/
ann.py
402 lines (326 loc) · 15.6 KB
/
ann.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
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torchvision import datasets, transforms, models
from torch.utils.data.dataloader import DataLoader
import torch.backends.cudnn as cudnn
from tensorboard_logger import configure, log_value
import torchvision
#import torchvision.transforms as transforms
import numpy as np
import os
import argparse
import math
from vgg_ann_models import *
#from utils import progress_bar
import time
def rin(input,b=4,s=2):
x=int(((input.shape[2]-b)/s)+1)*b
y=int(((input.shape[3]-b)/s)+1)*b
output = torch.zeros(input.shape[0],input.shape[1],x,y)
m=-1
for i in range(0, input.shape[2] - b + 1, s):
m=m+1
n=-1
for j in range(0, input.shape[3] - b + 1, s):
n=n+1
output[:,:,m*b : (m+1)*b,n*b : (n+1)*b]=input[:, :, i:i+b, j:j+b]
return output
class DCT2(nn.Module):
def __init__(self, block_size=4, p=0, mode = 'random', mean = None, std=None, device = 'cpu'):
super(DCT2, self).__init__()
### forming the cosine transform matrix
self.block_size = block_size
self.device = device
self.mean =mean
self.std =std
self.Q = torch.zeros((self.block_size,self.block_size)).to(self.device)
self.Q[0] = math.sqrt( 1.0/float(self.block_size) )
for i in range (1,self.block_size,1):
for j in range(self.block_size):
self.Q[i,j] = math.sqrt( 2.0/float(self.block_size) ) * math.cos( float((2*j+1)*math.pi*i) /float(2.0*self.block_size) )
def rgb_to_ycbcr(self,input):
# input is mini-batch N x 3 x H x W of an RGB image
#output = Variable(input.data.new(*input.size())).to(self.device)
output = torch.zeros_like(input).to(self.device)
input = (input * 255.0)
output[:, 0, :, :] = input[:, 0, :, :] * 0.299+ input[:, 1, :, :] * 0.587 + input[:, 2, :, :] * 0.114
output[:, 1, :, :] = input[:, 0, :, :] * -0.168736 - input[:, 1, :, :] *0.331264+ input[:, 2, :, :] * 0.5 + 128
output[:, 2, :, :] = input[:, 0, :, :] * 0.5 - input[:, 1, :, :] * 0.418688- input[:, 2, :, :] * 0.081312+ 128
return output/255.0
def ycbcr_to_freq(self,input):
output = torch.zeros_like(input).to(self.device)
a=int(input.shape[2]/self.block_size)
b=int(input.shape[3]/self.block_size)
# Compute DCT in block_size x block_size blocks
for i in range(a):
for j in range(b):
output[:,:,i*self.block_size : (i+1)*self.block_size,j*self.block_size : (j+1)*self.block_size] = torch.matmul(torch.matmul(self.Q, input[:, :, i*self.block_size : (i+1)*self.block_size, j*self.block_size : (j+1)*self.block_size]), self.Q.permute(1,0).contiguous() )
return output
def forward(self, x):
#return self.ycbcr_to_freq( self.rgb_to_ycbcr(x) )
if (x.shape[1]==3):
return self.ycbcr_to_freq( self.rgb_to_ycbcr(x) )
else:
return self.ycbcr_to_freq(x )
class AverageMeter(object):
"""Computes and stores the average and current value"""
def __init__(self):
self.reset()
def reset(self):
self.val = 0
self.avg = 0
self.sum = 0
self.count = 0
def update(self, val, n=1):
self.val = val
self.sum += val * n
self.count += n
self.avg = self.sum / self.count
def adjust_learning_rate(optimizer, epoch):
"""Sets the learning rate to the initial LR decayed by 2 every 30 epochs"""
lr = args.lr * (0.001 ** (epoch // 100))
for param_group in optimizer.param_groups:
param_group['lr'] = lr
parser = argparse.ArgumentParser(description='PyTorch tinyimagenet Training')
parser.add_argument('--lr', default=0.1, type=float, help='learning rate')
parser.add_argument('-b', '--batch_size', default=128, type=int,
metavar='N', help='mini-batch size (default: 128)')
#parser.add_argument('--resume', '-r', action='store_true', help='resume from checkpoint')
parser.add_argument('--seed', default=0, type=int, help='Random seed')
parser.add_argument('--ckpt_dir', default=None, type=str, help='Checkpoint dir. If set to none, default dir is used')
parser.add_argument('--ckpt_intrvl', default=1, type=int, help='Number of epochs between successive checkpoints')
parser.add_argument('--num_epochs', default=312, type=int, help='Number of epochs for backpropagation')
parser.add_argument('--resume_from_ckpt', default=0, type=int, help='Resume from checkpoint?')
