Predict Table Throwing runtime error #14605

mridulbirla · 2025-01-23T22:20:40Z

mridulbirla
Jan 23, 2025

🔎 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)

Unable to predict_table with en models

python table/predict_table.py --det_model_dir=inference/en_PP-OCRv3_det_infer --rec_model_dir=inference/en_PP-OCRv3_rec_infer --table_model_dir=inference/en_ppstructure_mobile_v2.0_SLANet_infer --rec_char_dict_path=../ppocr/utils/dict/en_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --image_dir=/home/mbirla/table_0.jpg     --output=../output/table1
[2025/01/23 14:15:25] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0
E0123 14:15:25.063272 21845 analysis_config.cc:125] Please use PaddlePaddle with GPU version.
[2025/01/23 14:15:25] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0
E0123 14:15:25.478474 21845 analysis_config.cc:125] Please use PaddlePaddle with GPU version.
[2025/01/23 14:15:25] ppocr WARNING: The first GPU is used for inference by default, GPU ID: 0
E0123 14:15:25.713666 21845 analysis_config.cc:125] Please use PaddlePaddle with GPU version.
Traceback (most recent call last):
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 241, in <module>
    main(args)
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 161, in main
    table_sys = TableSystem(args)
                ^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 79, in __init__
    self.table_structurer = predict_strture.TableStructurer(args)
                            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_structure.py", line 97, in __init__
    ) = utility.create_predictor(args, "table", logger)
        ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/tools/infer/utility.py", line 364, in create_predictor
    predictor = inference.create_predictor(config)
                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
RuntimeError: (PreconditionNotMet) Tensor's dimension is out of bound.Tensor's dimension must be equal or less than the size of its memory.But received Tensor's dimension is 8, memory's size is 0.
  [Hint: Expected numel() * SizeOf(dtype()) <= memory_size(), but received numel() * SizeOf(dtype()):8 > memory_size():0.] (at /paddle/paddle/phi/core/dense_tensor_impl.cc:55)
  [operator < scale > error]

🏃‍♂️ Environment (运行环境)

OS: Linux
Python 3.11
PaddleOCR paddleocr-2.9.1
Paddlepaddle-gpu 2.6
Cuda 12.2

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

python table/predict_table.py --det_model_dir=inference/en_PP-OCRv3_det_infer --rec_model_dir=inference/en_PP-OCRv3_rec_infer --table_model_dir=inference/en_ppstructure_mobile_v2.0_SLANet_infer --rec_char_dict_path=../ppocr/utils/dict/en_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --image_dir=/home/mbirla/table_0.jpg     --output=../output/table1

GreatV · 2025-01-24T01:40:15Z

GreatV
Jan 24, 2025
Maintainer

You can try installing paddle 3.0 rc https://www.paddlepaddle.org.cn/install/quick

0 replies

mridulbirla · 2025-01-24T05:36:37Z

mridulbirla
Jan 24, 2025
Author

Thanks upgrading paddle 3.0rc solved the issue. But then i get wierd output for en model in

python table/predict_table.py --det_model_dir=inference/en_PP-OCRv3_det_infer --rec_model_dir=inference/en_PP-OCRv3_rec_infer --table_model_dir=inference/en_ppstructure_mobile_v2.0_SLANet_infer --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --imag
e_dir=/home/mbirla/table_1.jpg     --output=../output/table1

[2025/01/23 21:29:04] ppocr INFO: <html><body><table><thead><tr><td>0hδΓ&lt;strike&gt;`η➔|&lt;/sub&gt;➔ζ&lt;/sub&gt;Δ|Δ➔δtη➔&lt;/sub&gt;|h`Γ θΓf|`7&lt;/sub&gt;ddΔ εδpδtζ`·➔`tη➔&lt;/sub&gt;d➔x`Γf&lt;/sub&gt;η</td><td>&#x27;</td><td>─$ /☆&lt;/overline&gt;$▼ $$α &#x27;</td><td></td><td>☆$ /☆☆▼ ⋅&lt;/overline&gt; &#x27;</td><td></td><td>&lt;/overline&gt;⋅ /☆α▼ ⋅&lt;/overline&gt;←</td></tr></thead><tbody><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr><tr><td></td><td></td><td></td><td></td><td></td><td></td><td></td></tr></tbody></table></body></html>

