[PyTorch] Deferred backward quantization state update (removes first-FP8-module ownership) - #34

[PyTorch] Deferred backward quantization state update (removes first-FP8-module ownership)#34

pggPL wants to merge 8 commits into

mainpggPL/TransformerEngine:mainfrom

bwd_state_update_refactorpggPL/TransformerEngine:bwd_state_update_refactorCopy head branch name to clipboard