parser.add_argument('--tensorboard', default=0, type=int, help='Log progress to TensorBoard')
global args
args = parser.parse_args()
device = 'cuda' if torch.cuda.is_available() else 'cpu'
best_acc = 0 # best test accuracy
# Initialize seed
#--------------------------------------------------
seed = args.seed
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
num_train = 50000
num_test = 10000
img_size = 32
inp_maps = 3
num_cls = 10
test_error_best = 100
start_epoch = 0
num_epochs = args.num_epochs
end_epoch = start_epoch+num_epochs
batch_size = args.batch_size
ckpt_dir = args.ckpt_dir
ckpt_intrvl = args.ckpt_intrvl
resume_from_ckpt = True if args.resume_from_ckpt else False
#model_str_use = 'vgg11_cifar100_ann'+'_bs'+str(batch_size)+'_new_'+str(args.lr)+'lrby5_every30epoch'
model_str_use = 'vgg9_cifar10_ann_lr.1_.1by100'+'_bs'+str(batch_size)+'_pixelexpanded_4avgpool'
#model_str_use = 'vgg13_tinyimgnet_4*4dctbnmaxpool_ann_lr.01_.1by100'+'_bs'+str(batch_size)+'_wd1e-4'
if(ckpt_dir is None):
ckpt_dir = '/home/vgg9_snn_surrgrad_backprop/CHECKPOINTS/'+model_str_use
ckpt_dir = os.path.expanduser(ckpt_dir)
if(ckpt_intrvl > 0):
if(not os.path.exists(ckpt_dir)):
os.mkdir(ckpt_dir)
ckpt_fname = ckpt_dir+'/ckpt.pth'
# Use TensorBoard?
tensorboard = True if args.tensorboard else False
# Data
print('==> Preparing data..')
#dataset = 'tinyIMAGENET' # {'CIFAR10', 'CIFAR100', 'IMAGENET'}
dataset = 'CIFAR10'
#usual
normalize = transforms.Normalize(mean = [0.5, 0.5, 0.5], std = [0.5, 0.5, 0.5])
# usual imgnet stat from repos
#normalize = transforms.Normalize(mean = [0.485, 0.456, 0.406], std = [0.229, 0.224, 0.225])
# calculated itiny-mgnet stat
#normalize = transforms.Normalize(mean = [0.48, 0.448, 0.3975], std = [0.277, 0.269, 0.282])
transform_train = transforms.Compose([
transforms.RandomCrop(32, padding=4),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize])
transform_test = transforms.Compose([transforms.ToTensor(), normalize])
if dataset == 'CIFAR10':
trainset = datasets.CIFAR10(root = './cifar_data', train = True, download = True, transform = transform_train)
testset = datasets.CIFAR10(root='./cifar_data', train=False, download=True, transform= transform_test)
labels = 10
elif dataset == 'CIFAR100':
trainset = datasets.CIFAR100(root = './cifar_data', train = True, download = True, transform = transform_train)
testset = datasets.CIFAR100(root='./cifar_data', train=False, download=True, transform= transform_test)
labels = 100
elif dataset == 'IMAGENET':
labels = 1000
traindir = os.path.join('/local/scratch/a/imagenet/imagenet2012/', 'train')
valdir = os.path.join('/local/scratch/a/imagenet/imagenet2012/', 'val')
trainset = datasets.ImageFolder(
traindir,
transforms.Compose([
transforms.RandomResizedCrop(224),
transforms.RandomHorizontalFlip(),
transforms.ToTensor(),
normalize,
]))
testset = datasets.ImageFolder(
valdir,
transforms.Compose([
transforms.Resize(256),
transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
]))
elif dataset == 'tinyIMAGENET':
