clmr
copied
You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Readme
Files and versions
67 lines
1.8 KiB
67 lines
1.8 KiB
import torch
|
|
import torch.nn as nn
|
|
from .model import Model
|
|
|
|
|
|
class SampleCNN(Model):
|
|
def __init__(self, strides, supervised, out_dim):
|
|
super(SampleCNN, self).__init__()
|
|
|
|
self.strides = strides
|
|
self.supervised = supervised
|
|
self.sequential = [
|
|
nn.Sequential(
|
|
nn.Conv1d(1, 128, kernel_size=3, stride=3, padding=0),
|
|
nn.BatchNorm1d(128),
|
|
nn.ReLU(),
|
|
)
|
|
]
|
|
|
|
self.hidden = [
|
|
[128, 128],
|
|
[128, 128],
|
|
[128, 256],
|
|
[256, 256],
|
|
[256, 256],
|
|
[256, 256],
|
|
[256, 256],
|
|
[256, 256],
|
|
[256, 512],
|
|
]
|
|
|
|
assert len(self.hidden) == len(
|
|
self.strides
|
|
), "Number of hidden layers and strides are not equal"
|
|
for stride, (h_in, h_out) in zip(self.strides, self.hidden):
|
|
self.sequential.append(
|
|
nn.Sequential(
|
|
nn.Conv1d(h_in, h_out, kernel_size=stride, stride=1, padding=1),
|
|
nn.BatchNorm1d(h_out),
|
|
nn.ReLU(),
|
|
nn.MaxPool1d(stride, stride=stride),
|
|
)
|
|
)
|
|
|
|
# 1 x 512
|
|
self.sequential.append(
|
|
nn.Sequential(
|
|
nn.Conv1d(512, 512, kernel_size=3, stride=1, padding=1),
|
|
nn.BatchNorm1d(512),
|
|
nn.ReLU(),
|
|
)
|
|
)
|
|
|
|
self.sequential = nn.Sequential(*self.sequential)
|
|
|
|
if self.supervised:
|
|
self.dropout = nn.Dropout(0.5)
|
|
self.fc = nn.Linear(512, out_dim)
|
|
|
|
def forward(self, x):
|
|
out = self.sequential(x)
|
|
if self.supervised:
|
|
out = self.dropout(out)
|
|
|
|
out = out.reshape(x.shape[0], out.size(1) * out.size(2))
|
|
logit = self.fc(out)
|
|
return logit
|