Closed sahilkhose closed 3 years ago
https://github.com/sahilkhose/ganctober/blob/919fe491f310ffe06e28e7306cbc26dad70d2d86/src/layers.py#L73 Conv2d default bias=True : https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html
A very common mistake is to follow Conv2d with BatchNorm2d and to not mention bias=False in conv2d BatchNorm2d has bias : https://pytorch.org/docs/stable/generated/torch.nn.BatchNorm2d.html
Hence 2 biases just adds to the model parameters without any help in the computation It is a common practice to set bias=False in conv whenever it is followed by a BatchNorm
https://github.com/sahilkhose/ganctober/blob/919fe491f310ffe06e28e7306cbc26dad70d2d86/src/layers.py#L73 Conv2d default bias=True : https://pytorch.org/docs/stable/generated/torch.nn.Conv2d.html
A very common mistake is to follow Conv2d with BatchNorm2d and to not mention bias=False in conv2d BatchNorm2d has bias : https://pytorch.org/docs/stable/generated/torch.nn.BatchNorm2d.html
Hence 2 biases just adds to the model parameters without any help in the computation It is a common practice to set bias=False in conv whenever it is followed by a BatchNorm