The module seems not effective at all

[DingLei14]

For my datasets the visualization of the residual part seems to be meaningless noises.

[kiyohiro8]

How did you find out that the GloRe unit seems ineffective?

[DingLei14]

How did you find out that the GloRe unit seems ineffective?

I compared the ourput before & after using the module, and also visualized the residual part of the module (the 'y' part in 'x+y' in your code).
Actually, I found another implementation effective. I am not sure what's the difference between this implementation and yours though.

[kiyohiro8]

Thanks! I think some bugs are in my implementation. I will review and fix the code.

[gogowhy]

could. you point out how to visualize the residual part?

[HauWong]

Thanks! I think some bugs are in my implementation. I will review and fix the code.

Hello, kiyohiro8. I wonder if you have managed this bug fix work? If yes, could you please commit it? Thanks very much.

[kiyohiro8]

I've been busy and haven't been able to get my hands on it, but I'll be making corrections within the next month. Please wait a little longer.

[NeverGiveU]

Thanks! I think some bugs are in my implementation. I will review and fix the code.

Maybe it is because a line y = torch.transpose(y, 1, 2) is missed between #37 and #38 in layers/py ?