PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/PP-StructureV3 #3651

0 paddle_infer::Predictor::Predictor(paddle::AnalysisConfig const&)
1 paddle::AnalysisPredictor::~AnalysisPredictor()
2 paddle::AnalysisPredictor::~AnalysisPredictor()
3 std::_Sp_counted_base<(__gnu_cxx::_Lock_policy)2>::_M_release()
4 pir::Program::~Program()
5 pir::ModuleOp::Destroy()
6 pir::Operation::Destroy()
7 pir::Region::~Region()
8 pir::Region::clear()
9 pir::Block::ClearOps()
10 pir::Block::pop_back()
11 pir::Operation::Destroy()

Error Message Summary:

FatalError: Process abort signal is detected by the operating system.
[TimeInfo: *** Aborted at 1743565700 (unix time) try "date -d @1743565700" if you are using GNU date ***]
[SignalInfo: *** SIGABRT (@0x3e800002867) received by PID 10343 (TID 0x7fded2e8d4c0) from PID 10343 ***]

Aborted (core dumped)

1 reply

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

最近么有做更新，不确定是不是aistudio某块的环境升级导致的，建议在本机试一下呢

code-frog-one · 2025-04-09T06:59:35Z

code-frog-one
Apr 9, 2025 — with giscus

1.请问这个是可以商用的吗？我看文档有的地方说需要购买，我点击进入又到了模型产线部署的页面，里面又说的需要获取序列号，但是序列号又是完全免费的。所以对于是否能免费使用和商用很混乱。
2.如果能私有化部署，可以完全离线部署吗？我在本地部署体验的时候很多依赖是联网下载的。

1 reply

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

paddlex的协议是Apache协议，完全可以商用。至于序列号，目前只针对高稳定性服务化部署需要，目的是为了监控哪些产线大家使用的比较多，我们好针对性优化。注意到这块确实容易造成理解上的混乱，所以下一个版本会下掉
如果希望完全离线部署，建议使用docker，相关的环境是ready的

sleepy-duo · 2025-04-11T01:42:22Z

sleepy-duo
Apr 11, 2025 — with giscus

使用paddlex --serve发布成服务后，看接口文档里说对于超过10页的PDF文件，只有前10页的内容会被使用。如何才能对整篇pdf文档整体解析？还是说要自己先对pdf文件做切分再去解析？

2 replies

Bobholamovic Apr 11, 2025 — with giscus
Maintainer

你好，目前有一个尚未在文档中体现的接口可以解除页数限制。具体而言，在产线配置文件中增加如下设置：

Serving:
    extra:
        max_num_input_imgs: null

Danee-wawawa Apr 23, 2025 — with giscus

你好，想问下以下两种方式速度上有区别吗？
1、将pdf整个传给服务，即"fileType"为0
2、先读取pdf，然后一页一页传给服务，即"fileType"为1
不确定内部实现方式，速度上1会更快吗？

Albus-Tim-Brown · 2025-04-15T01:52:10Z

Albus-Tim-Brown
Apr 15, 2025 — with giscus

最新PaddleX3.0.0rc版本，使用上面的示例代码报错：ValueError: (InvalidArgument) The size of inputs must be equal to 2.
[Hint: Expected input_size == 2, but received input_size:1 != 2:2.] (at paddle\fluid\pir\dialect\operator\ir\pd_op3.cc:13372)。无论换成图片还是pdf都一样

2 replies

Albus-Tim-Brown Apr 15, 2025 — with giscus

使用Python 3.12版本，CUDA12.6，PaddlePaddle为3.0.0版本，PaddleX插件都已经安装。在PaddleX的Github中有对应issues描述与我遇到的类似问题但未解决

Albus-Tim-Brown Apr 18, 2025 — with giscus

已将PaddlePaddle-gpu版本回退到3.0.0rc0版本，Python版本换成3.10，重新以插件模式安装PaddleX和执行paddlex --install后不再出现此类问题。暂且未知是什么原因，已在对应Github issues中回复了这个“临时解决方法”，希望继续跟进

hyp530 · 2025-04-24T08:39:10Z

hyp530
Apr 24, 2025 — with giscus

高性能推理没有支持PP-StructureV3，可否添加

我试用image_classification是能跑通的，用PP-StructureV3显示找不到模型（而PP-StructureV3不添加use_hpip也能跑通）。

