-
Notifications
You must be signed in to change notification settings - Fork 2
/
train.py
142 lines (116 loc) · 4.67 KB
/
train.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
import torchvision
import cv2
import numpy as np
import matplotlib.pyplot as plt
import os
import neptune.new as neptune
from tqdm import tqdm
from transformer import VisionTransformer
def neptune_monitoring(config):
PARAMS = {}
for key, val in config.__dict__.items():
if key not in ["__module__", "__dict__", "__weakref__", "__doc__"]:
PARAMS[key] = val
return PARAMS
def train_Engine(n_epochs,
train_data,
val_data,
model,
optimizer,
loss_fn,
device,
monitoring=True):
train_accuracy = 0
val_accuracy = 0
best_accuracy = 0
for epoch in range(1, n_epochs + 1):
total = 0
with tqdm(train_data, unit="iteration") as train_epoch:
train_epoch.set_description(f"Epoch {epoch}")
for i, (data, target) in enumerate(train_epoch):
total_samples = len(train_data.dataset)
#device
model = model.to(device)
x = data.to(device)
y = target.to(device)
optimizer.zero_grad()
logits, attn_weights = model(x)
proba = F.log_softmax(logits, dim=1)
loss = F.nll_loss(proba, y, reduction='sum')
loss.backward()
optimizer.step()
_, pred = torch.max(logits, dim=1) #
train_accuracy += torch.sum(pred==y).item()
total += target.size(0)
accuracy_=(100 * train_accuracy/ total)
train_epoch.set_postfix(loss=loss.item(), accuracy=accuracy_)
if monitoring:
run['Training_loss'].log(loss.item())
run['Training_acc'].log(accuracy_)
if accuracy_ > best_accuracy:
best_accuracy = accuracy_
best_model = model
torch.save(best_model, f'/metadata/model.pth')
total_samples = len(val_data.dataset)
correct_samples = 0
total_ = 0
model.eval()
with torch.no_grad():
with tqdm(val_data, unit="iteration") as val_epoch:
val_epoch.set_description(f"Epoch {epoch}")
for i, (data, target) in enumerate(val_epoch):
model = model.to(device)
x = data.to(device)
y = target.to(device)
logits,attn_weights = model(x)
proba = F.log_softmax(logits, dim=1)
val_loss = F.nll_loss(proba, y, reduction='sum')
_, pred = torch.max(logits, dim=1)#
val_accuracy += torch.sum(pred==y).item()
total_ += target.size(0)
val_accuracy_ = (100 * val_accuracy/ total_)
val_epoch.set_postfix(loss=val_loss.item(), accuracy=val_accuracy_)
if monitoring:
run['Val_accuracy '].log(val_accuracy_)
run['Val_loss'].log(loss.item())
if __name__ == "__main__":
from preprocessing import Dataset
from config import Config
config = Config()
params = neptune_monitoring(Config)
run = neptune.init(
project="nielspace/ViT-bird-classification",
api_token="eyJhcGlfYWRkcmVzcyI6Imh0dHBzOi8vYXBwLm5lcHR1bmUuYWkiLCJhcGlfdXJsIjoiaHR0cHM6Ly9hcHAubmVwdHVuZS5haSIsImFwaV9rZXkiOiJkYjRhYzI0Ny0zZjBmLTQ3YjYtOTY0Yi05ZTQ4ODM3YzE0YWEifQ==",
)
run["parameters"] = params
model = VisionTransformer(
img_size=config.IMG_SIZE,
num_classes=config.NUM_CLASSES,
hidden_size=config.HIDDEN_SIZE,
in_channels=config.IN_CHANNELS,
num_layers=config.NUM_LAYERS,
num_attention_heads=config.NUM_ATTENTION_HEADS,
linear_dim=config.LINEAR_DIM,
dropout_rate=config.DROPOUT_RATE,
attention_dropout_rate=config.ATTENTION_DROPOUT_RATE,
eps=config.EPS,
std_norm=config.STD_NORM,
)
train_data, val_data, test_data = Dataset(
config.BATCH_SIZE, config.IMG_SIZE, config.DATASET_SAMPLE
) # neptune.save_checkpoint(
optimizer = optim.Adam(model.parameters(), lr=0.003)
train_Engine(
n_epochs=config.N_EPOCHS,
train_data=train_data,
val_data=val_data,
model=model,
optimizer=optimizer,
loss_fn="nll_loss",
device=config.DEVICE[1],
monitoring=True,
)