请教一下，怎么使用vllm serve部署rerank模型， 然后实现接口调用

[lilikkk10908]

[RC-Qiao]

我用接口输出的图像重排结果0.5403098464012146,这是为什么

[lilikkk10908]

的图像重排
我可以用vllm serve启动服务，但是请求的时候应该怎么请求？documents中是图文
vllm启动是成功的：
(APIServer pid=27) INFO 02-05 07:32:14 [api_server.py:1014] Supported tasks: ['token_classify', 'classify', 'score']
...
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/messages, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/chat/completions, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/completions, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/audio/transcriptions, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/audio/translations, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /ping, Methods: GET
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /ping, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /invocations, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /classify, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/embeddings, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /score, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/score, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /rerank, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v1/rerank, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /v2/rerank, Methods: POST
(APIServer pid=27) INFO 02-05 07:32:14 [launcher.py:46] Route: /pooling, Methods: POST
(APIServer pid=27) INFO: Started server process [27]
(APIServer pid=27) INFO: Waiting for application startup.
(APIServer pid=27) INFO: Application startup complete.

[RC-Qiao]

https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/vision_score_api_online.py看这个， vllm serve Qwen/Qwen3-VL-Reranker-2B
--runner pooling
--max-model-len 4096
--hf_overrides '{"architectures": ["Qwen3VLForSequenceClassification"],"classifier_from_token": ["no", "yes"],"is_original_qwen3_reranker": true}'
--chat-template examples/pooling/score/template/qwen3_vl_reranker.jinja可以请求图像，但是图像重排得分都是0.5几

[lilikkk10908]

https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/vision_score_api_online.py看这个， vllm serve Qwen/Qwen3-VL-Reranker-2B --runner pooling --max-model-len 4096 --hf_overrides '{"architectures": ["Qwen3VLForSequenceClassification"],"classifier_from_token": ["no", "yes"],"is_original_qwen3_reranker": true}' --chat-template examples/pooling/score/template/qwen3_vl_reranker.jinja可以请求图像，但是图像重排得分都是0.5几
https://github.com/vllm-project/vllm/blob/main/examples/pooling/score/vision_score_api_online.py
可以看一下你的请求体和请求方式吗？我试过这个可是并不对
print("Query: string & Document: list") prompt = { "model": model, "queries": query, "documents": [ document, {"content": [documents[0]]}, {"content": [documents[1]]}, {"content": [documents[0], documents[1]]}, ], } response = requests.post(score_url, json=prompt) pprint.pprint(response.json())