File size: 7,331 Bytes
88ebb5a
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
"""

https://github.com/marrrcin/pytorch-resnet-mnist/blob/master/pytorch-resnet-mnist.ipynb

https://github.com/huyvnphan/PyTorch_CIFAR10/tree/master/cifar10_models

"""
import os, sys
from argparse import ArgumentParser

from pytorch_lightning import Trainer, seed_everything
from pytorch_lightning.callbacks import ModelCheckpoint
# from pytorch_lightning.accelerators import GPUAccelerator

sys.path.append('/home/yiming/ContrastDebugger/Model-mnist')
from data import MNISTData
# from module import MNISTModule
import pytorch_lightning as pl
import torch
# from torchmetrics import Accuracy
from pytorch_lightning.metrics import Accuracy

from cifar10_models.densenet import densenet121, densenet161, densenet169
from cifar10_models.googlenet import googlenet
from cifar10_models.inception import inception_v3
from cifar10_models.mobilenetv2 import mobilenet_v2
from cifar10_models.resnet import resnet18, resnet34, resnet50
from cifar10_models.vgg import vgg11_bn, vgg13_bn, vgg16_bn, vgg19_bn
from cifar10_models.mlp import mlp3
from cifar10_models.convnet import convnet
from schduler import WarmupCosineLR
from torch import nn
# from pytorch_lightning.core.decorators import auto_move_data

from torchvision.transforms import ToTensor
from torchvision.datasets import MNIST
from torch.utils.data import DataLoader
import json
import os

parser = ArgumentParser()

# PROGRAM level args
parser.add_argument("--data_dir", type=str, default="data")
parser.add_argument("--test_phase", type=int, default=0, choices=[0, 1])
parser.add_argument("--dev", type=int, default=0, choices=[0, 1])

# TRAINER args
parser.add_argument("--classifier", type=str, default="resnet18")
parser.add_argument("--precision", type=int, default=32, choices=[16, 32])
parser.add_argument("--batch_size", type=int, default=128)
parser.add_argument("--max_epochs", type=int, default=20)
parser.add_argument("--num_workers", type=int, default=2)
parser.add_argument("--gpu_id", type=str, default="0")

parser.add_argument("--learning_rate", type=float, default=5e-3)
parser.add_argument("--weight_decay", type=float, default=1e-2)
parser.add_argument("--filepath", type=str, default="Model")
parser.add_argument("--period", type=int, default=1)
parser.add_argument("--save_top_k", type=int, default=-1)

args = parser.parse_args(args=[])

def main(args):

    seed_everything(0)
    os.environ["CUDA_VISIBLE_DEVICES"] = args.gpu_id

    checkpoint = ModelCheckpoint(
        # dirpath=os.path.join(args.filepath, args.classifier),
        # filename="{epoch:03d}",
        filepath=os.path.join(args.filepath, args.classifier, "{epoch:03d}"),
        monitor="acc/val",
        mode="max",
        # save_last=False,
        period=args.period,
        save_top_k=args.save_top_k,
        save_weights_only=True,
    )

    trainer = Trainer(
        fast_dev_run=bool(args.dev),
        gpus=args.gpu_id,
        deterministic=True,
        weights_summary=None,
        log_every_n_steps=1,
        max_epochs=args.max_epochs,
        checkpoint_callback=checkpoint,
        precision=args.precision,

    )

    model = MNISTModule(args)
    # data = MNISTData(args)
    # trainloader = data.train_dataloader()
    # data.save_train_data(trainloader, args.filepath)
    # testloader = data.test_dataloader()
    # data.save_test_data(testloader, args.filepath)

    # if bool(args.test_phase):
    #     trainer.test(model, data.test_dataloader())
    # else:
    #     trainer.fit(model, data)
    #     trainer.test()

all_classifiers = {
    "vgg11_bn": vgg11_bn(),
    "vgg13_bn": vgg13_bn(),
    "vgg16_bn": vgg16_bn(),
    "vgg19_bn": vgg19_bn(),
    "resnet18": resnet18(),
    "resnet34": resnet34(),
    "resnet50": resnet50(),
    "densenet121": densenet121(),
    "densenet161": densenet161(),
    "densenet169": densenet169(),
    "mobilenet_v2": mobilenet_v2(),
    "googlenet": googlenet(),
    "inception_v3": inception_v3(),
    "mlp":mlp3(),
    "convnet":convnet()
}


class MNISTModule(pl.LightningModule):
    def __init__(self, my_hparams):
        super().__init__()
        self.my_hparams = my_hparams

        self.criterion = torch.nn.CrossEntropyLoss()
        self.accuracy = Accuracy()

        self.model = all_classifiers[self.my_hparams.classifier]
        if self.my_hparams.classifier not in ["mlp", "convnet"]:
            self.model.conv1 = nn.Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)

    def forward(self, batch):
        images, labels = batch
        predictions = self.model(images)
        loss = self.criterion(predictions, labels)
        accuracy = self.accuracy(predictions, labels)
        return loss, accuracy * 100

    # @auto_move_data
    # def forward(self, x):
    #     return self.model(x)

    def training_step(self, batch, batch_nb):
        loss, accuracy = self.forward(batch)
        self.log("loss/train", loss)
        self.log("acc/train", accuracy)
        return loss

    def validation_step(self, batch, batch_nb):
        loss, accuracy = self.forward(batch)
        self.log("loss/val", loss)
        self.log("acc/val", accuracy)

    def test_step(self, batch, batch_nb):
        loss, accuracy = self.forward(batch)
        self.log("acc/test", accuracy)

    def train_dataloader(self):
        transform = ToTensor()
        dataset = MNIST("mnist", train=True, download=True, transform=transform)
        dataloader = DataLoader(
            dataset,
            batch_size=self.my_hparams.batch_size,
            num_workers=self.my_hparams.num_workers,
            shuffle=True,
        )
        return dataloader
    
    def val_dataloader(self):
        transform = ToTensor()
        dataset = MNIST("mnist", train=False, download=True, transform=transform)
        dataloader = DataLoader(
            dataset,
            batch_size=self.my_hparams.batch_size,
            num_workers=self.my_hparams.num_workers,
            drop_last=True,
            pin_memory=True,
        )
        return dataloader

    def test_dataloader(self):
        return self.val_dataloader()

    def configure_optimizers(self):
        optimizer = torch.optim.SGD(
            self.model.parameters(),
            lr=self.my_hparams.learning_rate,
            weight_decay=self.my_hparams.weight_decay,
            momentum=0.9,
            nesterov=True,
        )
        total_steps = self.my_hparams.max_epochs * len(self.train_dataloader())
        scheduler = {
            "scheduler": WarmupCosineLR(
                optimizer, warmup_epochs=total_steps * 0.3, max_epochs=total_steps
            ),
            "interval": "step",
            "name": "learning_rate",
        }
        return [optimizer], [scheduler]

    def on_train_epoch_end(self, epoch_output):
        epoch = self.trainer.current_epoch
        state_dict = self.model.state_dict()
        save_dir = "/home/yiming/EXP/mnist_resnet18/Model/Epoch_" + str(self.current_epoch + 1)
        os.makedirs(save_dir, exist_ok=True)
        save_path = os.path.join(save_dir, "subject_model.pth")
        torch.save(state_dict, save_path)        
    
main(args)