search

cai4cai / torchsparsegradutils

A collection of utility functions to work with PyTorch sparse tensors
Apache License 2.0
24 stars 3 forks source link

RuntimeError: "addmm_sparse_cuda" not implemented for 'Half' #53

Closed EvgenyUgolkov closed 8 months ago

EvgenyUgolkov commented 8 months ago

Dear Team, good day! I try to run the sparse mm with the float16 tensors, and faced the following error:

RuntimeError: "addmm_sparse_cuda" not implemented for 'Half'

Could you tell me, do you know a potential solutions to it?

Regards, Evgeny

tvercaut commented 8 months ago

This is out of scope for this repo and likely a pytorch limitation. See e.g. https://github.com/pytorch/pytorch/issues/41069 -> closing this

EvgenyUgolkov commented 8 months ago

This is out of scope for this repo and likely a pytorch limitation. See e.g. pytorch/pytorch#41069 -> closing this

@tvercaut

Thank you!