xvdp

syncing your code I see that you removed LayerNorm and replaced it by a GroupNorm, although you still ahve LayerNorm in the LinearAttention Blocks. I still dont understand that 'LayerNorm',...

thank you for the answer: ill make an example on a static object - although when i went thru the code I noticed that you have a no warp option,...

@keunhong I got around to run the exact same test with one simple change: `ModelConfig.use_warp = False` now it renders perfectly, as I said, the character scanned was very still...

Writing the transform with the .copy() op I realized that it makes no sense to have torchvision transformations see accimage and PIL as the same when their behaviour is different,...

no worry. I only touched 3 functions - that I needed. if you really want to speed it up you could write it in cuda and save yourself allocation of...

Same thing happened to me yesterday - after an odd event in my ubuntu 16. I opened ubuntu after leaving it on overnight. It was logged in as guest instead...

@nateanl I did do an PR but didnt add any tests. a. I figured that any existing checks would cover this. b. locally i compared both versions of the resample...

@adefossez larger kernels are an issue if you need exact resampling, but if you want to randomize it then one doesnt need the exact gdc... , Theres more than one...

Also, sure, i understand that the version using arrays instead of a loop to create the kernel is NOT the solution for creation of lots of huge kernels. But it...

@adefossez you do have a point, quality does matter, but both audio resampling and shifting pitch degrade in quality quite quickly with range, I will look at float64 but I...