Add support for Python 3.13

[stweil]

The new Debian stable "Trixie" comes with Python 3.13. Therefore support for this Python release would be nice.

Here I collect my current results with kraken and Python 3.13.

• scikit-image 0.24.0 does not work with Python 3.13 ("KeyError: 'area_bbox'"), so I upgraded to scikit-image-0.25.2.

diff --git a/setup.cfg b/setup.cfg
index a936baae..902df91d 100644
--- a/setup.cfg
+++ b/setup.cfg
@@ -38,7 +38,7 @@ max_line_length = 160
 exclude = tests/*
 
 [options]
-python_requires = >=3.9,<3.13
+python_requires = >=3.9,<3.14
 install_requires =
     jsonschema
     lxml
@@ -55,7 +55,7 @@ install_requires =
     torchvision>=0.5.0
     torch>=2.4.0,<2.8.0
     scikit-learn~=1.5.0
-    scikit-image~=0.24.0
+    scikit-image>=0.24.0,<2.26.0
     shapely>=2.0.6,~=2.0.6
     pyarrow
     htrmopo>=0.3,~=0.3

pytest -k 'not test_train' fails in some tests:

============================================================================================== test session starts ==============================================================================================
platform linux -- Python 3.13.5, pytest-8.4.1, pluggy-1.6.0
rootdir: /home/stweil/src/github/mittagessen/kraken
configfile: pytest.ini
testpaths: tests
collected 226 items / 14 deselected / 212 selected                                                                                                                                                              

tests/test_align.py .                                                                                                                                                                                     [  0%]
tests/test_arrow_dataset.py ......                                                                                                                                                                        [  3%]
tests/test_binarization.py .....                                                                                                                                                                          [  5%]
tests/test_cli.py ...                                                                                                                                                                                     [  7%]
tests/test_codec.py ..............................                                                                                                                                                        [ 21%]
tests/test_dataset.py ............                                                                                                                                                                        [ 26%]
tests/test_layers.py .............................                                                                                                                                                        [ 40%]
tests/test_lineest.py ....                                                                                                                                                                                [ 42%]
tests/test_merging.py .......                                                                                                                                                                             [ 45%]
tests/test_models.py .                                                                                                                                                                                    [ 46%]
tests/test_newpolygons.py ......FFF...FFFFFFF...                                                                                                                                                          [ 56%]
tests/test_pageseg.py ..                                                                                                                                                                                  [ 57%]
tests/test_readingorder.py ......x...x....                                                                                                                                                                [ 64%]
tests/test_repo.py ...                                                                                                                                                                                    [ 66%]
tests/test_rpred.py .........................                                                                                                                                                             [ 77%]
tests/test_serialization.py .................                                                                                                                                                             [ 85%]
tests/test_transcribe.py .                                                                                                                                                                                [ 86%]
tests/test_vgsl.py ...s.....                                                                                                                                                                              [ 90%]
tests/test_xml.py ....................                                                                                                                                                                    [100%]

=================================================================================================== FAILURES ====================================================================================================
[...]
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_new_model - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_new_model_force_legacy - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_old_model - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_new_model - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_new_model_force_legacy - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_old_model - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_old_model_force_legacy - AssertionError: Expected 'mock' to have been called.
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_new_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_old_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_old_model_force_new - AssertionError: 1 != 0
======================================================= 10 failed, 199 passed, 1 skipped, 14 deselected, 2 xfailed, 1278818 warnings in 261.44s (0:04:21) =======================================================

[mittagessen]

Due to coremltools we're still stuck with python 3.12 (they tend to be quite slow with new python version support). Once they've merged it there shouldn't be anything keeping us from enabling 3.13. Here's the report on their repo.

As this is a somewhat recurring theme: the new recognizer and segmenter will use safetensors serialization as it is much more suitable as a simple serialization format. Coreml isn't really one but at the time there weren't any alternatives. It should be possible to write a parser that doesn't depend on the library, it is only a protobuf container after all, to deprecate this dependency while retaining backwards compatibility.

