minke.yu/sglang
1
0
Fork 0
Code Issues Pull Requests Actions 12 Packages Projects Releases Wiki Activity

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

Browse Source
This commit is contained in:
Jimmy Shong
2026-04-15 13:25:36 -07:00
committed by GitHub
parent 8686f42acb
commit 28e915b474
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/srt/models/glm4_moe_nextn.py
+2 -1
View File
@@ -129,7 +129,8 @@ class Glm4MoeForCausalLMNextN(Glm4MoeForCausalLM):
self.config = config self.config = config
self.tp_size = get_tensor_model_parallel_world_size() self.tp_size = get_tensor_model_parallel_world_size()
self.needs_quant_draft = ( 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 quant_config = quant_config if self.needs_quant_draft else None
self.model = Glm4MoeModelNextN( self.model = Glm4MoeModelNextN(