pydantic error

[breisfeld]

Hi, When running nougat on a test pdf file

$ nougat /Users/XXXXX/Downloads/pharmaceutics-16-00226.pdf -o output

I get the following traceback:

Traceback (most recent call last):
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/bin/nougat", line 5, in <module>
    from predict import main
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/predict.py", line 18, in <module>
    from nougat import NougatModel
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/nougat/__init__.py", line 7, in <module>
    from .model import NougatConfig, NougatModel
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/nougat/model.py", line 34, in <module>
    from nougat.transforms import train_transform, test_transform
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/nougat/transforms.py", line 146, in <module>
    alb.ElasticTransform(
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/albumentations/core/validation.py", line 35, in custom_init
    config = dct["InitSchema"](**full_kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/Users/XXXXX/miniconda3/envs/extract_pdf_tables/lib/python3.12/site-packages/pydantic/main.py", line 193, in __init__
    self.__pydantic_validator__.validate_python(data, self_instance=self)
pydantic_core._pydantic_core.ValidationError: 1 validation error for InitSchema
alpha_affine
  Input should be None [type=none_required, input_value=1.2, input_type=float]
    For further information visit https://errors.pydantic.dev/2.8/v/none_required

I get the same traceback when running nougat with no arguments.

I have tried running this same command under MacOS, Linux, and Windows. On all platforms, nougat was installed into a fresh conda virtual environment using pip.

Any advice on solving the problem?

Thanks.

[YuCheng-Qi]

Modify alb.ElasticTransform( p=1, alpha=50, sigma=120 0.1, alpha_affine=120 0.01, border_mode=0, value=(255, 255, 255), ), to alb.ElasticTransform in transforms.py (p=1, alpha=50, sigma=120 * 0.1, alpha_affine=None, border_mode=0, value=(255, 255, 255), ),

[YuCheng-Qi]

@breisfeld

[eesyfep]

mode Even I changed the parameters, it shows the same warning

[shenxianovo]

same issue

[breisfeld]

Modify alb.ElasticTransform( p=1, alpha=50, sigma=120 0.1, alpha_affine=120 0.01, border_mode=0, value=(255, 255, 255), ), to alb.ElasticTransform in transforms.py (p=1, alpha=50, sigma=120 * 0.1, alpha_affine=None, border_mode=0, value=(255, 255, 255), ),

That seemed to take care of the issue for me. Thanks!

[YodaGitMaster]

Thanks

[e-hossam96]

Modify alb.ElasticTransform( p=1, alpha=50, sigma=120 0.1, alpha_affine=120 0.01, border_mode=0, value=(255, 255, 255), ), to alb.ElasticTransform in transforms.py (p=1, alpha=50, sigma=120 * 0.1, alpha_affine=None, border_mode=0, value=(255, 255, 255), ),

It worked and the model started downloading. However, after downloading the model checkpoint, it immediately throws a connection error. (WSL2)

[SichangHe]

Can't even make a PR to fix this lol. They locked it down to collaborator-only.

[samiahhassan]

I am having the same issues with nougat. I am new to OCR. Please assist me with where I should make the change for the ElasticTransform function or any other way. Thanks for any advice.

Modify alb.ElasticTransform( p=1, alpha=50, sigma=120 0.1, alpha_affine=120 0.01, border_mode=0, value=(255, 255, 255), ), to alb.ElasticTransform in transforms.py (p=1, alpha=50, sigma=120 * 0.1, alpha_affine=None, border_mode=0, value=(255, 255, 255), ),

[anshumankmr]

Getting some strange error after making that edit,

No module named nougat.main; 'nougat' is a package and cannot be directly executed

Windows 11

[snimavat]

Same error, Macbook pro Python 3.11.2

[dimazig]

same issue

Traceback (most recent call last):
  File "/home/dima/nougat/bin/nougat", line 5, in <module>
    from predict import main
  File "/home/dima/nougat/lib/python3.12/site-packages/predict.py", line 18, in <module>
    from nougat import NougatModel
  File "/home/dima/nougat/lib/python3.12/site-packages/nougat/__init__.py", line 7, in <module>
    from .model import NougatConfig, NougatModel
  File "/home/dima/nougat/lib/python3.12/site-packages/nougat/model.py", line 34, in <module>
    from nougat.transforms import train_transform, test_transform
  File "/home/dima/nougat/lib/python3.12/site-packages/nougat/transforms.py", line 146, in <module>
    alb.ElasticTransform(
  File "/home/dima/nougat/lib/python3.12/site-packages/albumentations/core/validation.py", line 35, in custom_init
    config = dct["InitSchema"](**full_kwargs)
             ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
  File "/home/dima/nougat/lib/python3.12/site-packages/pydantic/main.py", line 209, in __init__
    validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
                     ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
pydantic_core._pydantic_core.ValidationError: 1 validation error for InitSchema
alpha_affine
  Input should be None [type=none_required, input_value=1.2, input_type=float]
    For further information visit https://errors.pydantic.dev/2.9/v/none_required

[SichangHe]

Try my fork for now, I merged the corresponding PRs:

python3 -m pip install git+https://github.com/SichangHe/facebookresearch--nougat.git
# Or, using Pipx:
pipx install git+https://github.com/SichangHe/facebookresearch--nougat.git
# Or, if you are using Rye:
rye add --git https://github.com/SichangHe/facebookresearch--nougat.git nougat-ocr

[Fireblossom]

pip install albumentations==1.4.8

[breisfeld]

pip install albumentations==1.4.8

@SichangHe , should the existing line in setup.py in your fork be replaced by this if it is the true dependency?

[SichangHe]

pip install albumentations==1.4.8

@SichangHe , should the existing line in setup.py in your fork be replaced by this if it is the true dependency?

I doublechecked, I have 1.4.14 in my environment and it still works.

Also, I am moving away to marker-pdf because it seems to give fewer spurious errors, although their header detection and math decoding really suck.

[anshumankmr]

Instead of calling the nougat thing via the CLI, I chose to install nougat_api and it worked for me. That was sufficient for my use case.