gpu环境并发请求：RuntimeError: could not execute a primitive

[June-Li]

🔎 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)

我用的fastapi（flaskapi也试过，问题一样）；
这是fastapi中的代码片段，如果保留async，他其实是单线程（不报错），因为你同时发两个请求，只有一个在处理并且会阻塞另一个请求，也就是另一个请求在排队，如果把async删掉的话，就可以多线程并发处理了，但是会报错，这个报错是从paddleocr传出来的。但是我们的服务是需要多线程的，不能阻塞在那；
@app.post(
"/pdf_parse",
tags=["projects"],
summary="Parse PDF files (supports local files and S3)",
)
async def pdf_parse(
pdf_file: UploadFile = None,
pdf_path: str = None,
parse_method: str = "auto",
is_json_md_dump: bool = False,
output_dir: str = "output",
return_layout: bool = False,
return_info: bool = False,
return_content_list: bool = False,
return_images: bool = False,
):

🏃‍♂️ Environment (运行环境)

How to reproduce the bug | 如何复现
flaskapi或者fastapi部署服务后，同时发两个请求就会出现以上报错

Operating system | 操作系统
Linux

Python version | Python 版本
3.10

Software version | 软件版本 (magic-pdf --version)
1.0.x

Device mode | 设备模式
cuda

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

fastapi部署服务后，发多次请求会报错，以下提供mineru的代码库，可以直接用mineru的fastapi服务（去掉async def pdf_parse()中的async可多线程并发）进行复现；
https://github.com/opendatalab/MinerU?tab=readme-ov-file
opendatalab/MinerU#1891

[GreatV]

可以试试升级一下paddlepaddle版本到3.0 rc1
Related issues:

• RuntimeError: could not execute a primitive #14103

[Luux]

I have this problem also when using in a threaded environment (even when doing everything related paddleocr initialization and inference with a threadlock!) and when using it in pytest sessions.

Tried with paddlepaddle 3.0 rc1.

It works with paddlepaddle-2.6.2, but I'm running into a memory leak there.

Kinda breaks PaddleOCR for me currently :/

[mauryaland]

Same as Luux here.

[Luux]

This is the memory leak I'm talking about: #7823

[Luux]

Oh and btw.: I'm using it on CPU. So the problem is also there.