Using instancenorm with resnest

[jaideep11061982]

I try to replace BNs with instancenorm
but I get an error

AttributeError: 'NoneType' object has no attribute 'zero_' 

model=timm.create_model(
    'resnest26d',
    pretrained=True,norm_layer=torch.nn.InstanceNorm2d)

Since my batch size is 1 ,resenest dsnt work with bs as 1

[rwightman]

@jaideep11061982 fix on main now (6902c48)

You should not use pretrained=True if you're changing the norm layer...

You can make it work prior to the fix if you enable the affine params (probably works a bit better).

from functools import partial
model=timm.create_model(
    'resnest26d',
    pretrained=False,
    norm_layer=partial(torch.nn.InstanceNorm2d, affine=True))