[PyTorch] Add optional caller-provided output/grad-input buffers to GroupedLinear and fused grouped MLP#3161

timmoon10 merged 11 commits into

NVIDIA:mainNVIDIA/TransformerEngine:mainfrom

phu0ngng:pyt-gg-w-symmphu0ngng/TransformerEngine:pyt-gg-w-symmCopy head branch name to clipboard