[ROCm][MiniMax-M3.1] Fuse MXFP8 activation quant into the Gemma input_layernorm for qkv_proj - #63637

[ROCm][MiniMax-M3.1] Fuse MXFP8 activation quant into the Gemma input_layernorm for qkv_proj#63637

shen-shanshan merged 7 commits into

vllm-project:mainvllm-project/vllm:mainfrom

RuibinCheung:cursor/minimax-m3-rocm-gemma-norm-mxfp8-quant-edd1RuibinCheung/vllm:cursor/minimax-m3-rocm-gemma-norm-mxfp8-quant-edd1Copy head branch name to clipboard