From 07452cbe8e278be4a4cfa950d76ac33b4a4b7b73 Mon Sep 17 00:00:00 2001 From: Chunyuan WU Date: Mon, 14 Jul 2025 16:32:43 +0800 Subject: [PATCH] [CPU] fix no attribute 'can_fuse_mlp_allreduce' error (#8010) --- python/sglang/srt/models/deepseek_v2.py | 8 +++++--- 1 file changed, 5 insertions(+), 3 deletions(-) diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index 5399d6904..aa47daa30 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -462,7 +462,7 @@ class DeepseekV2MoE(nn.Module): if hasattr(self, "shared_experts") and use_intel_amx_backend( self.shared_experts.gate_up_proj ): - return self.forward_cpu(hidden_states) + return self.forward_cpu(hidden_states, can_fuse_mlp_allreduce) shared_output = self._forward_shared_experts(hidden_states) # router_logits: (num_tokens, n_experts) @@ -479,7 +479,9 @@ class DeepseekV2MoE(nn.Module): final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states) return final_hidden_states - def forward_cpu(self, hidden_states: torch.Tensor) -> torch.Tensor: + def forward_cpu( + self, hidden_states: torch.Tensor, can_fuse_mlp_allreduce: bool = False + ) -> torch.Tensor: # router_logits: (num_tokens, n_experts) router_logits = self.gate(hidden_states) fused_experts_out = self.experts( @@ -528,7 +530,7 @@ class DeepseekV2MoE(nn.Module): None, # a2_scale True, # is_vnni ) - if self.tp_size > 1 and not self.can_fuse_mlp_allreduce: + if self.tp_size > 1 and not can_fuse_mlp_allreduce: final_hidden_states = tensor_model_parallel_all_reduce(final_hidden_states) return final_hidden_states