PaddleOCR VL识别多页pdf文件会卡住

[hhxxttxslcc]

您好，请问一下，在识别短文档的时候能够正常运行出结果，在识别长文档时，会一直卡在如图所示的位置，并且显存也是一直占用着的，但是gpu利用率为0，这是为什么呢？我的显卡是5070

[aniruddhaadak80]

It seems like you are asking about PDF processing with the VLM path and possibly how to scale it across GPUs, but the current text is too garbled to tell whether this is a performance question or a bug report. Those are very different debugging paths.

If you can restate the issue with a short reproduction example, the number of pages, your GPU setup, and what part is slow or failing, the discussion becomes much easier to answer. At the moment the most important missing piece is the exact bottleneck you are hitting.