labels = 200
# adding the tinyimagenet directory
traindir = os.path.join('/home/nano01/a/banerj11/srinivg_BackProp_CIFAR10/sayeed/tiny-imagenet-200/', 'train')
valdir = os.path.join('/home/nano01/a/banerj11/srinivg_BackProp_CIFAR10/sayeed/tiny-imagenet-200/', 'val')
# traindir = os.path.join('/local/scratch/a/chowdh23/data/tiny-imagenet-200/', 'train')
# valdir = os.path.join('/local/scratch/a/chowdh23/data/tiny-imagenet-200/', 'val')
trainset = datasets.ImageFolder(
traindir,
transforms.Compose([
transforms.RandomResizedCrop(64),
transforms.RandomHorizontalFlip(),
transforms.RandomVerticalFlip(),
transforms.ToTensor(),
normalize,
]))
testset = datasets.ImageFolder(
valdir,
transforms.Compose([
#transforms.Resize(256),
#transforms.CenterCrop(224),
transforms.ToTensor(),
normalize,
]))
trainloader = DataLoader(trainset, batch_size=batch_size, shuffle=True)
testloader = DataLoader(testset, batch_size=batch_size, shuffle=False)
# Model
print('==> Building model..')
model = VGG('VGG9', labels=labels)
# net = ResNet18()
# net = PreActResNet18()
# net = GoogLeNet()
# net = DenseNet121()
# net = ResNeXt29_2x64d()
# net = MobileNet()
# net = MobileNetV2()
# net = DPN92()
# net = ShuffleNetG2()
# net = SENet18()
# net = ShuffleNetV2(1)
#net = EfficientNetB0()
model = model.cuda()
model = torch.nn.DataParallel(model).cuda()
use_cuda =torch.cuda.is_available()
device = torch.device("cuda" if use_cuda else "cpu")
m=DCT2(block_size=4, device = device).to(device)
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=args.lr, momentum=0.9, weight_decay=1e-4)
#optimizer = optim.Adam(model.parameters(), lr=args.lr, weight_decay=5e-4, amsgrad=False)
if(resume_from_ckpt):
ckpt = torch.load(ckpt_fname)
start_epoch = ckpt['start_epoch']
end_epoch = start_epoch+num_epochs
test_error_best = ckpt['test_error_best']
epoch_best = ckpt['epoch_best']
# train_time = ckpt['train_time']
model.load_state_dict(ckpt['model_state_dict'])
optimizer.load_state_dict(ckpt['optim_state_dict'])
print('##### Loaded ANN_VGG from {}\n'.format(ckpt_fname))
print('********** ANN training and evaluation **********')
for epoch in range(start_epoch, end_epoch):
train_loss = AverageMeter()
test_loss = AverageMeter()
# model.use_max_out_over_time = use_max_out_over_time
# model.module.updt_tend(t_end_updt)
model.train()
adjust_learning_rate(optimizer, epoch)
for i, data in enumerate(trainloader):
# print("Epoch: {}/{};".format(epoch+1, end_epoch), "Training batch:{}/{};".format(i+1, math.ceil(num_train/batch_size)))
# start_time = time.time()
# Load the inputs and targets
inputs, targets = data
#targets=torch.from_numpy(np.eye(num_cls)[targets])
inputs, targets = inputs.cuda(), targets.cuda()
if dataset=='CIFAR10' or dataset=='CIFAR100':
inputs =rin(inputs)
#inputs =m(inputs)
# Reset the gradients
optimizer.zero_grad()
# Perform forward pass and compute the target loss
output = model(inputs)
#output= F.softmax(output,dim=1)
#b=targets.float()
loss = criterion(output, targets)
train_loss.update(loss.item(), targets.size(0))
