-
Notifications
You must be signed in to change notification settings - Fork 3
/
Copy pathocr_mnist_torch.py
229 lines (205 loc) · 11 KB
/
ocr_mnist_torch.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
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
# use pytorch Implement MNIST
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
from torchvision import datasets, transforms
from torch.utils.data import DataLoader
### 3. Define Hyperparameters
batch_size = 64
learning_rate = 0.01
epochs = 100
### 4. Prepare the MNIST Dataset
transform = transforms.Compose([
transforms.ToTensor(),
transforms.Normalize((0.1307,), (0.3081,))
])
train_dataset = datasets.MNIST(root='./data', train=True, download=True, transform=transform)
test_dataset = datasets.MNIST(root='./data', train=False, download=True, transform=transform)
train_loader = DataLoader(dataset=train_dataset, batch_size=batch_size, shuffle=True)
test_loader = DataLoader(dataset=test_dataset, batch_size=batch_size, shuffle=False)
### 5. Build the Neural Network Model
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.fc1 = nn.Linear(28 * 28, 128)
self.fc2 = nn.Linear(128, 64)
self.fc3 = nn.Linear(64, 10)
def forward(self, x):
x = x.view(-1, 28 * 28)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
### 6. Initialize the Model, Loss Function, and Optimizer
model = Net()
criterion = nn.CrossEntropyLoss()
optimizer = optim.SGD(model.parameters(), lr=learning_rate)
### 7. Train the Model
for epoch in range(epochs):
model.train()
for batch_idx, (data, target) in enumerate(train_loader):
optimizer.zero_grad()
output = model(data)
loss = criterion(output, target)
loss.backward()
optimizer.step()
if batch_idx % 100 == 0:
print(f'Epoch: {epoch+1}/{epochs} [Batch: {batch_idx*len(data)}/{len(train_loader.dataset)}] Loss: {loss.item():.6f}')
### 8. Test the Model
model.eval()
test_loss = 0
correct = 0
with torch.no_grad():
for data, target in test_loader:
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.dataset)
accuracy = 100. * correct / len(test_loader.dataset)
print(f'Test set: Average loss: {test_loss:.4f}, Accuracy: {correct}/{len(test_loader.dataset)} ({accuracy:.2f}%)')
### 9. Save the Model (Optional)
torch.save(model.state_dict(), "mnist_model.pth")
### Summary
# This script loads the MNIST dataset, defines a simple neural network model, trains it on the dataset, and evaluates its performance on the test set. The network architecture used here is quite basic, and you could improve it with more layers, dropout, or other advanced techniques for better performance.
## ----- very fast ------ in m3 ------
# @ python MNIST_torch.py
# Downloading http://yann.lecun.com/exdb/mnist/train-images-idx3-ubyte.gz
# Failed to download (trying next):
# HTTP Error 403: Forbidden
#
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-images-idx3-ubyte.gz to ./data/MNIST/raw/train-images-idx3-ubyte.gz
# 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 9912422/9912422 [00:01<00:00, 6148112.22it/s]
# Extracting ./data/MNIST/raw/train-images-idx3-ubyte.gz to ./data/MNIST/raw
#
# Downloading http://yann.lecun.com/exdb/mnist/train-labels-idx1-ubyte.gz
# Failed to download (trying next):
# HTTP Error 403: Forbidden
#
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/train-labels-idx1-ubyte.gz to ./data/MNIST/raw/train-labels-idx1-ubyte.gz
