Hi,

to make easy use of the layer, I noticed that it can help to add a groups argument to the initialization of the layer. This groups argument can then be given to the group_std calculation to adjust for the amount of groups that one wants to define.
Otherwise you are stuck with the default value of 32.

Kind regards

Hi, thank you for the nice implementation. I've got a question about how you use the EvoNorm S0 in the BasicBlock (your block definition):

        self.evo = EvoNorm2D(planes)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(planes, planes)
        self.bn2 = norm_layer(planes)

What would be the norm_layer in this case? My understanding of the original paper is that the norm_lyaer(planes) has to actually be EvoNorm2D(planes, non_linear=False). At least that is what I understood from Figure 5's capture and a footnote there (see the original paper).

What is your take on that?

作者你好，我在使用你的代码训练模型的时候，发现内存一直在涨，请问你是否发现有内存泄漏的问题？

Hi, i have read your code,but i want to know whether evo_norm could be used for sequence-data, for example, the dims in data just likes batch_size, seq_len, embedding_dims.
In other words, N, C, H, W = x.size(), does C correspond to embedding_dims?

I noticed that the first conv layer of ResNet is followed by BatchNorm Layer, I think it should also be followed by EvoNorm as well.

self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False) self.bn1 = nn.BatchNorm2d(64) self.relu = nn.ReLU(inplace=True)

How to judge which evonorm sample2d or evonorm batch2d to use

def __init__(self, input, non_linear = True, version = 'S0', efficien affine = True,

There is a space between efficien and affine. Not sure what should be written instead (erase "efficien" ?)

I am using this on my own computer vision dataset. My image size is small enough that when the image size ends up being (batch_size,n_channels,1,1), the EvoNorm begins returning NaN's. I know the reason is because the width=channel=1 because when I make my network smaller with less convolution layers i.e. width,channel>1, the NaN's go away. Is there any reason why this is the case?

Maybe this is my misunderstanding. But it looks like evonorm2d.py line 55 which is using self.eps, but there is no self.eps definition at init function.

Implement and example could refer to impl and train code.

I didn't struggle with training settings btw.