0 replies

zyxzs · 2025-05-09T08:19:02Z

zyxzs
May 9, 2025 — with giscus

图片可以正常运行，PDF报错。
命令：
paddlex --pipeline PP-StructureV3
--input test.pdf
--use_doc_orientation_classify False
--use_doc_unwarping False
--use_textline_orientation False
--save_path ./output
--device cpu
错误：

C++ Traceback (most recent call last):

0 paddle::AnalysisPredictor::ZeroCopyRun(bool)
1 paddle::framework::NaiveExecutor::RunInterpreterCore(std::vector<std::string, std::allocator<std::string > > const&, bool, bool)
2 paddle::framework::InterpreterCore::Run(std::vector<std::string, std::allocator<std::string > > const&, bool, bool, bool, bool)
3 paddle::framework::PirInterpreter::Run(std::vector<std::string, std::allocator<std::string > > const&, bool, bool, bool, bool)
4 paddle::framework::PirInterpreter::TraceRunImpl()
5 paddle::framework::PirInterpreter::TraceRunInstructionList(std::vector<std::unique_ptr<paddle::framework::InstructionBase, std::default_deletepaddle::framework::InstructionBase >, std::allocator<std::unique_ptr<paddle::framework::InstructionBase, std::default_deletepaddle::framework::InstructionBase > > > const&)
6 paddle::framework::PirInterpreter::RunInstructionBase(paddle::framework::InstructionBase*)
7 paddle::framework::PhiKernelInstruction::Run()
8 phi::KernelImpl<void ()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, std::vector<int, std::allocator > const&, int, std::string const&, phi::DenseTensor), &(void phi::ConvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, std::vector<int, std::allocator > const&, int, std::string const&, phi::DenseTensor*))>::Compute(phi::KernelContext*)
9 void phi::ConvKernelImpl<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, int, std::vector<int, std::allocator > const&, std::string const&, phi::DenseTensor*)
10 phi::funcs::Im2ColFunctor<(phi::funcs::ColFormat)0, phi::CPUContext, float>::operator()(phi::CPUContext const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, phi::DenseTensor*, common::DataLayout)

Error Message Summary:

FatalError: Segmentation fault is detected by the operating system.
[TimeInfo: *** Aborted at 1746778530 (unix time) try "date -d @1746778530" if you are using GNU date ***]
[SignalInfo: *** SIGSEGV (@0x7f21788b9320) received by PID 3818 (TID 0x7f2763780740) from PID 2022413088 ***]

Segmentation fault (core dumped)

0 replies

zhangft7478191 · 2025-05-09T08:19:34Z

zhangft7478191
May 9, 2025

你的邮件我已收到，谢谢合作！

0 replies

Damoooon · 2025-05-12T07:36:39Z

Damoooon
May 12, 2025 — with giscus

您好，我的paddlepaddle=3.0.0, paddlex=3.0.0rc1,
我在运行PP-StructureV3 pipeline示例脚本时报错，请问是什么原因：

Traceback (most recent call last):
File "G:\PythonProject\molar\ocr_test\paddle_test.py", line 3, in
pipeline = create_pipeline(pipeline="G:\PythonProject\molar\ocr_test\PP-StructureV3.yaml")
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines_init_.py", line 161, in create_pipeline
pipeline = BasePipeline.get(pipeline_name)(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\utils\deps.py", line 195, in wrapper
return old_init_func(self, *args, **kwargs)
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines\layout_parsing\pipeline_v2.py", line 69, in init
self.inintial_predictor(config)
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines\layout_parsing\pipeline_v2.py", line 164, in inintial_predictor
self.formula_recognition_pipeline = self.create_pipeline(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines\base.py", line 132, in create_pipeline
pipeline = create_pipeline(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines_init.py", line 161, in create_pipeline
pipeline = BasePipeline.get(pipeline_name)(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\utils\deps.py", line 195, in wrapper
return old_init_func(self, *args, **kwargs)
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines\formula_recognition\pipeline.py", line 109, in init
self.formula_recognition_model = self.create_model(formula_recognition_config)
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\pipelines\base.py", line 99, in create_model
model = create_predictor(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models_init.py", line 77, in create_predictor
return BasePredictor.get(model_name)(
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models\formula_recognition\predictor.py", line 64, in init
self.pre_tfs, self.infer, self.post_op = self._build()
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models\formula_recognition\predictor.py", line 84, in _build
infer = self.create_static_infer()
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models\base\predictor\base_predictor.py", line 248, in create_static_infer
return PaddleInfer(self.model_dir, self.MODEL_FILE_PREFIX, self._pp_option)
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models\common\static_infer.py", line 319, in init
self.predictor = self._create()
File "F:\anaconda\setup\envs\paddle\lib\site-packages\paddlex\inference\models\common\static_infer.py", line 486, in _create
predictor = paddle.inference.create_predictor(config)
ValueError: (InvalidArgument) The size of inputs must be equal to 2.
[Hint: Expected input_size == 2, but received input_size:1 != 2:2.] (at paddle\fluid\pir\dialect\operator\ir\pd_op3.cc:13372)

