Skip to content

fix: use device-agnostic autocast and tensor placement for ROCm/XPU compatibility#15921

Open
zhihuidu-amd wants to merge 1 commit into
NVIDIA-NeMo:mainfrom
zhihuidu-amd:rocm/device-agnostic-autocast
Open

fix: use device-agnostic autocast and tensor placement for ROCm/XPU compatibility#15921
zhihuidu-amd wants to merge 1 commit into
NVIDIA-NeMo:mainfrom
zhihuidu-amd:rocm/device-agnostic-autocast