-
Notifications
You must be signed in to change notification settings - Fork 1
/
Copy pathmain.py
154 lines (127 loc) · 6.55 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
import argparse
import torch
import torchvision
import torchvision.transforms as transforms
import random
from torch import optim, nn
from train_test.train import train
from train_test.test import test
parser = argparse.ArgumentParser(description='Sparse Double Descent in Vision Transformers: real or phantom threat?')
parser.add_argument('--learning_rate', default=1e-4, type=float, help='learning rate')
parser.add_argument('--num_classes', type=int, default='10', help='10 for CIFAR-10; 100 for CIFAR-100')
parser.add_argument('--model', default='vit', help='vit, swint or resnet-18')
parser.add_argument('--batch_size', type=int, default='512')
parser.add_argument('--epochs', type=int, default='200')
parser.add_argument('--weight_decay', type=float, default='0.03')
parser.add_argument('--amount_noise', type=float, default='0.1')
args = parser.parse_args()
## Using GPU
GPU_id = torch.cuda.current_device()
cuda = "cuda:"+str(GPU_id)
device = torch.device(cuda)
## Preprocessing
transform = transforms.Compose([
transforms.RandomHorizontalFlip(),
transforms.RandomCrop(32,4),
transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))])
## CIFAR Dataset
if args.num_classes ==100:
train_dataset = torchvision.datasets.CIFAR100(root='~/data/cifar-100/',
train=True,
transform=transform,
download=True)
test_dataset = torchvision.datasets.CIFAR100(root='~/data/cifar-100/',
train=False,
transform=transforms.Compose([transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))]),
download=True)
else :
train_dataset = torchvision.datasets.CIFAR10(root='~/data/cifar-10/',
train=True,
transform=transform,
download=True)
test_dataset = torchvision.datasets.CIFAR10(root='~/data/cifar-10/',
train=False,
transform=transforms.Compose([transforms.ToTensor(),
transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))]),
download=True)
## Noise injection :
for i in range(int(len(train_dataset)*args.amount_noise)):
label = train_dataset.targets[i]
a=random.randint(0,args.num_classes-1)
while label == a:
a=random.randint(0,args.num_classes-1)
train_dataset.targets[i] = a
## Dataloaders
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
batch_size=args.batch_size,
shuffle=True,
num_workers=32)
test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
batch_size=args.batch_size,
shuffle=False,
num_workers=32)
## Load model
if args.model == "vit":
from models.vit import ViT
model = ViT(
image_size = 32,
patch_size = 4,
num_classes = args.num_classes,
dim = 512,
depth = 6,
heads = 8,
mlp_dim = 512,
dropout = 0.1,
emb_dropout = 0.1
).to(device)
optimizer = optim.AdamW(model.parameters(), lr=args.learning_rate, weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, args.epochs)
elif args.model == "swin":
from models.swin import swin_t
model = swin_t(window_size= 4,
num_classes=args.num_classes,
downscaling_factors=(2,2,2,1)).to(device)
optimizer = optim.AdamW(model.parameters(), lr=args.learning_rate, weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, args.epochs)
elif args.model == "resnet-18":
from models.resnet import ResNet
model = ResNet(torchvision.models.resnet.BasicBlock, [2, 2, 2, 2], num_classes=args.num_classes).to(device)
optimizer = optim.SGD(model.parameters(), lr=args.learning_rate, momentum=0.9, weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[80, 120], gamma=0.1)
## Loss, optimizer and scheduler:
loss_fn=nn.CrossEntropyLoss().to(device)
## Training and Evalutations :
train_losses=[]
train_accu=[]
eval_losses=[]
eval_accu=[]
iteration = 0
for epoch in range(0,args.epochs):
train_acc, train_loss = train(model, epoch, train_loader, device, loss_fn, optimizer, train_accu, train_losses)
test_acc, test_loss = test(model, test_loader, device, loss_fn, eval_accu, eval_losses)
scheduler.step()
torch.save(model,"CIFAR-"+str(args.num_classes)+"_"+args.model)
## Pruning loop :
for i in range(1, 35):
sparsity = 1-(1-0.2)**i
if args.model == "vit" or args.model == "swin":
layers_to_prune = [(module, "weight") for module in filter(lambda m: type(m) == torch.nn.Linear, model.modules())]
optimizer = optim.AdamW(model.parameters(), lr=args.learning_rate, weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.CosineAnnealingLR(optimizer, args.epochs)
else:
layers_to_prune = [(module, "weight") for module in filter(lambda m: type(m) == torch.nn.Conv2d or type(m) == torch.nn.Linear, model.modules())]
optimizer = optim.SGD(model.parameters(), lr=args.learning_rate, momentum=0.9, weight_decay=args.weight_decay)
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones=[80, 120], gamma=0.1)
torch.nn.utils.prune.global_unstructured(layers_to_prune, pruning_method=torch.nn.utils.prune.L1Unstructured, amount=0.2)
loss_fn=nn.CrossEntropyLoss().to(device)
train_losses=[]
train_accu=[]
eval_losses=[]
eval_accu=[]
for epoch in range(0,args.epochs):
train_acc, train_loss = train(model, epoch, train_loader, device, loss_fn, optimizer, train_accu, train_losses)
test_acc, test_loss = test(model, test_loader, device, loss_fn, eval_accu, eval_losses)
scheduler.step()
torch.save(model, "CIFAR-"+str(args.num_classes)+"_"+args.model+"_pruned_"+str(sparsity))