# 100%|██████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 28881/28881 [00:00<00:00, 43471.05it/s]
# Extracting ./data/MNIST/raw/train-labels-idx1-ubyte.gz to ./data/MNIST/raw
#
# Downloading http://yann.lecun.com/exdb/mnist/t10k-images-idx3-ubyte.gz
# Failed to download (trying next):
# HTTP Error 403: Forbidden
#
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw/t10k-images-idx3-ubyte.gz
# 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1648877/1648877 [00:01<00:00, 1620685.15it/s]
# Extracting ./data/MNIST/raw/t10k-images-idx3-ubyte.gz to ./data/MNIST/raw
#
# Downloading http://yann.lecun.com/exdb/mnist/t10k-labels-idx1-ubyte.gz
# Failed to download (trying next):
# HTTP Error 403: Forbidden
#
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz
# Downloading https://ossci-datasets.s3.amazonaws.com/mnist/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz
# 100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 4542/4542 [00:00<00:00, 38500.31it/s]
# Extracting ./data/MNIST/raw/t10k-labels-idx1-ubyte.gz to ./data/MNIST/raw
#
# Epoch: 1/10 [Batch: 0/60000] Loss: 2.285485
# Epoch: 1/10 [Batch: 6400/60000] Loss: 1.787282
# Epoch: 1/10 [Batch: 12800/60000] Loss: 1.077599
# Epoch: 1/10 [Batch: 19200/60000] Loss: 0.655586
# Epoch: 1/10 [Batch: 25600/60000] Loss: 0.473320
# Epoch: 1/10 [Batch: 32000/60000] Loss: 0.526482
# Epoch: 1/10 [Batch: 38400/60000] Loss: 0.373468
# Epoch: 1/10 [Batch: 44800/60000] Loss: 0.517464
# Epoch: 1/10 [Batch: 51200/60000] Loss: 0.446572
# Epoch: 1/10 [Batch: 57600/60000] Loss: 0.406734
# Epoch: 2/10 [Batch: 0/60000] Loss: 0.674305
# Epoch: 2/10 [Batch: 6400/60000] Loss: 0.377594
# Epoch: 2/10 [Batch: 12800/60000] Loss: 0.215526
# Epoch: 2/10 [Batch: 19200/60000] Loss: 0.293928
# Epoch: 2/10 [Batch: 25600/60000] Loss: 0.245266
# Epoch: 2/10 [Batch: 32000/60000] Loss: 0.297042
# Epoch: 2/10 [Batch: 38400/60000] Loss: 0.283664
# Epoch: 2/10 [Batch: 44800/60000] Loss: 0.340037
# Epoch: 2/10 [Batch: 51200/60000] Loss: 0.383973
# Epoch: 2/10 [Batch: 57600/60000] Loss: 0.453223
# Epoch: 3/10 [Batch: 0/60000] Loss: 0.213752
# Epoch: 3/10 [Batch: 6400/60000] Loss: 0.314306
# Epoch: 3/10 [Batch: 12800/60000] Loss: 0.150672
# Epoch: 3/10 [Batch: 19200/60000] Loss: 0.273192
# Epoch: 3/10 [Batch: 25600/60000] Loss: 0.248815
# Epoch: 3/10 [Batch: 32000/60000] Loss: 0.481632
# Epoch: 3/10 [Batch: 38400/60000] Loss: 0.284605
# Epoch: 3/10 [Batch: 44800/60000] Loss: 0.215667
# Epoch: 3/10 [Batch: 51200/60000] Loss: 0.496624
# Epoch: 3/10 [Batch: 57600/60000] Loss: 0.227222
# Epoch: 4/10 [Batch: 0/60000] Loss: 0.260029
# Epoch: 4/10 [Batch: 6400/60000] Loss: 0.215243
# Epoch: 4/10 [Batch: 12800/60000] Loss: 0.171734
# Epoch: 4/10 [Batch: 19200/60000] Loss: 0.211567
# Epoch: 4/10 [Batch: 25600/60000] Loss: 0.272625
# Epoch: 4/10 [Batch: 32000/60000] Loss: 0.345563
# Epoch: 4/10 [Batch: 38400/60000] Loss: 0.251216
# Epoch: 4/10 [Batch: 44800/60000] Loss: 0.210512