# Perform backward pass and update the weights
loss.backward()
optimizer.step()
# end_time = time.time()
# train_time += (end_time-start_time)/3600
# Print error measures and log progress to TensorBoard
train_loss_per_epoch = train_loss.avg
# print("Epoch: {}/{};".format(epoch+1, end_epoch), "########## Training loss: {}".format(train_loss_per_epoch))
# if(tensorboard):
# log_value('train_loss', train_loss_per_epoch, epoch)
# Evaluate classification accuracy on the test set
# model.use_max_out_over_time = False
# model.module.updt_tend(t_end)
correct_pred_top1 = 0
correct_pred_topk = 0
model.eval()
with torch.no_grad():
for j, data in enumerate(testloader, 0):
# print("Epoch: {}/{};".format(epoch+1, end_epoch), "Test batch: {}/{}".format(j+1, math.ceil(num_test/batch_size)))
images, labels = data
images, labels = images.cuda(), labels.cuda()
if dataset=='CIFAR10' or dataset=='CIFAR100':
images =rin(images)
#images =m(images)
out = model(images)
loss1 = criterion(out, labels)
test_loss.update(loss1.item(), labels.size(0))
_, predicted = out.max(1)
# total += targets.size(0)
correct_pred_top1 += predicted.eq(labels).sum().item()
#print(correct_pred_top1)
# _, pred = out.topk(topk, 1, True, True)
# pred = pred.t()
# correct = pred.eq(labels.view(1, -1).expand_as(pred))
# correct_pred_top1 += correct[:1].view(-1).float().sum(0, keepdim=True)
# correct_pred_topk += correct[:topk].view(-1).float().sum(0, keepdim=True)
test_loss_per_epoch = test_loss.avg
# print("Epoch: {}/{};".format(epoch+1, end_epoch), "########## Test loss: {}".format(test_loss_per_epoch))
if(tensorboard):
log_value('test_loss', test_loss_per_epoch, epoch)
# Print error measures and log progress to TensorBoard
test_error_top1 = (1-(correct_pred_top1/num_test))*100
# test_error_topk = (1-(correct_pred_topk/num_test))*100
test_error_chgd = False
if(test_error_top1 < test_error_best):
test_error_best = test_error_top1
epoch_best = epoch
test_error_chgd = True
print("Epoch: {}/{};".format(epoch_best+1, end_epoch), "########## Test error (top1-best): {:.2f}%".format(test_error_best))
print("Epoch: {}/{};".format(epoch+1, end_epoch), "########## Test error (top1-cur) : {:.2f}%".format(test_error_top1))
# print("Epoch: {}/{};".format(epoch+1, end_epoch), "########## Test error (top"+str(topk)+"-cur) : {:.2f}%".format(test_error_topk[0]))
if(tensorboard):
log_value('test_error (top1-best)', test_error_best, epoch)
log_value('test_error (top1)', test_error_top1, epoch)
# log_value('test_error (top'+str(topk)+')', test_error_topk, epoch)
# Checkpoint SNN training and evaluation states
if((ckpt_intrvl > 0) and ((epoch == end_epoch-1) or test_error_chgd)):
print('=========== Checkpointing ANN training and evaluation states')
ckpt = {'model_state_dict': model.state_dict(),
'optim_state_dict': optimizer.state_dict(),
'start_epoch' : epoch+1,
'test_error_best' : test_error_best,
'epoch_best' : epoch_best}
# 'train_time' : train_time}
if(test_error_chgd):
torch.save(ckpt, ckpt_fname)