Skip FlashInfer FP8 autotune for MXFP8 quantized models (#29462)

2026-06-26 18:15:08 -07:00
parent 13b5bd962a
commit 09ca4fc96b
1 changed files with 7 additions and 1 deletions
@@ -302,7 +302,13 @@ class BaseRunner(ABC):
            "modelopt_fp8",
            "modelopt_mixed",
        )
-        fp8_gemm_needs_autotune = (
+        # Online MXFP8 (microscaling) linears dispatch to flashinfer's
+        # ``mm_mxfp8``, which the flashinfer fp8 autotune dummy run does not
+        # exercise correctly -- it triggers an illegal memory access inside the
+        # mxfp8 cutlass cubin. The mxfp8 gemm is fixed-config and needs no
+        # tuning, so skip autotune for these models.
+        model_uses_mxfp8 = "mxfp8" in (mr.model_config.quantization or "")
+        fp8_gemm_needs_autotune = not model_uses_mxfp8 and (
            get_fp8_gemm_runner_backend().is_flashinfer_cutlass()
            or (model_uses_modelopt_fp8 and is_sm100_supported())
        )