The newest paddleocr 3.1.1 is not compatible with paddlepaddle-gpu? TypeError: PaddlePredictorOption.__init__() takes 1 positional argument but 2 were given

[YIXIN-YAO]

🔎 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)

Got TypeError: PaddlePredictorOption.__init__() takes 1 positional argument but 2 were given when init PaddleOCR.

Tried paddleocr==3.1.1 and paddleocr==3.1.0, they all have this issue.
The latest working version is paddleocr==3.0.1

🏃‍♂️ Environment (运行环境)

Using this image on windows: paddlepaddle/paddle:3.1.0-gpu-cuda11.8-cudnn8.9

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

Launch container

• docker run -it --rm --gpus all paddlepaddle/paddle:3.1.0-gpu-cuda11.8-cudnn8.9 bash

Run all the commands below in container:

• pip install paddleocr==3.1.1

For some reason these opencv related libs are missing, need to install them before import paddleocr

apt-get update
apt-get install -y libgl1 libglib2.0-0

Now launch python in terminal:

• python

• from paddleocr import PaddleOCR

• ocr = PaddleOCR(use_doc_orientation_classify=False, use_doc_unwarping=False, use_textline_orientation=False)

Then error occurs:

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.10/dist-packages/paddleocr/_pipelines/ocr.py", line 161, in __init__
    super().__init__(**base_params)
  File "/usr/local/lib/python3.10/dist-packages/paddleocr/_pipelines/base.py", line 66, in __init__
    self.paddlex_pipeline = self._create_paddlex_pipeline()
  File "/usr/local/lib/python3.10/dist-packages/paddleocr/_pipelines/base.py", line 99, in _create_paddlex_pipeline
    kwargs = prepare_common_init_args(None, self._common_args)
  File "/usr/local/lib/python3.10/dist-packages/paddleocr/_common_args.py", line 75, in prepare_common_init_args
    pp_option = PaddlePredictorOption(
TypeError: PaddlePredictorOption.__init__() takes 1 positional argument but 2 were given

[SuiyueYoung]

解决方法：把 paddlex 改回 3.1 版本。我的 paddleocr==3.1.1, paddlepaddle-gpu==3.0.0, paddlex=3.1.0 是没问题的。#16286

[YIXIN-YAO]

解决方法：把 paddlex 改回 3.2 版本。#16286

改回3.1？

[bigminions]

Add paddlex==3.1.3 to requirement.txt

[Bobholamovic]

Sorry, this is because PaddleX 3.2 introduced an interface change, while the metadata of PaddleOCR 3.1 did not impose an upper bound on the required PaddleX version. For now, please consider specifying the PaddleX version <3.2 when installing, for example:

pip install paddleocr "paddlex<3.2"

We will release PaddleOCR 3.2 soon (which will be compatible with PaddleX 3.2) and improve the metadata of the Python package.

[annjawn]

Tried this

paddlex<3.2
paddlepaddle-gpu==3.0.0
paddleocr==3.1.1

Didn't work. Got the same error. Also tried this -

paddleocr==3.1.1 
paddlepaddle-gpu==3.0.0
paddlex=3.1.0

Still the same error. I am having to use cuda 11.8.0, with extra-index in requirements.txt as below-

--extra-index-url https://www.paddlepaddle.org.cn/packages/stable/cu118/

[alichohann]

Same error, did anyone find any solution for this?

[YIXIN-YAO]

This dockerfile worked for me:

FROM paddlepaddle/paddle:3.1.0-gpu-cuda11.8-cudnn8.9

RUN apt-get update && \
    apt-get install -y libgl1 libglib2.0-0

# Downgrade paddlex to 3.1.0 to avoid the bug
RUN pip install paddleocr==3.1.1 paddlex==3.1.0

[annjawn]

Unfortunately we cannot use the Paddle container.

@alichohann none of the solutions listed here or in the other thread worked for me. It’s seems it’s officially broken.

@YIXIN-YAO what version of paddlepaddle-gpu should we try?

[annjawn]

ok i finally made the following work

paddlex<3.2
paddlepaddle-gpu==3.1.0
paddleocr==3.1.0

But here's the catch, the *_model_name parameters in PaddleOCR() seems to be mandatory now because now paddlex checks it with this assertion.

https://github.com/PaddlePaddle/PaddleX/blob/0d042824651819b8a1c5557354440b0ab7a6624e/paddlex/inference/models/__init__.py#L74

So i had to list all the model names in PaddleOCR() initialization for it to work because I am loading the models from local dir. Previously, (2 days ago), the names were not required and it worked just fine. I am unclear if this assertion is intentional or a leftover test artifact.

[alichohann]

@annjawn can you share the code that worked for you?

[YIXIN-YAO]

This is my working env:
In this container: #16287 (comment)

[annjawn]

@alichohann . Note, now there is a new assert in paddlex that checks explicitly for the *_model_name parameters and these are the values.

from paddleocr import PaddleOCR

_ocr = PaddleOCR(
    use_doc_orientation_classify=True,
    use_doc_unwarping=False,
    use_textline_orientation=True,
    text_detection_model_name='PP-OCRv5_server_det',
    text_detection_model_dir="/opt/paddle_models/PP-OCRv5_server_det_infer",  # if loading from local dir
    text_recognition_model_name="PP-OCRv5_server_rec",
    text_recognition_model_dir="/opt/paddle_models/PP-OCRv5_server_rec_infer",   # if loading from local dir
    textline_orientation_model_name="PP-LCNet_x0_25_textline_ori",
    textline_orientation_model_dir="/opt/paddle_models/PP-LCNet_x0_25_textline_ori_infer",  # if loading from local dir
    doc_orientation_classify_model_name="PP-LCNet_x1_0_doc_ori",
    doc_orientation_classify_model_dir="/opt/paddle_models/PP-LCNet_x1_0_doc_ori_infer" # if loading from local dir
)

[cuicheng01]

Sorry for the inconvenience caused by this bug. If the versions of paddleocr and paddlex are compatible, you can directly install paddleocr version 3.2.

[TingquanGao]

The issue has no response for a long time and will be closed. You can reopen or new another issue if are still confused.

---

From Bot