高性能服务hps不兼容旧版本paddleocr-vl吗，能否兼容一下呢

[zhujlfree]

🔎 Search before asking

• I have searched the PaddleOCR Docs and found no similar bug report.
• I have searched the PaddleOCR Issues and found no similar bug report.
• I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

使用教程安装：https://github.com/PaddlePaddle/PaddleOCR/blob/main/deploy/paddleocr_vl_docker/hps/README.md

在paddleocr-vl-1.5中速度正常，每张图片2-3s，但是识别效果差，幻觉严重，所以想更换成paddleocr-vl，但是速度需要18-20s一张。

paddleocr-vl-api | 2026-05-15 02:31:16,529 - app - _process_triton_request - INFO - Processing 'layout-parsing' request 825f35fc-a4a9-4d09-904e-fb5b8f0a100a
paddleocr-vl-tritonserver | [ INFO] [2026-05-15 02:31:16,530] [1f6f47aa960244d983b9eb77d32118ec] [f9c9942a54c943ae93c7b1b4cdf6bb18] - Received batch of size 1
paddleocr-vl-tritonserver | [ INFO] [2026-05-15 02:31:16,530] [1f6f47aa960244d983b9eb77d32118ec] [f9c9942a54c943ae93c7b1b4cdf6bb18] - Request 825f35fc-a4a9-4d09-904e-fb5b8f0a100a received
paddleocr-vlm-server | (APIServer pid=1) INFO: 127.0.0.1:57150 - "GET /health HTTP/1.1" 200 OK
paddleocr-vlm-server | (APIServer pid=1) INFO: 127.0.0.1:50046 - "GET /health HTTP/1.1" 200 OK
paddleocr-vl-tritonserver | [ INFO] [2026-05-15 02:31:34,956] [1f6f47aa960244d983b9eb77d32118ec] [f9c9942a54c943ae93c7b1b4cdf6bb18] - Time taken: 18425.459 ms
paddleocr-vl-api | 2026-05-15 02:31:34,966 - app - _process_triton_request - INFO - Completed 'layout-parsing' request 825f35fc-a4a9-4d09-904e-fb5b8f0a100a: 18.437s
paddleocr-vl-api | INFO: 172.22.0.1:46014 - "POST /layout-parsing HTTP/1.1" 200 OK

paddleocr-vl使用方法：
1、下载https://paddle-model-ecology.bj.bcebos.com/paddlex/PaddleX3.0/deploy/paddlex_hps/public/sdks/v3.4/paddlex_hps_PaddleOCR-VL_sdk.tar.gz

2、将paddleocr-vl-tritonserver 和 paddleocr-vlm-server的模型替换成paddleocr-vl-0.9b，其他保持一致。

🏃‍♂️ Environment (运行环境)

NVIDIA RTX 4090
CUDA Version: 12.8

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

使用教程中的镜像

[zhujlfree]

已解决
将paddlex_hps_PaddleOCR-VL_sdk/server/pipeline_config.yaml中的
genai_config:
backend: native
改成vllm：
genai_config:
backend: vllm-server
server_url: http://paddleocr-vlm-server:8080/v1