EngineX-Hygon/sglang
5
0
Fork 0
Code Issues Pull Requests Actions 7 Projects Releases Wiki Activity

[ci] disable memory imbalance check for draft worker (#8108)

Browse Source
This commit is contained in:
Cheng Wan
2025-07-16 20:41:47 -07:00
committed by GitHub
parent 8a7a7770e5
commit 9069884b51
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
python/sglang/srt/model_executor/model_runner.py
View File

@@ -561,7 +561,7 @@ class ModelRunner:
# Check memory for tensor parallelism # Check memory for tensor parallelism
local_gpu_memory = get_available_gpu_memory(self.device, self.gpu_id) local_gpu_memory = get_available_gpu_memory(self.device, self.gpu_id)
if self.tp_size > 1: if self.tp_size > 1 and not self.is_draft_worker:
if min_per_gpu_memory < local_gpu_memory * 0.9: if min_per_gpu_memory < local_gpu_memory * 0.9:
if get_bool_env_var("SGL_DISABLE_TP_MEMORY_INBALANCE_CHECK"): if get_bool_env_var("SGL_DISABLE_TP_MEMORY_INBALANCE_CHECK"):
logger.warning( logger.warning(