Unknown matfile type in scipy loadmat running inside Github action

Question

I'm using scipy.loadmat to load a Matlab .mat file inside my Python unit tests. This arrangement works just fine on my desktop, and inside a Docker container running on my servers. However, I'm getting an error when I try to run the unit tests inside a Github action (my CI flow):

    model_dict_tmp = loadmat(model_path)
  File "/usr/local/lib/python3.6/dist-packages/scipy/io/matlab/mio.py", line 223, in loadmat
    MR, _ = mat_reader_factory(f, **kwargs)
  File "/usr/local/lib/python3.6/dist-packages/scipy/io/matlab/mio.py", line 72, in mat_reader_factory
    mjv, mnv = get_matfile_version(byte_stream)
  File "/usr/local/lib/python3.6/dist-packages/scipy/io/matlab/miobase.py", line 231, in get_matfile_version
    raise ValueError('Unknown mat file type, version %s, %s' % ret)
ValueError: Unknown mat file type, version 50, 52

The .mat file is stored inside a Github submodule (as an LFS object) and should be cloned along with the code. So I'm expecting a well-known location for the file:

mainproject/
   submodule/
      path/to/matfile.mat
   tests/
      test_matrix_loading.py  # unit test that attempts to load the matrix 

I cannot reproduce this error when spinning the same Docker image manually and running the CI as:

python -m unittest discover -s tests/

Which is the same instruction I use inside my action script.

1. Why is this happening? This error is very recent and this arrangement used to work previously.
2. How can I debug this?

Answer 1

This ended up being due to a git lfs problem. The scipy.loadmat error was just the manifestation.

See: Github Action: How to pull LFS files in submodule

For the solution.