[Bugfix] Preserve auto-detected quant_config for GLM NextN draft model (#22823)

2026-04-15 13:25:36 -07:00
parent 8686f42acb
commit 28e915b474
1 changed files with 2 additions and 1 deletions
@@ -129,7 +129,8 @@ class Glm4MoeForCausalLMNextN(Glm4MoeForCausalLM):
        self.config = config
        self.tp_size = get_tensor_model_parallel_world_size()
        self.needs_quant_draft = (
-            get_global_server_args().speculative_draft_model_quantization
+            get_global_server_args().speculative_draft_model_quantization is not None
+            or quant_config is not None
        )
        quant_config = quant_config if self.needs_quant_draft else None
        self.model = Glm4MoeModelNextN(