom"+str( momentum)+'_'+str( r )+'_weight_decay_'+str( weight_decay)
scheduler = torch.optim.lr_scheduler.ReduceLROnPlateau(optimizer, 'max', 0.1, 10*len(train_loader)/log_interval)
logger = SummaryWriter(folder_name)
file = open(file_name+str(".txt"),"w")
file.write('epoch,iteration,testing_loss,testing_accuracy\n')
#train
data_iter = iter(train_loader)
iter_per_epoch = len(train_loader)
total_step =epochs*iter_per_epoch
# Start training
for step in range(total_step):
# Reset the data_iter
if (step+1) % iter_per_epoch == 0:
data_iter = iter(train_loader)
# Fetch images and labels
images, labels = next(data_iter)
images, labels = images.to(device), labels.to(device)
# Forward pass
outputs = net(images)
loss = loss_function(outputs, labels)
# Backward and optimize
optimizer.zero_grad()
loss.backward()
denom_info= optimizer.step()
# Compute accuracy
_, argmax = torch.max(outputs, 1)
accuracy = (labels == argmax.squeeze()).float().mean()
if (step+1) % log_interval == 0:
print ('Step [{}/{}], Loss: {:.4f}, Acc: {:.2f}'
.format(step+1, total_step, loss.item(), accuracy.item()))
# ================================================================== #
# Tensorboard Logging #
# ================================================================== #
# 1. Log scalar values (scalar summary)
info = { 'training_loss': loss.item(), 'training_accuracy': accuracy.item() }
for tag, value in info.items():
logger.add_scalar(tag, value, step+1)
testing_loss, testing_accuracy = test( net, device, test_loader,loss_function)
logger.add_scalar( 'lr', optimizer.param_groups[0]['lr'], step+1)
scheduler.step( accuracy)
info = { 'testing_loss': testing_loss, 'testing_accuracy': testing_accuracy }
for tag, value in info.items():
logger.add_scalar(tag, value, step+1)
file.write(str((step+1)/log_interval) +','+str(step+1)+','+ str(testing_loss)+','+str(testing_accuracy)+'\n')
if denom_info and denom_info['denom']:
for tag, value in denom_info.items():
logger.add_histogram(tag, torch.cat(value, dim=0).cpu().numpy(), step+1)
#logger.add_histogram(tag+'/grad', value.grad.data.cpu().numpy(), step+1)
for tag, value in net.named_parameters():
tag = tag.replace('.', '/')
logger.add_histogram(tag, value.data.cpu().numpy(), step+1)
logger.add_histogram(tag+'/grad', value.grad.data.cpu().numpy(), step+1)
file.close()
if __name__ == '__main__':
opts = [ 'Padam']#,'sgd']
lrs = [ 0.1, 0.01, 0.001]
for r in range( 1 ):
for opt in opts:
for lr in lrs:
print( "opt : {}, lr : {}".format(opt, lr))
if opt == 'sgd':
for mom in [0.9]:
main( opt , lr, mom, r)
else:
for transf in ['sigmoid']
for amsgrad in [False, True]:
if transf == 'softplus':
main( opt=opt, lr=lr, r=r, amsgrad=amsgrad, transformer=transf, test_type='Padam', hist=True, grad_transf='square')
elif transf == 'Padam':
for par in [1/4, 1/8, 1/16]:
for eps in [1e-3, 1e-8]:
main( opt=opt, lr=lr, eps=eps, r=r, partial=par, amsgrad=amsgrad, transformer=transf, test_type='Padam', hist=True, grad_transf='abs')