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
| import torch
import torch.nn as nn
import torch.optim as optim
from torch.utils.data import DataLoader
from torchvision import datasets, transforms
import wandb
wandb.init(
project="mnist-demo",
name="my-first-run",
config={
"architecture": "CNN",
"epochs": 5,
"batch_size": 64,
"learning_rate": 0.001,
"optimizer": "Adam",
}
)
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])
train_dataset = datasets.MNIST('./data', train=True, download=True, transform=transform)
test_dataset = datasets.MNIST('./data', train=False, transform=transform)
train_loader = DataLoader(train_dataset, batch_size=wandb.config.batch_size, shuffle=True)
test_loader = DataLoader(test_dataset, batch_size=wandb.config.batch_size, shuffle=False)
class SimpleCNN(nn.Module):
def __init__(self):
super().__init__()
self.conv1 = nn.Conv2d(1, 32, 3, 1)
self.conv2 = nn.Conv2d(32, 64, 3, 1)
self.dropout1 = nn.Dropout2d(0.25)
self.dropout2 = nn.Dropout2d(0.5)
self.fc1 = nn.Linear(9216, 128)
self.fc2 = nn.Linear(128, 10)
def forward(self, x):
~~~~~~~~~~~~~
return output
model = SimpleCNN()
optimizer = optim.Adam(model.parameters(), lr=wandb.config.learning_rate)
criterion = nn.CrossEntropyLoss()
wandb.watch(model, log="all", log_freq=10)
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model.to(device)
for epoch in range(wandb.config.epochs):
model.train()
running_loss = 0.0
correct = 0
total = 0
for batch_idx, (data, target) in enumerate(train_loader):
data, target = data.to(device), target.to(device)
~~~~~~~~~~~~~~~~~~~~~
if batch_idx % 100 == 0:
wandb.log({
"batch_loss": loss.item(),
"batch_accuracy": 100. * correct / total,
"batch_idx": batch_idx
})
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in test_loader:
data, target = data.to(device), target.to(device)
output = model(data)
test_loss += criterion(output, target).item()
pred = output.argmax(dim=1, keepdim=True)
correct += pred.eq(target.view_as(pred)).sum().item()
test_loss /= len(test_loader)
accuracy = 100. * correct / len(test_loader.dataset)
wandb.log({
"epoch": epoch + 1,
"train_loss": running_loss / len(train_loader),
"train_accuracy": 100. * correct / total,
"test_loss": test_loss,
"test_accuracy": accuracy,
})
wandb.log({"sample_images": [wandb.Image(img) for img in images]})
table = wandb.Table(columns=["id", "prediction", "label"], data=[[1, "cat", "cat"], [2, "dog", "cat"]])
wandb.log({"predictions": table})
import matplotlib.pyplot as plt
plt.plot([1, 2, 3], [1, 4, 9])
wandb.log({"chart": wandb.Plotly(plt.gcf())})
print(f'Epoch {epoch+1}: Train Loss: {running_loss/len(train_loader):.4f}, '
f'Test Acc: {accuracy:.2f}%')
torch.save(model.state_dict(), "mnist_model.pth")
wandb.save("mnist_model.pth")
wandb.finish()
|
微信
