From 4f204db57cf96b3c85df427868f9d2f9949c46df Mon Sep 17 00:00:00 2001 From: Yineng Zhang Date: Tue, 17 Jun 2025 16:22:46 -0700 Subject: [PATCH] fix: resolve b200 dsv3 mtp issue (#7286) --- python/sglang/srt/models/deepseek_v2.py | 2 ++ 1 file changed, 2 insertions(+) diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index 4cf898d43..017c103ba 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -1932,6 +1932,8 @@ class DeepseekV2ForCausalLM(nn.Module): self._weight_requant_ue8m0() def _weight_requant_ue8m0(self): + if self.config.architectures[0] == "DeepseekV3ForCausalLMNextN": + return weight_block_size = self.quant_config.weight_block_size moe_layers = list(