-
Notifications
You must be signed in to change notification settings - Fork 0
/
CNN.py
26 lines (19 loc) · 1.26 KB
/
CNN.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
import torch
class SimpleCNN(torch.nn.Module):
def __init__(self, n_in_channels: int = 2, n_hidden_layers: int = 3, n_kernels: int = 32, kernel_size: int = 7):
"""Simple CNN with `n_hidden_layers`, `n_kernels`, and `kernel_size` as hyperparameters"""
super(SimpleCNN, self).__init__()
cnn = []
for i in range(n_hidden_layers):
cnn.append(torch.nn.Conv2d(in_channels=n_in_channels, out_channels=n_kernels, kernel_size=kernel_size,
bias=True, padding=int(kernel_size / 2)))
cnn.append(torch.nn.ReLU())
n_in_channels = n_kernels
self.hidden_layers = torch.nn.Sequential(*cnn)
self.output_layer = torch.nn.Conv2d(in_channels=n_in_channels, out_channels=1,
kernel_size=kernel_size, bias=True, padding=3) # padding=int(kernel_size / 2)
def forward(self, x):
"""Apply CNN to input `x` of shape (N, n_channels, X, Y), where N=n_samples and X, Y are spatial dimensions"""
cnn_out = self.hidden_layers(x) # apply hidden layers (N, n_in_channels, X, Y) -> (N, n_kernels, X, Y)
pred = self.output_layer(cnn_out) # apply output layer (N, n_kernels, X, Y) -> (N, 1, X, Y)
return pred