xinyun/vllm
1
0
Fork 0
mirror of https://git.datalinker.icu/vllm-project/vllm.git synced 2026-05-24 08:31:18 +08:00
Code Issues Packages Projects Releases Wiki Activity

[Benchmark] Use truncation by default for pooling benchmarks (#26992)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung 2025-10-16 16:02:39 +08:00 committed by GitHub
parent 44c8555621
commit 17838e50ef
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 9 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
vllm/benchmarks/lib/endpoint_request_func.py
View File

@ -527,6 +527,9 @@ async def async_request_openai_embeddings(
if request_func_input.model_name if request_func_input.model_name
else request_func_input.model, else request_func_input.model,
"input": request_func_input.prompt, "input": request_func_input.prompt,
# Many embedding models have short context length,
# this is to avoid dropping some of the requests.
"truncate_prompt_tokens": -1,
} }
_update_payload_common(payload, request_func_input) _update_payload_common(payload, request_func_input)
@ -564,6 +567,9 @@ async def async_request_vllm_rerank(
else request_func_input.model, else request_func_input.model,
"query": request_func_input.prompt[0], "query": request_func_input.prompt[0],
"documents": request_func_input.prompt[1:], "documents": request_func_input.prompt[1:],
# Many reranker models have short context length,
# this is to avoid dropping some of the requests.
"truncate_prompt_tokens": -1,
} }
headers = { headers = {
@ -599,6 +605,9 @@ async def async_request_openai_embeddings_chat(
"messages": [ "messages": [
{"role": "user", "content": content}, {"role": "user", "content": content},
], ],
# Many embedding models have short context length,
# this is to avoid dropping some of the requests.
"truncate_prompt_tokens": -1,
} }
_update_payload_common(payload, request_func_input) _update_payload_common(payload, request_func_input)
@ -634,13 +643,6 @@ def _preprocess_clip(request_func_input: RequestFuncInput):
# Image input # Image input
request_func_input.prompt = "" request_func_input.prompt = ""
# max_model_len=77 is too short for most datasets,
# so by default we truncate the prompt to max_model_len
if request_func_input.extra_body is None:
request_func_input.extra_body = {}
if "truncate_prompt_tokens" not in request_func_input.extra_body:
request_func_input.extra_body["truncate_prompt_tokens"] = -1
def _preprocess_vlm2vec(request_func_input: RequestFuncInput): def _preprocess_vlm2vec(request_func_input: RequestFuncInput):
if request_func_input.multi_modal_content: if request_func_input.multi_modal_content: