[RL] Fix FlashInfer TRTLLM MXFP8 dense weight layout (#28459)

This commit is contained in:
Ziang Li
2026-06-17 10:33:35 +00:00
committed by GitHub
parent 2f1390fcb1
commit 3fb65ebabd
3 changed files with 45 additions and 6 deletions
@@ -171,7 +171,7 @@ class FlashinferTrtllmGenMoeBackendMXFP8MixedBF16Base:
"--kv-cache-dtype",
"bf16",
"--fp8-gemm-backend",
"flashinfer_cutlass",
"flashinfer_trtllm",
"--moe-runner-backend",
cls.backend,
"--tp-size",