diff --git a/python/sglang/srt/models/deepseek_v2.py b/python/sglang/srt/models/deepseek_v2.py index ab67e03a61b6..2b1c3c04d6cf 100644 --- a/python/sglang/srt/models/deepseek_v2.py +++ b/python/sglang/srt/models/deepseek_v2.py @@ -3202,7 +3202,6 @@ def forward( class DeepseekV2ForCausalLM(nn.Module): # for quark model load packed_modules_mapping = {} - model_cls = DeepseekV2Model def __init__( self, @@ -3228,7 +3227,7 @@ def __init__( self.quant_config = quant_config self.determine_num_fused_shared_experts() self.use_nsa = is_deepseek_nsa(config) - self.model = self.model_cls( + self.model = DeepseekV2Model( config, quant_config, prefix=add_prefix("model", prefix) ) if self.pp_group.is_last_rank: