From 5239d79568f3b5ce55106cb3c9d9bee7cc8e7477 Mon Sep 17 00:00:00 2001 From: fzyzcjy <5236035+fzyzcjy@users.noreply.github.com> Date: Mon, 21 Apr 2025 09:12:01 +0800 Subject: [PATCH] Speedup shared expert weight construction by avoid cloning (#5188) --- python/sglang/srt/models/deepseek_v2.py | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index 26c5e617a..2cd64d532 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -1628,7 +1628,7 @@ class DeepseekV2ForCausalLM(nn.Module): f"mlp.experts." f"{self.config.n_routed_experts + num_repeat}" f".{suffix}", - weights_dict[shared_expert_weight_name].clone(), + weights_dict[shared_expert_weight_name], ) ) names_to_remove += [shared_expert_weight_name]