| 24 | # Here's a simple CNN and loss function: |
| 25 | |
| 26 | class SimpleCNN(nn.Module): |
| 27 | def __init__(self): |
| 28 | super(SimpleCNN, self).__init__() |
| 29 | self.conv1 = nn.Conv2d(1, 32, 3, 1) |
| 30 | self.conv2 = nn.Conv2d(32, 64, 3, 1) |
| 31 | self.fc1 = nn.Linear(9216, 128) |
| 32 | self.fc2 = nn.Linear(128, 10) |
| 33 | |
| 34 | def forward(self, x): |
| 35 | x = self.conv1(x) |
| 36 | x = F.relu(x) |
| 37 | x = self.conv2(x) |
| 38 | x = F.relu(x) |
| 39 | x = F.max_pool2d(x, 2) |
| 40 | x = torch.flatten(x, 1) |
| 41 | x = self.fc1(x) |
| 42 | x = F.relu(x) |
| 43 | x = self.fc2(x) |
| 44 | output = F.log_softmax(x, dim=1) |
| 45 | return output |
| 46 | |
| 47 | def loss_fn(predictions, targets): |
| 48 | return F.nll_loss(predictions, targets) |