0 replies

nioehwrs · 2025-06-19T02:42:59Z

nioehwrs
Jun 19, 2025 — with giscus

希望产线处理时，能在提取印章后，从图片中去除印章效果（比如按颜色去除）再执行其他文字识别。否则印章文字会和原本的文本产生识别冲突。

1 reply

nioehwrs Jun 19, 2025 — with giscus

如果能够根据印章坐标，在最终识别结果中补充回印章识别结果就更好了

zhangft7478191 · 2025-06-19T02:43:31Z

zhangft7478191
Jun 19, 2025

你的邮件我已收到，谢谢合作！

0 replies

wangxy-hub · 2025-06-25T10:37:09Z

wangxy-hub
Jun 25, 2025 — with giscus

在cpu版本的docker里边运行，推理样例都跑不通，一直提示

C++ Traceback (most recent call last):

0 paddle::AnalysisPredictor::ZeroCopyRun(bool)
1 paddle::framework::NaiveExecutor::RunInterpreterCore(std::vector<std::string, std::allocator<std::string > > const&, bool, bool)
2 paddle::framework::InterpreterCore::Run(std::vector<std::string, std::allocator<std::string > > const&, bool, bool, bool, bool)
3 paddle::framework::PirInterpreter::Run(std::vector<std::string, std::allocator<std::string > > const&, bool, bool, bool, bool)
4 paddle::framework::PirInterpreter::TraceRunImpl()
5 paddle::framework::PirInterpreter::TraceRunInstructionList(std::vector<std::unique_ptr<paddle::framework::InstructionBase, std::default_deletepaddle::framework::InstructionBase >, std::allocator<std::unique_ptr<paddle::framework::InstructionBase, std::default_deletepaddle::framework::InstructionBase > > > const&)
6 paddle::framework::PirInterpreter::RunInstructionBase(paddle::framework::InstructionBase*)
7 paddle::framework::PhiKernelInstruction::Run()
8 phi::KernelImpl<void ()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, std::vector<int, std::allocator > const&, int, std::string const&, phi::DenseTensor), &(void phi::ConvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, std::vector<int, std::allocator > const&, int, std::string const&, phi::DenseTensor*))>::Compute(phi::KernelContext*)
9 void phi::ConvKernelImpl<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::string const&, int, std::vector<int, std::allocator > const&, std::string const&, phi::DenseTensor*)
10 phi::funcs::Im2ColFunctor<(phi::funcs::ColFormat)0, phi::CPUContext, float>::operator()(phi::CPUContext const&, phi::DenseTensor const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, std::vector<int, std::allocator > const&, phi::DenseTensor*, common::DataLayout)

Error Message Summary:

FatalError: Segmentation fault is detected by the operating system.
[TimeInfo: *** Aborted at 1750847369 (unix time) try "date -d @1750847369" if you are using GNU date ***]
[SignalInfo: *** SIGSEGV (@0x7f40d6b77400) received by PID 8126 (TID 0x7f47546cb740) from PID 18446744073016931328 ***]

Segmentation fault (core dumped)

