Fix shared experts fusion error without quantization (#5632)

2025-04-23 00:22:26 +08:00
parent 70d040f904
commit 76d17c7ecb
1 changed files with 1 additions and 1 deletions
--- a/python/sglang/srt/models/deepseek_v2.py
+++ b/python/sglang/srt/models/deepseek_v2.py
@@ -1604,7 +1604,7 @@ class DeepseekV2ForCausalLM(nn.Module):
        if self.n_share_experts_fusion > 0:
            weights_list = list(weights)
            weights_dict = dict(weights_list)
-            if self.quant_config.get_name() == "w8a8_int8":
+            if self.quant_config is None or self.quant_config.get_name() == "w8a8_int8":
                suffix_list = [
                    "down_proj.weight",
                    "down_proj.weight_scale",