pyrox.dev
1import torch
2from torch import nn
3from torch.utils.data import DataLoader
4from torchvision import datasets
5from torchvision.transforms import ToTensor
6
7# Download training data from open datasets.
8training_data = datasets.FashionMNIST(
9 root="data",
10 train=True,
11 download=False,
12 transform=ToTensor(),
13)
14
15# Download test data from open datasets.
16test_data = datasets.FashionMNIST(
17 root="data",
18 train=False,
19 download=False,
20 transform=ToTensor(),
21)
22
23batch_size = 64
24
25# Create data loaders.
26train_dataloader = DataLoader(training_data, batch_size=batch_size)
27test_dataloader = DataLoader(test_data, batch_size=batch_size)
28
29for X, y in test_dataloader:
30 print(f"Shape of X [N, C, H, W]: {X.shape}")
31 print(f"Shape of y: {y.shape} {y.dtype}")
32 break
33
34device = (
35 torch.accelerator.current_accelerator().type
36 if torch.accelerator.is_available()
37 else "cpu"
38)
39print(f"Using {device} device")
40
41
42# Define model
43class NeuralNetwork(nn.Module):
44 def __init__(self):
45 super().__init__()
46 self.flatten = nn.Flatten()
47 self.linear_relu_stack = nn.Sequential(
48 nn.Linear(28 * 28, 512),
49 nn.ReLU(),
50 nn.Linear(512, 512),
51 nn.ReLU(),
52 nn.Linear(512, 10),
53 )
54
55 def forward(self, x):
56 x = self.flatten(x)
57 logits = self.linear_relu_stack(x)
58 return logits
59
60
61model = NeuralNetwork().to(device)
62print(model)
63
64loss_fn = nn.CrossEntropyLoss()
65optimizer = torch.optim.SGD(model.parameters(), lr=1e-3)
66
67
68def train(
69 dataloader,
70 model,
71 loss_fn,
72 optimizer,
73):
74 size = len(dataloader.dataset)
75 model.train()
76 for batch, (X, y) in enumerate(dataloader):
77 X, y = (
78 X.to(device),
79 y.to(device),
80 )
81
82 # Compute prediction error
83 pred = model(X)
84 loss = loss_fn(pred, y)
85
86 # Backpropagation
87 loss.backward()
88 optimizer.step()
89 optimizer.zero_grad()
90
91 if batch % 100 == 0:
92 loss, current = (
93 loss.item(),
94 (batch + 1) * len(X),
95 )
96 print(f"loss: {loss:>7f} [{current:>5d}/{size:>5d}]")
97
98
99def test(dataloader, model, loss_fn):
100 size = len(dataloader.dataset)
101 num_batches = len(dataloader)
102 model.eval()
103 test_loss, correct = 0, 0
104 with torch.no_grad():
105 for X, y in dataloader:
106 X, y = (
107 X.to(device),
108 y.to(device),
109 )
110 pred = model(X)
111 test_loss += loss_fn(pred, y).item()
112 correct += (pred.argmax(1) == y).type(torch.float).sum().item()
113 test_loss /= num_batches
114 correct /= size
115 print(
116 f"Test Error: \n Accuracy: {(100 * correct):>0.1f}%, Avg loss: {test_loss:>8f} \n"
117 )
118
119
120epochs = 5
121for t in range(epochs):
122 print(f"Epoch {t + 1}\n-------------------------------")
123 train(
124 train_dataloader,
125 model,
126 loss_fn,
127 optimizer,
128 )
129 test(test_dataloader, model, loss_fn)
130print("Done!")
131
132torch.save(model.state_dict(), "model.pth")
133print("Saved PyTorch Model State to model.pth")
134
135model = NeuralNetwork().to(device)
136model.load_state_dict(torch.load("model.pth", weights_only=True))
137
138classes = [
139 "T-shirt/top",
140 "Trouser",
141 "Pullover",
142 "Dress",
143 "Coat",
144 "Sandal",
145 "Shirt",
146 "Sneaker",
147 "Bag",
148 "Ankle boot",
149]
150
151model.eval()
152x, y = test_data[0][0], test_data[0][1]
153with torch.no_grad():
154 x = x.to(device)
155 pred = model(x)
156 predicted, actual = (
157 classes[pred[0].argmax(0)],
158 classes[y],
159 )
160 print(f'Predicted: "{predicted}", Actual: "{actual}"')