-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathmodel.py
169 lines (133 loc) · 7.07 KB
/
model.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
import torchmetrics
import numpy as np
import random
import torch
from torch.utils.data import DataLoader
from utils.load_data import HDF5Dataset, HDF5Dataset_Labels, MergedDataset
from utils.utils import cosine_scheduler, weight_decay_filter, load_model, LARS, load_path
from utils.loss import loss_function
from utils.augmentations import load_augmentations
import pytorch_lightning as pl
import torchvision
import psutil
import heapq
import os
import pickle
class LitModel(pl.LightningModule):
def __init__(self, hyperparameters):
super().__init__()
self.hyperparameters = hyperparameters
self.transform = self.hyperparameters['augmentation']
self.learning_rate = self.hyperparameters['learning_rate']
self.accuracy = torchmetrics.Accuracy(task='multiclass',num_classes= self.hyperparameters['num_classes'])
self.val_accuracy = torchmetrics.Accuracy(task='multiclass',num_classes= self.hyperparameters['num_classes'])
self.test_accuracy = torchmetrics.Accuracy(task='multiclass',num_classes= self.hyperparameters['num_classes'])
self.model = load_model(self.hyperparameters['model'], self.hyperparameters['num_classes'], self.hyperparameters["pretrained"], self.hyperparameters["kind"])
self.LARS = self.hyperparameters['LARS']
self.cosine_decay_lr_momentum = self.hyperparameters['cosine_decay_lr_momentum']
self.momentum_start = self.hyperparameters['momentum_start']
self.weight_decay = self.hyperparameters['weight_decay']
self.loss = loss_function(self.hyperparameters['loss'])
self.min_lr = self.hyperparameters['min_lr']
self.epochs = self.hyperparameters['num_epochs']
self.warmup_epochs = self.hyperparameters['warmup_epochs']
self.data_loader_len= self.hyperparameters['data_loader_len']
self.prediction_differences = [] ## used for LRP to store references during test #currently deactivated due to clean-up
def forward(self, x):
return self.model(x)
def training_step(self, batch, batch_idx):
x, y = batch
y_hat = self.forward(x)
loss = self.loss(y_hat.float(), y) # CrossEntropy
preds = torch.argmax(y_hat, dim=1)
self.log('accuracy', self.accuracy(preds, y), on_epoch=True)
self.log('train_loss', loss)
return loss
def validation_step(self, batch, batch_idx):
x, y = batch
y_hat = self.forward(x)
loss = self.loss(y_hat.float(), y)
preds = torch.argmax(y_hat, dim=1)
self.log('val_accuracy', self.val_accuracy(preds, y), on_epoch=True)
self.log('val_loss', loss, on_epoch=True)
def test_step(self, batch, batch_idx):
x, y = batch
y_hat = self.forward(x)
loss = self.loss(y_hat.float(), y)
preds = torch.argmax(y_hat, dim=1)
self.log('test_acc', self.test_accuracy(preds, y), on_epoch=True)
self.log('test_loss', loss, on_epoch=True)
def configure_optimizers(self):
if self.LARS:
optimizer = LARS(self.model.parameters(), lr=self.learning_rate, weight_decay=self.weight_decay, momentum=0.9, eta=0.001, weight_decay_filter=weight_decay_filter)
else:
optimizer = torch.optim.AdamW(self.model.parameters(), lr=self.learning_rate)
if self.cosine_decay_lr_momentum:
lr_scheduler = cosine_scheduler(
self.learning_rate,
self.min_lr,
self.epochs, self.data_loader_len,
warmup_epochs=self.warmup_epochs
)
momentum_scheduler = cosine_scheduler(
self.momentum_start,
1,
self.epochs, self.data_loader_len
)
lr_scheduler_config = {
'scheduler': lr_scheduler,
'interval': 'step',
'name': 'lr_scheduler',
}
momentum_scheduler_config = {
'scheduler': momentum_scheduler,
'interval': 'step',
'name': 'momentum_scheduler',
}
return {"optimizer": optimizer, "lr_schedulers": [lr_scheduler_config, momentum_scheduler_config]}
return optimizer
class LitDataModule(pl.LightningDataModule):
def __init__(self, hyperparameters):
super().__init__()
self.dataset = hyperparameters['dataset']
if self.dataset == "crc":
# handling for h5 file configuration
self.transform = load_augmentations(hyperparameters["augmentation"], True)
self.val_transform = load_augmentations(hyperparameters["augmentation_val"], True)
self.test_transform = load_augmentations(hyperparameters["augmentation_test"], True)
paths = load_path(hyperparameters["dataset"])
self.train_x_path = paths["train_data_file"]
self.train_y_path = paths["train_label_file"]
self.val_x_path = paths["valid_data_file"]
self.val_y_path = paths["valid_label_file"]
self.test_x_path = paths["test_data_file"]
self.test_y_path = paths["test_label_file"]
else:
self.transform = load_augmentations(hyperparameters["augmentation"])
self.val_transform = load_augmentations(hyperparameters["augmentation_val"])
self.test_transform = load_augmentations(hyperparameters["augmentation_test"])
self.dataset = hyperparameters["dataset"]
self.batch_size = hyperparameters['batch_size']
def setup(self, stage=None):
if self.dataset == "CRC-NoNorm":
self.C16_train = torchvision.datasets.ImageFolder('./data/CRC/NCT-CRC-HE-100K-NONORM', self.transform)
self.C16_val = torchvision.datasets.ImageFolder('./data/CRC/CRC-VAL-HE-7K', self.test_transform)
elif self.dataset == "CRC":
self.C16_train = torchvision.datasets.ImageFolder('./data/CRC/NCT-CRC-HE-100K', self.transform)
self.C16_val = torchvision.datasets.ImageFolder('./data/CRC/CRC-VAL-HE-7K', self.test_transform)
else:
x = HDF5Dataset(self.train_x_path, self.transform)
y = HDF5Dataset_Labels(self.train_y_path)
self.C16_train = MergedDataset(x,y)
x = HDF5Dataset(self.val_x_path, self.val_transform, train=False)
y = HDF5Dataset_Labels(self.val_y_path)
self.C16_val = MergedDataset(x,y)
x = HDF5Dataset(self.test_x_path, self.test_transform, train=False)
y = HDF5Dataset_Labels(self.test_y_path)
self.C16_test = MergedDataset(x,y)
def test_dataloader(self):
return DataLoader(self.C16_test, batch_size=self.batch_size, shuffle=False, num_workers=4)
def train_dataloader(self):
return DataLoader(self.C16_train, batch_size=self.batch_size, shuffle=True, num_workers=4, persistent_workers=True)
def val_dataloader(self):
return DataLoader(self.C16_val, batch_size=self.batch_size, shuffle=False, num_workers=4,persistent_workers=True)