maxymnaumchyk

Results 6 comments of maxymnaumchyk

fix: add cuda backend support for `to_raggedtensor` and `from_raggedtensor` functions

@jpivarski while trying to make the `to_raggedtensor` function keep the device of the original awkward array I stumbled upon an issue. The thing is, tensorflow automatically selects gpu for computation,...

feat: to/from PyTorch JaggedTensor

@jpivarski should I also leave out a "keep_regular" parameter since it does the same as ak.from_regular()? It's kind of the same situation we talked about today (about "padded" parameter).

feat: to/from PyTorch JaggedTensor

thanks for such a detailed answer!

docs: Add an example argmax reducer using cccl

Sure, here: Line 861 is where the `localindex` [cuda_kernel](https://github.com/scikit-hep/awkward/blob/main/src/awkward/_connect/cuda/cuda_kernels/awkward_ListArray_localindex.cu) is called.

docs: Add an example argmax reducer using cccl

Thank you for the feedback @shwina! How did you point down that the issue is related to `DeviceSegmentedReduce` function?

docs: Add an example argmax reducer using cccl

Thanks a lot for the explanation @shwina! Yes, now after adding `cp.cuda.Device().synchronize()` calls, I get similar results to yours. An unrelated question: the `cudaDeviceSynchronize` process that I see on host...