[stweil]

At least for the installation I had no problem with coremltools 8.3.0:

aiohappyeyeballs==2.6.1
aiohttp==3.12.15
aiosignal==1.4.0
atpublic==6.0.1
attrs==25.3.0
cattrs==25.1.1
certifi==2025.8.3
charset-normalizer==3.4.3
click==8.2.1
coremltools==8.3.0
filelock==3.18.0
flufl.lock==8.2.0
frozenlist==1.7.0
fsspec==2025.7.0
future==1.0.0
hocr-spec==0.2.0
htrmopo==0.3.0
idna==3.10
imageio==2.37.0
iniconfig==2.1.0
iso639-lang==2.6.3
Jinja2==3.1.6
joblib==1.5.1
jsonschema==4.25.0
jsonschema-specifications==2025.4.1
kraken @ file:///home/stweil/src/github/mittagessen/kraken
lazy_loader==0.4
lightning==2.4.0
lightning-utilities==0.15.2
lxml==6.0.0
Markdown==3.8.2
markdown-it-py==4.0.0
MarkupSafe==3.0.2
mdurl==0.1.2
mpmath==1.3.0
multidict==6.6.4
networkx==3.5
numpy==2.0.2
nvidia-cublas-cu12==12.6.4.1
nvidia-cuda-cupti-cu12==12.6.80
nvidia-cuda-nvrtc-cu12==12.6.77
nvidia-cuda-runtime-cu12==12.6.77
nvidia-cudnn-cu12==9.5.1.17
nvidia-cufft-cu12==11.3.0.4
nvidia-cufile-cu12==1.11.1.6
nvidia-curand-cu12==10.3.7.77
nvidia-cusolver-cu12==11.7.1.2
nvidia-cusparse-cu12==12.5.4.2
nvidia-cusparselt-cu12==0.6.3
nvidia-nccl-cu12==2.26.2
nvidia-nvjitlink-cu12==12.6.85
nvidia-nvtx-cu12==12.6.77
packaging==24.2
pillow==11.3.0
platformdirs==4.3.8
pluggy==1.6.0
propcache==0.3.2
protobuf==6.31.1
psutil==7.0.0
pyaml==25.7.0
pyarrow==21.0.0
Pygments==2.19.2
pytest==8.4.1
python-bidi==0.6.6
python-dateutil==2.9.0.post0
pytorch-lightning==2.5.2
PyYAML==6.0.2
referencing==0.36.2
regex==2025.7.34
requests==2.32.4
rich==14.1.0
rpds-py==0.27.0
scikit-image==0.25.2
scikit-learn==1.5.2
scipy==1.13.1
setuptools==80.9.0
shapely==2.0.7
Sickle==0.7.0
six==1.17.0
sympy==1.14.0
threadpoolctl==3.5.0
tifffile==2025.6.11
torch==2.7.1
torchmetrics==1.8.1
torchvision==0.22.1
tqdm==4.67.1
triton==3.3.1
typing_extensions==4.14.1
urllib3==2.5.0
yarl==1.20.1

[stweil]

A full protocol with Python 3.13 is now created in my GitHub action here:
https://github.com/stweil/kraken/actions/runs/16884630003/job/47829030222
https://github.com/stweil/kraken/actions/runs/16884835512/job/47829759508
https://github.com/mittagessen/kraken/actions/runs/16390476086/job/46315774399.

It looks like the CI tests also fail with older Python releases, maybe even with the same errors.

[mittagessen]

On 25/08/11 08:25AM, Stefan Weil wrote: stweil left a comment (mittagessen/kraken#718) At least for the installation I had no problem with coremltools 8.3.0:

It doesn't pull the wheel but a source distribution and builds the package. The metadata indicates that coremltools doesn't work with python 3.13 but pip only does this check for wheels as far as I understand. And at least on my system the wheel built from source by pip does not function complaining about missing libraries on import.

[stweil]

Meanwhile there is a coremltools-9.0b1 which looks promissing. pytest -k 'not test_train' passes on my MacBook with Python 3.12:

204 passed, 1 skipped, 14 deselected, 2 xfailed, 1275919 warnings in 387.26s (0:06:27)

With Python 3.13 (but otherwise the same Python packages), these tests fail:

FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_new_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_new_model_force_legacy - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_pretrain_old_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_new_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_new_model_force_legacy - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_old_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_ketoscli_train_old_model_force_legacy - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_new_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_old_model - AssertionError: 1 != 0
FAILED tests/test_newpolygons.py::TestNewPolygons::test_krakencli_ocr_old_model_force_new - AssertionError: 1 != 0