请问一下这个问题是内存不够导致的么？如果是的话产线能支持在cpu上最小的设备要求是什么？docker容器生成的时候需要增加什么设置么？

1 reply

Bobholamovic Jun 25, 2025
Maintainer

这是一个bug导致的，我们将在本周内发布的下个版本修复，抱歉给你带来了不便～

doggnol · 2025-06-25T14:29:08Z

doggnol
Jun 25, 2025 — with giscus

您好，请问版面区域检测模型可以支持单独部署吗？目前文档的说明是版面区域检测和OCR子产线是必选的。

4 replies

Bobholamovic Jun 26, 2025
Maintainer

可以部署版面区域检测产线～

doggnol Jun 26, 2025 — with giscus

您是指layout_parsing.yaml中所定义的产线吗？

Bobholamovic Jun 26, 2025
Maintainer

不好意思，我看错了，这个模块应该没有单独的产线。不过，你可以选择部署layout detection模块～

doggnol Jul 3, 2025 — with giscus

好的，如果能官方支持单独的版面区域检测产线就更好了。
有些特殊项目可以依赖开源定义的接口，但是不能依赖自己定义的接口，所以比较尴尬。
如果需要开发者参与，我可以用业余时间来提交一些PR。

doggnol · 2025-07-03T13:08:38Z

doggnol
Jul 3, 2025

您好，我成功部署了PP-structureV3，关于版面检测模型PP-DocLayout_plus-L有两点疑问，请指教：

algorithm类型我没测试出来是什么，我以为是代码块，但从检测结果看起来不是；单模块说明中介绍这个类型是“算法”，不太容易理解实际的文档元素是什么。
figure_table_chart_title在配置文件中显示是其中一个类型，但我测试到返回的类型是单独的chart_title，而不是类型列表里面的figure_table_chart_title，这里是有做什么优化处理么？

0 replies

F-people · 2025-07-14T03:36:43Z

F-people
Jul 14, 2025 — with giscus

您好，直接调用产线模型PPStructureV3( ) 最后识别结果包含后处理 Reading Order Recovering以及Layout Recovering，单独使用PP-DocLayout_plus-L模型并没有，请问阅读顺序有单独模型么

3 replies

Bobholamovic Jul 14, 2025
Maintainer

阅读顺序目前并不是通过模型实现的，而是基于规则的后处理。目前只支持在PPStructureV3产线中获得阅读顺序结果。

F-people Jul 14, 2025 — with giscus

好的，感谢回复

F-people Jul 28, 2025 — with giscus

您好，PPStructureV3产线我看很多检测模块可以关闭，但是没有找到ocr相关模块，是因为后处理版面还原
以及阅读顺序需要用到ocr识别结果么

Echo-Forever · 2025-07-15T09:39:24Z

Echo-Forever
Jul 15, 2025 — with giscus

大文件解析 visualize 设置为false ，解析返回的数据也造成缓冲区溢出，有什么参数可以控制只返回markdown字段吗

1 reply

Bobholamovic Jul 15, 2025
Maintainer

这个目前没有可以考虑在本地对文件分块后处理或者改造服务

hyp530 · 2025-07-17T04:38:56Z

hyp530
Jul 17, 2025

@Bobholamovic 请教一下：我运行pp_structure_v3，显存占用通常在9-10G左右，可以长期稳定运行。

但我想把24G的显存用尽，同时运行两个pipeline，则运行一段时间后，就会报错退出，看上去是GPU的某种冲突，但指向性不明。

请问paddlex支持同一个GPU下运行多个pipeline，或有什么方式充分利用显存吗？多谢。

17 replies

Bobholamovic Jul 28, 2025
Maintainer

请问你设置的是哪里的batch size呀？

hyp530 Jul 28, 2025

请问你设置的是哪里的batch size呀？

在产线配置文件PP-StructureV3.yaml中，见以下文件的第4行：batch_size: 8

https://github.com/PaddlePaddle/PaddleX/blob/a8ec0bdec40d4108859ebda48e079d2fdcfb5b82/paddlex/configs/pipelines/PP-StructureV3.yaml

