[Misc] Expand SUPPORTED_HIDDEN_SIZES for DeepEP low-latency kernels (#21818)

Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>

[Misc] Expand SUPPORTED_HIDDEN_SIZES for DeepEP low-latency kernels (#21818)
Signed-off-by: Jee Jee Li <pandaleefree@gmail.com>
0f7919fc · Jee Jee Li · GitHub · 61445453 · 0f7919fc
Unverified Commit 0f7919fc authored Jul 31, 2025 by Jee Jee Li Committed by GitHub Jul 30, 2025
Show whitespace changes
Inline Side-by-side

Showing with 1 addition and 1 deletion

vllm/model_executor/layers/fused_moe/deepep_ll_prepare_finalize.py ...l_executor/layers/fused_moe/deepep_ll_prepare_finalize.py +1 -1

No files found.
--- a/vllm/model_executor/layers/fused_moe/deepep_ll_prepare_finalize.py
+++ b/vllm/model_executor/layers/fused_moe/deepep_ll_prepare_finalize.py
@@ -40,7 +40,7 @@ class DeepEPLLPrepareAndFinalize(mk.FusedMoEPrepareAndFinalize):

    # DeepEP low-latency kernels are compiled only for certain
    # specific hidden sizes.
-    SUPPORTED_HIDDEN_SIZES = [2048, 2560, 4096, 5120, 7168]
+    SUPPORTED_HIDDEN_SIZES = [2048, 2560, 4096, 5120, 6144, 7168]

    def __init__(self,
                 buffer: deep_ep.Buffer,