|
|
@ -1,3 +1,21 @@
|
|
|
|
|
|
|
|
""" Convolution with Weight Standardization (StdConv and ScaledStdConv)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
StdConv:
|
|
|
|
|
|
|
|
@article{weightstandardization,
|
|
|
|
|
|
|
|
author = {Siyuan Qiao and Huiyu Wang and Chenxi Liu and Wei Shen and Alan Yuille},
|
|
|
|
|
|
|
|
title = {Weight Standardization},
|
|
|
|
|
|
|
|
journal = {arXiv preprint arXiv:1903.10520},
|
|
|
|
|
|
|
|
year = {2019},
|
|
|
|
|
|
|
|
}
|
|
|
|
|
|
|
|
Code: https://github.com/joe-siyuan-qiao/WeightStandardization
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
ScaledStdConv:
|
|
|
|
|
|
|
|
Paper: `Characterizing signal propagation to close the performance gap in unnormalized ResNets`
|
|
|
|
|
|
|
|
- https://arxiv.org/abs/2101.08692
|
|
|
|
|
|
|
|
Official Deepmind JAX code: https://github.com/deepmind/deepmind-research/tree/master/nfnets
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Hacked together by / copyright Ross Wightman, 2021.
|
|
|
|
|
|
|
|
"""
|
|
|
|
import torch
|
|
|
|
import torch
|
|
|
|
import torch.nn as nn
|
|
|
|
import torch.nn as nn
|
|
|
|
import torch.nn.functional as F
|
|
|
|
import torch.nn.functional as F
|
|
|
@ -5,12 +23,6 @@ import torch.nn.functional as F
|
|
|
|
from .padding import get_padding, get_padding_value, pad_same
|
|
|
|
from .padding import get_padding, get_padding_value, pad_same
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
def get_weight(module):
|
|
|
|
|
|
|
|
std, mean = torch.std_mean(module.weight, dim=[1, 2, 3], keepdim=True, unbiased=False)
|
|
|
|
|
|
|
|
weight = (module.weight - mean) / (std + module.eps)
|
|
|
|
|
|
|
|
return weight
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
class StdConv2d(nn.Conv2d):
|
|
|
|
class StdConv2d(nn.Conv2d):
|
|
|
|
"""Conv2d with Weight Standardization. Used for BiT ResNet-V2 models.
|
|
|
|
"""Conv2d with Weight Standardization. Used for BiT ResNet-V2 models.
|
|
|
|
|
|
|
|
|
|
|
@ -30,7 +42,7 @@ class StdConv2d(nn.Conv2d):
|
|
|
|
def forward(self, x):
|
|
|
|
def forward(self, x):
|
|
|
|
weight = F.batch_norm(
|
|
|
|
weight = F.batch_norm(
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
eps=self.eps, training=True, momentum=0.).reshape_as(self.weight)
|
|
|
|
training=True, momentum=0., eps=self.eps).reshape_as(self.weight)
|
|
|
|
x = F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
x = F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
return x
|
|
|
|
return x
|
|
|
|
|
|
|
|
|
|
|
@ -56,7 +68,7 @@ class StdConv2dSame(nn.Conv2d):
|
|
|
|
x = pad_same(x, self.kernel_size, self.stride, self.dilation)
|
|
|
|
x = pad_same(x, self.kernel_size, self.stride, self.dilation)
|
|
|
|
weight = F.batch_norm(
|
|
|
|
weight = F.batch_norm(
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
eps=self.eps, training=True, momentum=0.).reshape_as(self.weight)
|
|
|
|
training=True, momentum=0., eps=self.eps).reshape_as(self.weight)
|
|
|
|
x = F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
x = F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
return x
|
|
|
|
return x
|
|
|
|
|
|
|
|
|
|
|
@ -86,7 +98,7 @@ class ScaledStdConv2d(nn.Conv2d):
|
|
|
|
weight = F.batch_norm(
|
|
|
|
weight = F.batch_norm(
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
weight=(self.gain * self.scale).view(-1),
|
|
|
|
weight=(self.gain * self.scale).view(-1),
|
|
|
|
eps=self.eps, training=True, momentum=0.).reshape_as(self.weight)
|
|
|
|
training=True, momentum=0., eps=self.eps).reshape_as(self.weight)
|
|
|
|
return F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
return F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
@ -117,5 +129,5 @@ class ScaledStdConv2dSame(nn.Conv2d):
|
|
|
|
weight = F.batch_norm(
|
|
|
|
weight = F.batch_norm(
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
self.weight.view(1, self.out_channels, -1), None, None,
|
|
|
|
weight=(self.gain * self.scale).view(-1),
|
|
|
|
weight=(self.gain * self.scale).view(-1),
|
|
|
|
eps=self.eps, training=True, momentum=0.).reshape_as(self.weight)
|
|
|
|
training=True, momentum=0., eps=self.eps).reshape_as(self.weight)
|
|
|
|
return F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|
return F.conv2d(x, weight, self.bias, self.stride, self.padding, self.dilation, self.groups)
|
|
|
|