PaddleOCR 3.0.2 fails with paddlepaddle-gpu==2.6.2 due to missing set_optimization_level attribute

[Youssef-SH]

🔎 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)

Trying to use PaddleOCR 3.0.2 with GPU on Colab (Tesla T4), but I get this error:

AttributeError: 'paddle.base.libpaddle.AnalysisConfig' object has no attribute 'set_optimization_level'

Seems like an incompatibility between paddleocr and paddlepaddle-gpu.

I just wanted to know:
– Is paddlepaddle-gpu still supported?
– What version combo is recommended to run PaddleOCR 3.0.2 on GPU without downgrading PaddleOCR?

I'm still learning, so any help would be greatly appreciated. Thanks in advance.

🏃‍♂️ Environment (运行环境)

paddleocr: 3.0.2
paddlepaddle-gpu: 2.6.2
Python: 3.11 (Colab)
CUDA: Tesla T4
OS: Google Colab / Ubuntu 22.04
Fails on: GPU inference only
Works fine on: CPU inference

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

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    text_detection_model_name='PP-OCRv5_server_det',
    text_recognition_model_name='arabic_PP-OCRv3_mobile_rec',
    device='gpu'
)

ocr.predict('/content/test.png')

[BVishal-Geek]

@Youssef-SH Any luck on solving the error ?

[pmbhatiya97]

Hello team facing same issue on paddlepaddle-gpu == 2.6.2 version

Attribute error : paddle.base.libpaddle.Analysisconfig' object has no attribute 'set_optimization_level' : did you mean: 'tensorrt_optimization_level' ?

CUDA 117 installed
NVIDIA RTX 4000 ADA GPU used

can anyone help me on solving this

same code paddlepaddle cpu version working fine but it is taking 25 mins to do OCR on single page

any lead appriciated.

[pmbhatiya97]

@Youssef-SH Any luck on solving the error ?

🔎 Search before asking

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

🐛 Bug (问题描述)

Trying to use PaddleOCR 3.0.2 with GPU on Colab (Tesla T4), but I get this error:

AttributeError: 'paddle.base.libpaddle.AnalysisConfig' object has no attribute 'set_optimization_level'

Seems like an incompatibility between paddleocr and paddlepaddle-gpu.

I just wanted to know: – Is paddlepaddle-gpu still supported? – What version combo is recommended to run PaddleOCR 3.0.2 on GPU without downgrading PaddleOCR?

I'm still learning, so any help would be greatly appreciated. Thanks in advance.

🏃‍♂️ Environment (运行环境)

paddleocr: 3.0.2
paddlepaddle-gpu: 2.6.2
Python: 3.11 (Colab)
CUDA: Tesla T4
OS: Google Colab / Ubuntu 22.04
Fails on: GPU inference only
Works fine on: CPU inference

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

from paddleocr import PaddleOCR

ocr = PaddleOCR(
    text_detection_model_name='PP-OCRv5_server_det',
    text_recognition_model_name='arabic_PP-OCRv3_mobile_rec',
    device='gpu'
)

ocr.predict('/content/test.png')

did you found any solution ??

[Youssef-SH]

The issue was about CUDA incompatibility. I found the latest version of paddlepaddle-gpu in the official paddlepaddle website quick install https://www.paddlepaddle.org.cn/en. paddlepaddle-gpu 3.0.0 supports either CUDA 12.6 or 11.8. So I had to get a VM with CUDA 12.6 and everything worked perfectly with the latest version of paddleOCR==3.0.2.

[tvone]

Vấn đề là về sự không tương thích của CUDA. Tôi đã tìm thấy phiên bản mới nhất của paddlepaddle-gpu trong trang web chính thức của paddlepaddle cài đặt nhanh https://www.paddlepaddle.org.cn/en. paddlepaddle-gpu 3.0.0 hỗ trợ CUDA 12.6 hoặc 11.8. Vì vậy, tôi phải có một VM với CUDA 12.6 và mọi thứ hoạt động hoàn hảo với phiên bản mới nhất của paddleOCR==3.0.2.

google colab has cuda currently 12.4, which version should i use? Or is it mandatory to use exact version 12.6 or 11.8, thank you!

[Hemilibeatriz]

Vấn đề là về sự không tương thích của CUDA. Tôi đã tìm thấy phiên bản mới nhất của paddlepaddle-gpu trong trang web chính thức của paddlepaddle cài đặt nhanh https://www.paddlepaddle.org.cn/en. paddlepaddle-gpu 3.0.0 hỗ trợ CUDA 12.6 hoặc 11.8. Vì vậy, tôi phải có một VM với CUDA 12.6 và mọi thứ hoạt động hoàn hảo với phiên bản mới nhất của paddleOCR==3.0.2.

google colab has cuda currently 12.4, which version should i use? Or is it mandatory to use exact version 12.6 or 11.8, thank you!

Hi @tvone Did you get it to work?

[wooseok-AI]

I had a same issue,
python -m pip install paddlepaddle-gpu==3.2.0 -i https://www.paddlepaddle.org.cn/packages/stable/cu126/

this helped me to re install which is compatible with my environment.

https://www.paddleocr.ai/latest/en/version3.x/installation.html

As I had kind of hard time fixing it. for whom would may have same problem, please try this.