[Kernels] Add activation chunking logic to FusedMoEModularKernel (#19168)

Signed-off-by: Bill Nell <bnell@redhat.com>
2025-06-11 12:53:10 -04:00
parent b2d9be6f7d
commit 29fa5cac1c
15 changed files with 458 additions and 396 deletions
--- a/tests/kernels/moe/test_cutlass_moe.py
+++ b/tests/kernels/moe/test_cutlass_moe.py
@@ -29,6 +29,7 @@ MNK_FACTORS = [
    (224, 1024, 1536),
    (224, 3072, 1024),
    (224, 3072, 1536),
+    (1024 * 128, 1024, 1024),
 ]

 vllm_config = VllmConfig(parallel_config=ParallelConfig(