From 0f7919fca05d7cf60b773da26d898b72bc07a089 Mon Sep 17 00:00:00 2001 From: Jee Jee Li Date: Thu, 31 Jul 2025 11:41:12 +0800 Subject: [PATCH] [Misc] Expand SUPPORTED_HIDDEN_SIZES for DeepEP low-latency kernels (#21818) Signed-off-by: Jee Jee Li --- .../layers/fused_moe/deepep_ll_prepare_finalize.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/vllm/model_executor/layers/fused_moe/deepep_ll_prepare_finalize.py b/vllm/model_executor/layers/fused_moe/deepep_ll_prepare_finalize.py index 57871ca250ae3..cfc2bdcf02408 100644 --- 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,