Bobholamovic Jul 28, 2025
Maintainer

那么在产线并行调用的脚本中，是否也实际给每个instance分配了等于或大于batch大小的数据呢？

hyp530 Jul 29, 2025

那么在产线并行调用的脚本中，是否也实际给每个instance分配了等于或大于batch大小的数据呢？

我理解是有的，我采用的是这个示例脚本，input_dir目录中有上千个pdf文件【只是将create_pipeline改成yaml配置文件】https://paddlepaddle.github.io/PaddleX/latest/pipeline_usage/instructions/parallel_inference.html#_3

Bobholamovic Jul 29, 2025
Maintainer

由于产线比较复杂，所以“batch size”不止一处：配置文件中的batch_size决定产线并行处理的样本数量；示例脚本中还有一个--batch_size参数，会决定实际一次分配给产线的样本数量；此外，配置文件中各个模型的batch size会决定该模型应该同时处理多少个样本。

fennuzhicui · 2025-08-13T10:46:52Z

fennuzhicui
Aug 13, 2025 — with giscus

使用NVIDIA 4090显卡推理，一次大概要6s，这正常吗，而且把图像矫正啥的都关了。
推理参数：
pipeline.predict(
input=image,
use_doc_orientation_classify=False,
use_doc_unwarping=False,
use_textline_orientation=False,
use_seal_recognition=False,
use_table_recognition=False,
use_formula_recognition=False,
use_chart_recognition=False,
return_visualization=False,
)

0 replies

zhangft7478191 · 2025-08-13T10:47:31Z

zhangft7478191
Aug 13, 2025

你的邮件我已收到，谢谢合作！

0 replies

wangqifenggithub · 2025-09-02T14:16:22Z

wangqifenggithub
Sep 2, 2025 — with giscus

我调用这个接口：http://172.16.49.10:8000/v2/models/layout-parsing//infer
HTTP 状态码是400，这是为什么呢

0 replies

zhangft7478191 · 2025-09-02T14:16:55Z

zhangft7478191
Sep 2, 2025

你的邮件我已收到，谢谢合作！

0 replies

liuhannnnnnnnn · 2025-09-04T02:13:17Z

liuhannnnnnnnn
Sep 4, 2025 — with giscus

predict() 方法的参数及其说明中，参数名称错误。

1 reply

liuhannnnnnnnn Sep 4, 2025 — with giscus

use_wired_table_cells_trans_to_html 是否启用有线表单元格检测结果直转HTML，默认False，启用则直接基于有线表单元格检测结果的几何关系构建HTML。 float|None
bool：True 或者 False；
None：如果设置为None, 将默认使用产线初始化的该参数值，初始化为False；
False
use_wired_table_cells_trans_to_html 是否启用无有线表单元格检测结果直转HTML，默认False，启用则直接基于有线表单元格检测结果的几何关系构建HTML。 float|None
bool：True 或者 False；
None：如果设置为None, 将默认使用产线初始化的该参数值，初始化为False；
False。
这两个参数名称重复了

liuhannnnnnnnn · 2025-09-04T02:13:27Z

liuhannnnnnnnn
Sep 4, 2025 — with giscus

use_wired_table_cells_trans_to_html 是否启用有线表单元格检测结果直转HTML，默认False，启用则直接基于有线表单元格检测结果的几何关系构建HTML。 float|None
bool：True 或者 False；
None：如果设置为None, 将默认使用产线初始化的该参数值，初始化为False；
False
use_wired_table_cells_trans_to_html 是否启用无有线表单元格检测结果直转HTML，默认False，启用则直接基于有线表单元格检测结果的几何关系构建HTML。 float|None
bool：True 或者 False；
None：如果设置为None, 将默认使用产线初始化的该参数值，初始化为False；
False

0 replies

PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/PP-StructureV3 #3651

Uh oh!

giscus[bot] bot Mar 19, 2025

PaddleX/latest/pipeline_usage/tutorials/ocr_pipelines/PP-StructureV3

Replies: 26 comments · 40 replies

Uh oh!

yuanmengwei Mar 19, 2025 — with giscus

Uh oh!

cuicheng01 Mar 22, 2025 Maintainer

Uh oh!

yuanmengwei Mar 24, 2025 — with giscus

Uh oh!

zhangft7478191 Mar 20, 2025 — with giscus

Uh oh!

zzzcccxx Mar 21, 2025 — with giscus

Uh oh!

cuicheng01 Mar 22, 2025 Maintainer

Uh oh!

