Quantization support for GroupedTensor: FP8 per-tensor#3102

int-smart wants to merge 8 commits into

NVIDIA:mainNVIDIA/TransformerEngine:mainfrom

int-smart:feature/fp8_quantint-smart/TransformerEngine:feature/fp8_quantCopy head branch name to clipboard