-
Notifications
You must be signed in to change notification settings - Fork 2
/
model.py
117 lines (97 loc) · 3.79 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
from collections import OrderedDict
import torch.nn.functional as F
import torch
import pytorch_lightning as pl
import torchvision
from torch.utils.data import DataLoader, random_split
from torchvision import models
import pandas as pd
from data_loader import PlantDataset
from torch import optim
from torch import nn
from torchtoolbox.nn import LabelSmoothingLoss
from torchtoolbox.tools import mixup_data, mixup_criterion
import torchtoolbox.nn.loss as tloss
from model_finetune import *
from config import DefaultConfig as dc
import utils
class LitPlants(pl.LightningModule):
def __init__(self):
super().__init__()
self.net =Efficient(dc.backbone_name, dc.num_classes)
def forward(self, x):
y = self.net(x)
return y
def configure_optimizers(self):
optimizer = optim.Adam(self.parameters(), lr=dc.lr)
scheduler1 = optim.lr_scheduler.CosineAnnealingWarmRestarts(optimizer, T_0=10, T_mult=1)
# scheduler0 ={
# 'scheduler': optim.lr_scheduler.ReduceLROnPlateau(optimizer, mode='max',factor=0.5,patience=3,verbose=True),
# 'monitor': 'val_acc', # Default: val_loss
# 'interval': 'epoch',
# 'frequency': 1}
# scheduler = utils.GradualWarmupScheduler(optimizer, 8, 10, scheduler1)
return [optimizer], [scheduler1]
# def prepare_data(self):
# data = pd.read_csv("data/data.csv",sep="\t")
# split = int(len(data) * 0.5)
# data = data.sample(frac=1, random_state=2020) # 打乱
# data[:split].to_csv("data/train.csv", index=False, sep="\t")
# data[split:].to_csv("data/val.csv", index=False, sep="\t")
###############
def train_dataloader(self):
train_dataset = PlantDataset(
"data/train.csv",
transforms=dc.transforms_train)
train_loader = DataLoader(
dataset=train_dataset,
batch_size=dc.batch_size,
shuffle=True,
num_workers=4
)
return train_loader
def training_step(self, batch, batch_idx): # mixup + labelsmooth+ circle_loss
data, labels = batch
# mixed_x, labels_a, labels_b, lam = mixup_data(data, labels, 0.2)
# output = self(mixed_x)
# loss = mixup_criterion(LabelSmoothingLoss(4, smoothing=0.1), output, labels_a, labels_b, lam)
output = self(data)
loss = LabelSmoothingLoss(dc.num_classes, smoothing=0.1)(output, labels)
return {'loss': loss}
def training_epoch_end(self, outputs):
loss = torch.stack([output["loss"] for output in outputs]).mean()
tqdm_dict = {"loss": loss} # todo lr
result = {'log': tqdm_dict, 'train_loss': loss, 'progress_bar': tqdm_dict}
return result
##########
def val_dataloader(self):
val_dataset = PlantDataset(
"data/val.csv",
transforms=dc.transforms_val)
val_loader = DataLoader(
dataset=val_dataset,
batch_size=dc.batch_size,
shuffle=False,
num_workers=4
)
return val_loader
def validation_step(self, batch, batch_idx):
images, target = batch
output = self(images)
acc = self.accuracy(output, target)
output = OrderedDict({
'val_acc': acc,
})
return output
def validation_epoch_end(self, outputs):
val_acc = torch.stack([output["val_acc"] for output in outputs]).mean()
tqdm_dict = {"val_acc": val_acc}
result = {'log': tqdm_dict, 'progress_bar': tqdm_dict}
return result
#########
@classmethod
def accuracy(cls, output, target):
batch_size = target.size(0)
pred = torch.argmax(output, dim=1)
correct = (pred == target).float().sum() / batch_size
return correct