# Epoch: 4/10 [Batch: 51200/60000] Loss: 0.218367
# Epoch: 4/10 [Batch: 57600/60000] Loss: 0.270740
# Epoch: 5/10 [Batch: 0/60000] Loss: 0.171540
# Epoch: 5/10 [Batch: 6400/60000] Loss: 0.182449
# Epoch: 5/10 [Batch: 12800/60000] Loss: 0.193923
# Epoch: 5/10 [Batch: 19200/60000] Loss: 0.399605
# Epoch: 5/10 [Batch: 25600/60000] Loss: 0.176479
# Epoch: 5/10 [Batch: 32000/60000] Loss: 0.263907
# Epoch: 5/10 [Batch: 38400/60000] Loss: 0.143796
# Epoch: 5/10 [Batch: 44800/60000] Loss: 0.117058
# Epoch: 5/10 [Batch: 51200/60000] Loss: 0.102552
# Epoch: 5/10 [Batch: 57600/60000] Loss: 0.234998
# Epoch: 6/10 [Batch: 0/60000] Loss: 0.190606
# Epoch: 6/10 [Batch: 6400/60000] Loss: 0.161692
# Epoch: 6/10 [Batch: 12800/60000] Loss: 0.273620
# Epoch: 6/10 [Batch: 19200/60000] Loss: 0.274887
# Epoch: 6/10 [Batch: 25600/60000] Loss: 0.119512
# Epoch: 6/10 [Batch: 32000/60000] Loss: 0.187856
# Epoch: 6/10 [Batch: 38400/60000] Loss: 0.142739
# Epoch: 6/10 [Batch: 44800/60000] Loss: 0.127890
# Epoch: 6/10 [Batch: 51200/60000] Loss: 0.253540
# Epoch: 6/10 [Batch: 57600/60000] Loss: 0.135013
# Epoch: 7/10 [Batch: 0/60000] Loss: 0.139322
# Epoch: 7/10 [Batch: 6400/60000] Loss: 0.217386
# Epoch: 7/10 [Batch: 12800/60000] Loss: 0.127039
# Epoch: 7/10 [Batch: 19200/60000] Loss: 0.138367
# Epoch: 7/10 [Batch: 25600/60000] Loss: 0.167661
# Epoch: 7/10 [Batch: 32000/60000] Loss: 0.103273
# Epoch: 7/10 [Batch: 38400/60000] Loss: 0.158958
# Epoch: 7/10 [Batch: 44800/60000] Loss: 0.232634
# Epoch: 7/10 [Batch: 51200/60000] Loss: 0.269437
# Epoch: 7/10 [Batch: 57600/60000] Loss: 0.234816
# Epoch: 8/10 [Batch: 0/60000] Loss: 0.051134
# Epoch: 8/10 [Batch: 6400/60000] Loss: 0.094156
# Epoch: 8/10 [Batch: 12800/60000] Loss: 0.271425
# Epoch: 8/10 [Batch: 19200/60000] Loss: 0.155359
# Epoch: 8/10 [Batch: 25600/60000] Loss: 0.183958
# Epoch: 8/10 [Batch: 32000/60000] Loss: 0.182390
# Epoch: 8/10 [Batch: 38400/60000] Loss: 0.154530
# Epoch: 8/10 [Batch: 44800/60000] Loss: 0.133099
# Epoch: 8/10 [Batch: 51200/60000] Loss: 0.107830
# Epoch: 8/10 [Batch: 57600/60000] Loss: 0.141398
# Epoch: 9/10 [Batch: 0/60000] Loss: 0.130976
# Epoch: 9/10 [Batch: 6400/60000] Loss: 0.278118
# Epoch: 9/10 [Batch: 12800/60000] Loss: 0.233672
# Epoch: 9/10 [Batch: 19200/60000] Loss: 0.132633
# Epoch: 9/10 [Batch: 25600/60000] Loss: 0.118954
# Epoch: 9/10 [Batch: 32000/60000] Loss: 0.127910
# Epoch: 9/10 [Batch: 38400/60000] Loss: 0.176615
# Epoch: 9/10 [Batch: 44800/60000] Loss: 0.099357
# Epoch: 9/10 [Batch: 51200/60000] Loss: 0.089455
# Epoch: 9/10 [Batch: 57600/60000] Loss: 0.162767
# Epoch: 10/10 [Batch: 0/60000] Loss: 0.028447
# Epoch: 10/10 [Batch: 6400/60000] Loss: 0.058944
# Epoch: 10/10 [Batch: 12800/60000] Loss: 0.173397
# Epoch: 10/10 [Batch: 19200/60000] Loss: 0.111585
# Epoch: 10/10 [Batch: 25600/60000] Loss: 0.079948
# Epoch: 10/10 [Batch: 32000/60000] Loss: 0.082440
# Epoch: 10/10 [Batch: 38400/60000] Loss: 0.110963
# Epoch: 10/10 [Batch: 44800/60000] Loss: 0.058552
# Epoch: 10/10 [Batch: 51200/60000] Loss: 0.152358
# Epoch: 10/10 [Batch: 57600/60000] Loss: 0.090184
# Test set: Average loss: 0.0019, Accuracy: 9640/10000 (96.40%)
#