External-Attention-pytorch

External-Attention-pytorch copied to clipboard

1. padding is set to "same" because of the dialation, otherwise `res=res.expand_as(x)` in line 49 will fail. 2. removed unnecessary `b, c, _, _ = x.size()` since it's never used....

KevinOfCathay

如果是mmdetection框架该如何使用这个库呢

PeterSmith1

receptive_field_size = tensor[0][0].numel() IndexError index 0 is out of bounds for dimension 0

Lijinxisa

部分模块在使用GPU计算时报错(Most modules report errors when using GPU)

2

报错内容类似于： RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same 或者 Expected all tensors to be on the same device, but found at least two devices, cuda:0...

ImcwjHere

当输入为两个方向的图片,请问有没有利用计算相似度矩阵作为权重,把两个方向特征图进行融合的模块或者方法?谢谢!

Floyd809

使用SEA出现的问题

1

 请问一下如图在使用SEA出现的问题应该怎么解决呢

mmmmyolo

I notice the code in BAM is : ` def forward(self, x): b, c, _, _ = x.size() sa_out=self.sa(x) ca_out=self.ca(x) weight = self.sigmoid(sa_out*ca_out) weight=self.sigmoid(sa_out+ca_out) # here out=(1+weight)*x return out` here...

Luo-Z13

Can you add some comments in the code to explain the meaning of each parameter, which will be more friendly to novices, thank you. For example, what do d_model and...

scarLiu

亲爱的作者你好，一维卷积神经网络可以使用ECA或者SE注意力机制吗？

3

打扰一下，我正在做风电数据的故障诊断，数据是100000*80，我想利用1D-CNN来进行实验，但是我发现关于卷积常用的ECA,SE注意力机制，一般都是处理4维数据，也就是处理图像数据。冒昧地问一下，ECA或者SE注意力机制可以处理1D-CNN吗，期待您的回复

qiaolikui

SKAttention似乎有点小问题

1

单运行./model/attention/SKAttention.py时会报错，RuntimeError: Can't call numpy() on Tensor that requires grad. Use tensor.detach().numpy() instead. 目测应该是第40行U=sum(conv_outs)中调用sum时需要将tensor转numpy，而被转tensor参与梯度计算不能直接转？

AAAAADA