Vadim Kantorov

Also stumbled on this: https://github.com/pytorch/pytorch/issues/44544

My related attempts / investigation on ways of NumPy C# arrays interop: - https://github.com/pythonnet/pythonnet/discussions/2226 - I also discovered a zero-copy way of NumPy -> C# array interop: https://gist.github.com/vadimkantorov/c9ab97aff995ea01ea7f6248ddc733bc#file-csharpfrompython-cs-L23-L41 using some...

Related old discussion on this being a useful primitive (in the old days for collate_fn of data loading) and deserving more fame :) - https://github.com/pytorch/pytorch/issues/65156 One useful thing here is...

Maybe one way to auto-construct NJT from torch.stack([...]) call in default collate could be: 1) from `dataset.__getitem__` return dense tensors but wrapped in NJT (but the internal representation should be...

I used this compression trick in my Python CTC-based forced alignment: https://github.com/vadimkantorov/ctc/blob/master/ctc.py And this compression indeed works and helps

Hi! Thanks for interest in my hobby project :) Some my thoughts on data packaging: - I do not have huge experience in TeX packaging / TDS and making it...

I think your approach might work if your tex sources don't need to do any font compilation / generation / other post processing at dependent package installation step

I think I'm still not understanding your complete usecase :( Please explain your full usecase with an example

Will the fonts for your usecase get compiled at the latex compilation time? If so, how? `tlmgr` includes a perl binary which is needed for running updmap/fmtutil and other scripts,...

(I've fixed the master branch and cleaned it up)