[None][feat] Add W4A16 per-channel weight-only support for Cutlass fused MoE - #18313

[None][feat] Add W4A16 per-channel weight-only support for Cutlass fused MoE#18313

Dorijan10 wants to merge 25 commits into

NVIDIA:mainNVIDIA/TensorRT-LLM:mainfrom

Dorijan10:w4a16-moe-per-channelDorijan10/TensorRT-LLM-Optimisation:w4a16-moe-per-channelCopy head branch name to clipboard