-
Notifications
You must be signed in to change notification settings - Fork 11
/
spectra_encoder_model.py
39 lines (32 loc) · 1.39 KB
/
spectra_encoder_model.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
import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.nn.init as init
from torch.autograd import Variable
def conv_out_dim(length_in, kernel, stride, padding, dilation):
length_out = (length_in + 2 * padding - dilation * (kernel - 1) - 1)// stride + 1
return length_out
class Net1D(nn.Module):
def __init__(self, length_in):
super(Net1D, self).__init__()
out_1 = conv_out_dim(length_in, 200, 1, 100, 1)
out_2 = conv_out_dim(out_1, 50, 50, 0, 1)
out_3 = conv_out_dim(out_2, 200, 1, 100, 1)
out_4 = conv_out_dim(out_3, 50, 50, 0, 1)
self.cnn_out = 8*out_4
self.conv1 = nn.Conv1d(4, 4, 200, stride=1, padding=100, dilation=1)
self.norm1 = nn.BatchNorm1d(4)
self.pool1 = nn.MaxPool1d(50, stride=50, padding=0, dilation=1)
self.pool2 = nn.MaxPool1d(50, stride=50, padding=0, dilation=1)
self.conv2 = nn.Conv1d(4, 8, 200, stride=1, padding=100, dilation=1)
self.norm2 = nn.BatchNorm1d(8)
self.fc1 = nn.Linear(8*out_4, 512)
self.fc2 = nn.Linear(512, 512)
self.norm3 = nn.BatchNorm1d(512)
def forward(self, x):
x = self.pool1(F.relu(self.norm1(self.conv1(x))))
x = self.pool2(F.relu(self.norm2(self.conv2(x))))
x = x.view(-1, self.cnn_out)
x = F.relu(self.norm3(self.fc1(x)))
x = torch.tanh(self.fc2(x))
return x