yuanmengwei Mar 24, 2025 — with giscus

Uh oh!

cuicheng01 Apr 11, 2025 — with giscus Maintainer

Uh oh!

skiingwang Apr 2, 2025 — with giscus

Uh oh!

cuicheng01 Apr 11, 2025 — with giscus Maintainer

Uh oh!

fayxxxx Apr 2, 2025 — with giscus

Destroyed a op_result that is still in use by pd_op.while. The owner op type is: pd_op.full (at /paddle/paddle/pir/src/core/op_result_impl.cc:37)

C++ Traceback (most recent call last):

Error Message Summary:

Uh oh!

cuicheng01 Apr 11, 2025 — with giscus Maintainer

Uh oh!

code-frog-one Apr 9, 2025 — with giscus

Uh oh!

cuicheng01 Apr 11, 2025 — with giscus Maintainer

Uh oh!

sleepy-duo Apr 11, 2025 — with giscus

Uh oh!

Bobholamovic Apr 11, 2025 — with giscus Maintainer

Uh oh!

Danee-wawawa Apr 23, 2025 — with giscus

Uh oh!

Albus-Tim-Brown Apr 15, 2025 — with giscus

Uh oh!

Albus-Tim-Brown Apr 15, 2025 — with giscus

Uh oh!

Albus-Tim-Brown Apr 18, 2025 — with giscus

Uh oh!

hyp530 Apr 24, 2025 — with giscus

Uh oh!

zyxzs May 9, 2025 — with giscus

图片可以正常运行，PDF报错。 命令： paddlex --pipeline PP-StructureV3 --input test.pdf --use_doc_orientation_classify False --use_doc_unwarping False --use_textline_orientation False --save_path ./output --device cpu 错误：

C++ Traceback (most recent call last):

Error Message Summary:

Uh oh!

zhangft7478191 May 9, 2025

Uh oh!

Damoooon May 12, 2025 — with giscus

Uh oh!

nioehwrs Jun 19, 2025 — with giscus

Uh oh!

nioehwrs Jun 19, 2025 — with giscus

Uh oh!

zhangft7478191 Jun 19, 2025

Uh oh!

wangxy-hub Jun 25, 2025 — with giscus

在cpu版本的docker里边运行，推理样例都跑不通，一直提示

C++ Traceback (most recent call last):

Error Message Summary:

Uh oh!

Bobholamovic Jun 25, 2025 Maintainer

Uh oh!

doggnol Jun 25, 2025 — with giscus

Uh oh!

Bobholamovic Jun 26, 2025 Maintainer

Uh oh!

doggnol Jun 26, 2025 — with giscus

giscus[bot]
bot Mar 19, 2025

Replies: 26 comments 40 replies

yuanmengwei
Mar 19, 2025 — with giscus

cuicheng01 Mar 22, 2025
Maintainer

zhangft7478191
Mar 20, 2025 — with giscus

cuicheng01 Mar 22, 2025
Maintainer

yuanmengwei
Mar 24, 2025 — with giscus

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

skiingwang
Apr 2, 2025 — with giscus

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

fayxxxx
Apr 2, 2025 — with giscus

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

code-frog-one
Apr 9, 2025 — with giscus

cuicheng01 Apr 11, 2025 — with giscus
Maintainer

sleepy-duo
Apr 11, 2025 — with giscus

Bobholamovic Apr 11, 2025 — with giscus
Maintainer

Albus-Tim-Brown
Apr 15, 2025 — with giscus

hyp530
Apr 24, 2025 — with giscus

zyxzs
May 9, 2025 — with giscus

图片可以正常运行，PDF报错。
命令：
paddlex --pipeline PP-StructureV3
--input test.pdf
--use_doc_orientation_classify False
--use_doc_unwarping False
--use_textline_orientation False
--save_path ./output
--device cpu
错误：

zhangft7478191
May 9, 2025

Damoooon
May 12, 2025 — with giscus

nioehwrs
Jun 19, 2025 — with giscus

zhangft7478191
Jun 19, 2025

wangxy-hub
Jun 25, 2025 — with giscus

Bobholamovic Jun 25, 2025
Maintainer

doggnol
Jun 25, 2025 — with giscus

Bobholamovic Jun 26, 2025
Maintainer