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

Support LoRA in bench_serving oai interface (#11318)

Browse Source
This commit is contained in:
Lifu Huang
2025-10-08 01:28:58 -07:00
committed by GitHub
parent 6c0bb32711
commit 92473e2e34
1 changed files with 11 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

11
python/sglang/bench_serving.py
View File

@@ -209,6 +209,11 @@ async def async_request_openai_completions(
**request_func_input.extra_request_body, **request_func_input.extra_request_body,
} }
# hack to accommodate different LoRA conventions between SGLang and vLLM.
if request_func_input.lora_name:
payload["model"] = request_func_input.lora_name
payload["lora_path"] = request_func_input.lora_name
if request_func_input.image_data: if request_func_input.image_data:
payload.update({"image_data": request_func_input.image_data}) payload.update({"image_data": request_func_input.image_data})
@@ -326,6 +331,12 @@ async def async_request_openai_chat_completions(
"stream": not args.disable_stream, "stream": not args.disable_stream,
**request_func_input.extra_request_body, **request_func_input.extra_request_body,
} }
# hack to accommodate different LoRA conventions between SGLang and vLLM.
if request_func_input.lora_name:
payload["model"] = request_func_input.lora_name
payload["lora_path"] = request_func_input.lora_name
headers = get_auth_headers() headers = get_auth_headers()
output = RequestFuncOutput.init_new(request_func_input) output = RequestFuncOutput.init_new(request_func_input)