Janis Meyer

@damian-nplan: This is something else. The config files of layout/segmentation models have changed recently. Unfortunately, there is no automatic way to force downloading model config files and this has to...

> @JaMe76 > > when i try your suggestion it says AttributeError: module deepdoctection has no attribute pytorch_available. Did you mean: 'boto3_available'? > > also i am trying to run...

Check this [tutorial](https://deepdoctection.readthedocs.io/en/latest/tutorials/running_pre_trained_models_from_third_party_libraries_notebook/). It shows you how to save your models manually in the .cache and how to add the model to the `ModelCatalog`.

To get the .cache you can find it like that ``` from deepdoctection.utils.settings import dd_cache_home print(dd_cache_home) ``` You can also change the .cache dir by changing the env variable before...

`requirements.txt` only contains no model related dependencies like Transformers, Detectron2, DocTr, etc. When installing the project you should use `setup.py`.

There is a logic that tries to infer the device/framework to use based upon availability and it seems to find something mps-related ( which isn’t obviously working). Running DocTr on...

Thanks for your comments about this repo. Regarding your questions: The training of the private layout model follows exactly the training scripts you were referring to with the only difference...

Tatr v2 does not require padding as far as I remember. So, you reducing the default padding values to 0 in the configs should improve the segmentation results.

Thanks for your PR, super happy to merge, once I get time reviewing this next week! 1.) Could you maybe add a small test? Something very similar to [this](https://github.com/deepdoctection/deepdoctection/blob/master/tests/extern/test_texocr.py) would...

I think, you are running on an older version. The issue results from the fact that it tries to download some tokenizer information (which has nothing to do for running...