[Kernel] Enable fp8 support for pplx and BatchedTritonExperts. (#18864)

Signed-off-by: Bill Nell <bnell@redhat.com>
2025-07-03 17:55:40 -04:00
parent 2f2fcb31b8
commit 78fe77534b
25 changed files with 1277 additions and 663 deletions
--- a/vllm/model_executor/layers/fused_moe/modular_kernel.py
+++ b/vllm/model_executor/layers/fused_moe/modular_kernel.py
@@ -193,6 +193,10 @@ class FusedMoEPrepareAndFinalize(ABC):
        """
        raise NotImplementedError

+    @abstractmethod
+    def num_dispatchers(self) -> int:
+        raise NotImplementedError
+

 class FusedMoEPermuteExpertsUnpermute(ABC):
    """