ROCm / DeepSpeed

DeepSpeed is a deep learning optimization library that makes distributed training easy, efficient, and effective.
https://www.deepspeed.ai/
Apache License 2.0
4 stars 3 forks source link

Replace THCudaCheck with C10_CUDA_CHECK #51

Closed rraminen closed 2 years ago

rraminen commented 2 years ago

This issue occurs as THCudaCheck is deprecated and PyTorch now considers THCDeviceUtils.cuh, THC/THC.h.

DeepSpeed still builds successfully in ROCm4.3.1 docker (rocm/pytorch:rocm4.3.1_ubuntu18.04_py3.6_pytorch_1.10.0) which has an older Pytorch versions.

Fixes https://ontrack-internal.amd.com/browse/SWDEV-314993

Reference: https://github.com/ROCmSoftwarePlatform/apex/pull/55/commits/fec3141c33c23da9f790700817f7496158080712

cc: @jithunnair-amd