if as per #12783 (comment) if i use en_dict.txt i get

python table/predict_table.py --det_model_dir=inference/en_PP-OCRv3_det_infer --rec_model_dir=inference/en_PP-OCRv3_rec_infer --table_model_dir=inference/en_ppstructure_mobile_v2.0_SLANet_infer --rec_char_dict_path=../ppocr/utils/dict/en_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --image_dir=/home/mbirla/table_1.jpg     --output=../output/table1

Traceback (most recent call last):
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 241, in <module>
    main(args)
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 189, in main
    pred_res, _ = table_sys(img)
                  ^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 100, in __call__
    dt_boxes, rec_res, det_elapse, rec_elapse = self._ocr(copy.deepcopy(img))
                                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppstructure/table/predict_table.py", line 145, in _ocr
    rec_res, rec_elapse = self.text_recognizer(img_crop_list)
                          ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/tools/infer/predict_rec.py", line 805, in __call__
    rec_result = self.postprocess_op(
                 ^^^^^^^^^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppocr/postprocess/rec_postprocess.py", line 208, in __call__
    text = self.decode(
           ^^^^^^^^^^^^
  File "/home/mbirla/PaddleOCR/ppocr/postprocess/rec_postprocess.py", line 156, in decode
    char_list = [
                ^
  File "/home/mbirla/PaddleOCR/ppocr/postprocess/rec_postprocess.py", line 157, in <listcomp>
    self.character[text_id] for text_id in text_index[batch_idx][selection]
    ~~~~~~~~~~~~~~^^^^^^^^^

0 replies

GreatV · 2025-01-28T02:36:28Z

GreatV
Jan 28, 2025
Maintainer

The runtime error you encountered during table structure prediction in PaddleOCR indicates an issue with tensor dimensions or GPU configuration. Below is a detailed analysis and potential solutions:

Error Analysis

Primary Error:
```
RuntimeError: (PreconditionNotMet) Tensor's dimension is out of bound.Tensor's dimension must be equal or less than the size of its memory. But received Tensor's dimension is 8, memory's size is 0.
```
This suggests that the tensor dimensions expected by the model do not align with the memory allocated. This can happen due to:
- Incorrect model files or incompatible versions.
- Invalid input dimensions (e.g., incorrect preprocessing of images).
- Issues with GPU memory allocation or GPU configuration.
GPU Warning:
```
E0123 14:15:25.063272 21845 analysis_config.cc:125] Please use PaddlePaddle with GPU version.
```
This implies that PaddlePaddle was built without GPU support or there’s a mismatch in the GPU setup.
Referenced Issue (IndexError):
- The "IndexError" mentioned in the provided related issue (too many indices for array) occurs when the input arrays are not formatted correctly or are 1-dimensional instead of 2-dimensional. This error is common when:
  - OCR detection (det_model) or recognition (rec_model) outputs are invalid.
  - Preprocessing issues, such as incorrect padding or image resizing.

Potential Solutions

1. Verify Model Compatibility

Ensure that the models (det_model, rec_model, and table_model) are compatible with the PaddleOCR version you are using (paddleocr-2.9.1).
Use the official PaddleOCR model zoo to download the appropriate models:
PaddleOCR Model Zoo.

2. Check Model Paths and Files

Ensure the paths provided for the --det_model_dir, --rec_model_dir, and --table_model_dir point to valid and complete inference model files. Each directory should contain the following:
- inference.pdmodel
- inference.pdiparams
- inference.pdiparams.info

3. Verify GPU Compatibility

Ensure that you have installed the GPU version of PaddlePaddle (paddlepaddle-gpu).
Verify that your CUDA (12.2) and cuDNN versions are compatible with the installed PaddlePaddle version. Refer to the PaddlePaddle GPU Compatibility Table for details.
Use the command below to check GPU availability:
```
python -c "import paddle; print(paddle.is_compiled_with_cuda())"
```
If it returns False, reinstall paddlepaddle-gpu with a matching CUDA version.

