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

[args] Add Expert Parallelism Argument To SRT Runner (#18492)

Browse Source 
Co-authored-by: Qiaolin Yu <liin1211@outlook.com>
This commit is contained in:
Jonah Bernard
2026-03-03 14:16:35 -08:00
committed by GitHub
co-authored by Qiaolin Yu
parent f7897def96
commit fb37c0a400
1 changed files with 2 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
python/sglang/test/runners.py
+2
View File
@@ -522,6 +522,7 @@ class SRTRunner:
torch_dtype: torch.dtype, torch_dtype: torch.dtype,
model_type: str, model_type: str,
tp_size: int = 1, tp_size: int = 1,
ep_size: int = 1,
model_impl: str = "auto", model_impl: str = "auto",
port: int = DEFAULT_PORT_FOR_SRT_TEST_RUNNER, port: int = DEFAULT_PORT_FOR_SRT_TEST_RUNNER,
lora_paths: Optional[Union[List[str], List[dict[str, str]]]] = None, lora_paths: Optional[Union[List[str], List[dict[str, str]]]] = None,
@@ -588,6 +589,7 @@ class SRTRunner:
self.engine = Engine( self.engine = Engine(
model_path=model_path, model_path=model_path,
tp_size=tp_size, tp_size=tp_size,
ep_size=ep_size,
dtype=get_dtype_str(torch_dtype), dtype=get_dtype_str(torch_dtype),
port=port, port=port,
model_impl=model_impl, model_impl=model_impl,