From 0e0b0c0566fe2917195c8cdf161f8ed60aa06074 Mon Sep 17 00:00:00 2001 From: Yineng Zhang Date: Mon, 8 Dec 2025 20:06:52 -0800 Subject: [PATCH] =?UTF-8?q?Revert=20"[Bug]=20fix=20not=20desired=20disable?= =?UTF-8?q?=20fused=20share=20experts=20caused=20by=20r=E2=80=A6=20(#14676?= =?UTF-8?q?)?= MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit --- python/sglang/srt/models/deepseek_v2.py | 6 ++---- 1 file changed, 2 insertions(+), 4 deletions(-) diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index ea8ec92a5..ab67e03a6 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -3292,10 +3292,8 @@ class DeepseekV2ForCausalLM(nn.Module): "Only Deepseek V3/R1 on NV-platform with capability >= 80 " "or AMD-platform with capability >= gfx942(MI30x) can use shared experts fusion optimization." ) - elif ( - get_moe_expert_parallel_world_size() > 1 - and _is_hip - and torch.cuda.get_device_capability("cuda") < (9, 4) + elif get_moe_expert_parallel_world_size() > 1 and ( + not _is_hip or torch.cuda.get_device_capability("cuda") < (9, 4) ): disable_reason = "Only Deepseek V3/R1 on AMD-platform with capability >= gfx942(MI30x) can use shared experts fusion optimization under expert parallelism." elif disable_reason is None and get_moe_a2a_backend().is_deepep():