[Core] Fix MXFP8 grouped quantize for zero-sized groups in update_tma_descriptors - #2782

[Core] Fix MXFP8 grouped quantize for zero-sized groups in update_tma_descriptors#2782

jberchtold-nvidia merged 4 commits into

NVIDIA:mainNVIDIA/TransformerEngine:mainfrom

jberchtold-nvidia:jberchtold/fix-mxpf8-grouped-quantize-tma-for-groups-with-size-zerojberchtold-nvidia/TransformerEngine:jberchtold/fix-mxpf8-grouped-quantize-tma-for-groups-with-size-zeroCopy head branch name to clipboard