Add support for Relu2 in BF16 fused MoE#2864

Merged

aleozlx merged 7 commits intoflashinfer-ai:mainfrom

amitz-nv:support-bf16-relu2-trtllm-gen-fused-moe-kernel

Apr 13, 2026

Commits on Apr 13, 2026

Add support for Relu2 in BF16 fused MoE, update test_accordingly
amitz-nv
committed
Fix flashinfer/fused_moe.__init__.py import issues found by 'pre-commit run --all-files'
amitz-nv
committed
Use validateAndCastActivationType instead of unchecked static cast
amitz-nv
committed
Update trtllm-gen batched GEMM artifact path & checksum, update csrc/trtllm_batched_gemm_runner.cu access to BatchedGemmOptions.mNumStages as it was split to A and B
amitz-nv
committed
MoE test - Add FP4_NVFP4_NVFP4 to allowed quant_mode with logits_dtype of fp32
amitz-nv
committed
Update docstrings with gemm1_weights to include non-gated activation shape
amitz-nv
committed
clang format fix
amitz-nv
committed