Last active
April 10, 2020 22:40
-
-
Save kashif/ff44b17a6da18ec5128678d100c3818f to your computer and use it in GitHub Desktop.
EvoNorm-S0 in PyTorch from https://arxiv.org/pdf/2004.02967.pdf
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
import torch | |
import torch.nn as nn | |
class EvoNorm2d(nn.Module): | |
__constants__ = ['num_features', 'eps', 'nonlinearity'] | |
def __init__(self, num_features, eps=1e-5, nonlinearity=True): | |
super(EvoNorm2d, self).__init__() | |
self.num_features = num_features | |
self.eps = eps | |
self.nonlinearity = nonlinearity | |
self.weight = nn.Parameter(torch.Tensor(1, num_features, 1, 1)) | |
self.bias = nn.Parameter(torch.Tensor(1, num_features, 1, 1)) | |
if self.nonlinearity: | |
self.v = nn.Parameter(torch.Tensor(1, num_features, 1, 1)) | |
self.reset_parameters() | |
def reset_parameters(self): | |
nn.init.ones_(self.weight) | |
nn.init.zeros_(self.bias) | |
if self.nonlinearity: | |
nn.init.ones_(self.v) | |
def group_std(self, x, groups=32): | |
N, C, H, W = x.shape | |
x = torch.reshape(x, (N, groups, C//groups, H, W)) | |
std = torch.std(x, (2, 3, 4), keepdim=True).expand_as(x) | |
return torch.reshape(std + self.eps, (N, C, H, W)) | |
def forward(self, x): | |
if self.nonlinearity: | |
num = x * torch.sigmoid(self.v * x) | |
return num/self.group_std(x) * self.weight + self.bias | |
else: | |
return x * self.weight + self.bias |
@pinouchon https://github.com/digantamisra98/EvoNorm I have EvoNorm B0 here, however I just have one error of shape mismatch in the running variance calculation to solve.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Do you have the EvoNorm-B0 by any chance? It looks like this is EvoNorm-S0. I changed it to 1d and it seems to work fine, but I would still prefer the batch version