NotImplementedError for ConvertPirAttribute2RuntimeAttribute crash during Step 2

[pale-blue-dot-matrix]

Checklist

• I have searched existing issues to avoid duplicates
• I am using the latest version of VideOCR

Describe the bug

videocr-cli-GPU-v1.4.1-CUDA-12.9 version crashes once it reaches Step 2. The paddleocr_error.log contains the following error: "NotImplementedError: (Unimplemented) ConvertPirAttribute2RuntimeAttribute not support"

Possibly related to this Paddle bug: PaddlePaddle/Paddle#77340

Steps to reproduce

1. Invoke CLI version of VideOCR via C# windows console app:

Process p = new Process();
p.StartInfo.UseShellExecute = false;
p.StartInfo.RedirectStandardOutput = true;
p.StartInfo.FileName = videoOcrPath;
p.StartInfo.Arguments = string.Format("--video_path "{0}" --output "{1}" --lang en --use_fullframe true", inputVideoFileName, outputSrtName);
p.Start();

2. Application crashes on step 2
3. NotImplementedError appears in paddleocr_error.log

Expected behavior

Application does not crash during Step 2

Screenshots / Logs

[2026-03-04 22:06:16] PaddleOCR process failed with exit code 1.
Command: C:\TESTAPP\videocr-cli-GPU-v1.4.1-CUDA-12.9\PaddleOCR-GPU-v1.4.0-CUDA-12.9\paddleocr.exe ocr --input C:\Users\user\AppData\Local\Temp\videocr_temp_108916_36bcth49 --device cpu --use_textline_orientation false --use_doc_orientation_classify False --use_doc_unwarping False --lang en --text_detection_model_dir C:\TESTAPP\videocr-cli-GPU-v1.4.1-CUDA-12.9\PaddleOCR.PP-OCRv5.support.files\det\PP-OCRv5_mobile_det --text_detection_model_name PP-OCRv5_mobile_det --text_recognition_model_dir C:\TESTAPP\videocr-cli-GPU-v1.4.1-CUDA-12.9\PaddleOCR.PP-OCRv5.support.files\rec\en_PP-OCRv5_mobile_rec --text_recognition_model_name en_PP-OCRv5_mobile_rec

--- STDOUT ---

--- STDERR ---
�[33mConnectivity check to the model hoster has been skipped because PADDLE_PDX_DISABLE_MODEL_SOURCE_CHECK is enabled.�[0m
C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_utils\cli.py:98: UserWarning: lang and ocr_version will be ignored when model names or model directories are not None.
�[32mCreating model: ('PP-OCRv5_mobile_det', 'C:\TESTAPP\videocr-cli-GPU-v1.4.1-CUDA-12.9\PaddleOCR.PP-OCRv5.support.files\det\PP-OCRv5_mobile_det')�[0m
WARNING: Logging before InitGoogleLogging() is written to STDERR
I0304 22:06:14.442039 12384 onednn_context.cc:81] oneDNN v3.6.2
�[32mCreating model: ('en_PP-OCRv5_mobile_rec', 'C:\TESTAPP\videocr-cli-GPU-v1.4.1-CUDA-12.9\PaddleOCR.PP-OCRv5.support.files\rec\en_PP-OCRv5_mobile_rec')�[0m
Traceback (most recent call last):
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\wrapper.py", line 41, in
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_main_.py", line 26, in console_entry
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_cli.py", line 194, in main
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_cli.py", line 183, in _execute
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_pipelines\ocr.py", line 693, in execute_with_args
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddleocr_utils\cli.py", line 114, in perform_simple_inference_ocr
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\pipelines_parallel.py", line 139, in predict
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\pipelines\ocr\pipeline.py", line 357, in predict
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\models\base\predictor\base_predictor.py", line 281, in call
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\models\base\predictor\base_predictor.py", line 338, in apply
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\models\text_detection\predictor.py", line 133, in process
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\models\common\static_infer.py", line 298, in call
File "C:\TESTAPP\VIDEOC1.9\PADDLE1.9\paddlex\inference\models\common\static_infer.py", line 261, in call
NotImplementedError: (Unimplemented) ConvertPirAttribute2RuntimeAttribute not support [pir::ArrayAttributepir::DoubleAttribute] (at ..\paddle\fluid\framework\new_executor\instruction\onednn\onednn_instruction.cc:118)

🖥 Environment

OS: Windows 11
VideOCR Version: [1.4.1]
GPU: NVIDIA GeForce RTX 5090

[timminator]

Yes, thats because of that PaddleBug, thats why I compiled the CPU version with Paddle version 3.2.2 instead if you want to check my release notes:
https://github.com/timminator/PaddleOCR-Standalone/releases/tag/v1.4.0

As you have a GPU and downloaded the GPU version you should add the --use-gpu flag to your command and set it to true. That will use you GPU and resolve the bug.