ChengShen comments

Repositories
Issues
Comments

Results 12 comments of


                                            ChengShen

Question about the AdaptiveInstanceNorm2d Implementation with F.batch_norm

`x_reshaped = x.contiguous().view(1, b * c, *x.size()[2:])` I think it is the answer why they used BN function.

Curious about HRM on NLP tasks.

Hi! I noticed that your code feeds the model identical sequences for input and output. This encourages the network to memorize the identity mapping and overfit. During evaluation, do you...