-
Notifications
You must be signed in to change notification settings - Fork 6
Open
Description
Hello,
in unet.py you create ConvBlock-objects with (for example) this call:
self.conv1 = ConvBlock(n_channel_in, 32, residual, activation)
But the constructor of ConvBlock is __init__(self, in_channels, out_channels, dropout=False, norm='batch', residual=True, activation='leakyrelu', transpose=False), resulting in self.dropout being assigned to residual and self.norm being set to activation.
You probably wanted to call it like this:
self.conv1 = ConvBlock(n_channel_in, 32, residual=residual, activation=activation) and so on.
Minimal example:
class ConvBlock():
def __init__(self, in_channels, out_channels, dropout=False, norm='batch',
residual=True, activation='leakyrelu', transpose=False):
self.dropout = dropout
self.residual = residual
self.activation = activation
self.transpose = transpose
self.norm = norm
def printme(self):
print('Dropout: ', self.dropout)
print('Residual: ', self.residual)
print('Activation: ', self.activation)
print('Transpose: ', self.transpose)
print('Norm: ', self.norm)
residual = True
activation = 'relu'
c = ConvBlock(32, 64, residual, activation)
c.printme()
"""Dropout: True
Residual: True
Activation: leakyrelu
Transpose: False
Norm: relu
"""
c2 = ConvBlock(32, 64, residual=residual, activation=activation)
c2.printme()
"""
Dropout: False
Residual: True
Activation: relu
Transpose: False
Norm: batch
"""
Metadata
Metadata
Assignees
Labels
No labels