corentin/torch_utils
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

Fix batch norm training default value

Browse source
This commit is contained in:
corentin 2020-12-01 11:43:02 +09:00
commit 4b786943f5
1 changed files with 4 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

8
layers.py
View file

@ -19,7 +19,7 @@ class Layer(nn.Module):
ACTIVATION = F.leaky_relu
BATCH_NORM = True
BATCH_NORM_TRAINING = False
BATCH_NORM_TRAINING = True
BATCH_NORM_MOMENTUM = 0.01
IS_TRAINING = False
@ -28,7 +28,7 @@ class Layer(nn.Module):
LOGGER = DummyLogger()
def __init__(self, activation, batch_norm):
super(Layer, self).__init__()
super().__init__()
self.name = 'Layer'
self.info = LayerInfo()
@ -55,7 +55,7 @@ class Conv1d(Layer):
self.batch_norm = nn.BatchNorm1d(
out_channels,
momentum=Layer.BATCH_NORM_MOMENTUM,
track_running_stats=not Layer.BATCH_NORM_TRAINING) if self.batch_norm else None
track_running_stats=Layer.BATCH_NORM_TRAINING) if self.batch_norm else None
def forward(self, input_data: torch.Tensor) -> torch.Tensor:
return super().forward(self.conv(input_data))
@ -87,7 +87,7 @@ class Conv3d(Layer):
self.batch_norm = nn.BatchNorm3d(
out_channels,
momentum=Layer.BATCH_NORM_MOMENTUM,
track_running_stats=not Layer.BATCH_NORM_TRAINING) if self.batch_norm else None
track_running_stats=Layer.BATCH_NORM_TRAINING) if self.batch_norm else None
def forward(self, input_data: torch.Tensor) -> torch.Tensor:
return super().forward(self.conv(input_data))