diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index b0a3f20f6..330ad5133 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -360,6 +360,9 @@ class DeepseekV2MoE(nn.Module): correction_bias=self.correction_bias, routed_scaling_factor=self.routed_scaling_factor, num_token_non_padded=forward_batch.num_token_non_padded, + expert_location_dispatch_info=ExpertLocationDispatchInfo.init_new( + layer_id=self.layer_id, + ), ) else: topk_idx = torch.full(