4. Image Preprocessing

Ensure the input image is correctly preprocessed. Follow these steps:

Verify the image format is supported (e.g., .jpg, .png).
Ensure the image is not corrupted.
Resize the image if necessary, as excessively large images might cause memory issues.

Example preprocessing:

from PIL import Image
import cv2

# Load and preprocess the image
img_path = "/home/mbirla/table_0.jpg"
img = cv2.imread(img_path)
if img is None:
    raise ValueError("Image could not be loaded. Check the file path or format.")
img = cv2.resize(img, (1024, 1024))  # Resize if needed
cv2.imwrite("/home/mbirla/processed_table.jpg", img)

5. Update PaddleOCR and PaddlePaddle

There may be bugs in the current version you are using. Update both PaddleOCR and PaddlePaddle to the latest versions:
```
pip install --upgrade paddlepaddle-gpu paddleocr
```

6. Test with CPU

If GPU issues persist, test the pipeline with CPU to isolate the problem:

python table/predict_table.py \
  --det_model_dir=inference/en_PP-OCRv3_det_infer \
  --rec_model_dir=inference/en_PP-OCRv3_rec_infer \
  --table_model_dir=inference/en_ppstructure_mobile_v2.0_SLANet_infer \
  --rec_char_dict_path=../ppocr/utils/dict/en_dict.txt \
  --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt \
  --image_dir=/home/mbirla/table_0.jpg \
  --output=../output/table1 \
  --use_gpu=False

7. Debug Tensor Dimensions

Add debug prints in predict_table.py to check the shape of tensors before the error occurs. For example:
```
print("Tensor dimensions:", tensor.shape)
```

8. Related Issue Insights

The "IndexError" from the linked issue suggests preprocessing errors might also cause problems during table structure matching. Ensure the OCR detection and recognition results are valid and well-aligned.

Final Checklist

Verify GPU installation: paddlepaddle-gpu with correct CUDA/cuDNN versions.
Ensure model files are complete and compatible with your PaddleOCR version.
Preprocess the input image to avoid oversized or corrupted inputs.
Test with CPU (--use_gpu=False) to rule out GPU-specific issues.
Update to the latest PaddleOCR and PaddlePaddle versions.

If the issue persists after trying these steps, additional debugging or sharing input/output tensor details may be necessary.

Response generated by feifei-bot | chatgpt-4o-latest

0 replies

dimitri009 · 2025-02-21T16:40:51Z

dimitri009
Feb 21, 2025

@mridulbirla did you solve ?

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Predict Table Throwing runtime error #14605

Uh oh!

{{title}}

Uh oh!

Replies: 4 comments

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{editor}}'s edit

{{editor}}'s edit

Uh oh!

Uh oh!

{{title}}

Uh oh!

Uh oh!

{{title}}

Uh oh!

Select a reply

Uh oh!

Predict Table Throwing runtime error #14605

Uh oh!

mridulbirla Jan 23, 2025

🔎 Search before asking

🐛 Bug (问题描述)

🏃‍♂️ Environment (运行环境)

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

Replies: 4 comments

Uh oh!

GreatV Jan 24, 2025 Maintainer

Uh oh!

Uh oh!

mridulbirla Jan 24, 2025 Author

Uh oh!

GreatV Jan 28, 2025 Maintainer

Error Analysis

Potential Solutions

1. Verify Model Compatibility

2. Check Model Paths and Files

3. Verify GPU Compatibility

4. Image Preprocessing

5. Update PaddleOCR and PaddlePaddle

6. Test with CPU

7. Debug Tensor Dimensions

8. Related Issue Insights

Final Checklist

Uh oh!

dimitri009 Feb 21, 2025

mridulbirla
Jan 23, 2025

GreatV
Jan 24, 2025
Maintainer

mridulbirla
Jan 24, 2025
Author

GreatV
Jan 28, 2025
Maintainer

dimitri009
Feb 21, 2025