is there a way to force MATLAB to use a different HDF5 library?

I have an H5 version conflict when I try to use python functions from matlab.

pyenv is correctly assigned, but since the python package I'm trying to use has a specific version of H5, MATLAB crashes with an error:

UserWarning: h5py is running against HDF5 1.8.12 when it was built against 1.10.4, this may cause problems

'{0}.{1}.{2}'.format(*version.hdf5_built_version_tuple)

i.e. my python package has 1.10.4 while MATLAB has 1.8.12

Can I force MATLAB to use a different H5 library? I tried to change the python H5 version, but there is no easily available package with 1.8.12 specifically.

[ARUN]

Hello, I am facing a similar issue, did you got this resolved?

[Douglas Franz]

@ARUN the answer below worked for me (using `OutOfProcess`).

[ARUN]

No, I am still facing the issue.

I am trying to call a python module where I load a Neural network model as.H5 from MATLAB. Whenever I load .H5 file into Python I am getting MATLAB error so I thought it could be mismatch of versions between the two.

But with the OutOfprocess, it resulted the same.

[ARUN]

No, Still I am facing the issue.

I am trying to call a python module from MATLAB. I have a Neural network model loaded into the python module as .h5 file. Whenever I call the python with .h5 file loaded, matlab throws error. I thought it could be because of library version mismatch between two.

When I tried with OutOfprocess, it resulted the same

[Maxwell Twardowski]

I have the same issue, but with netcdf. My netcdf functions are using hdf5 v 1.8.16, but matlab has 1.8.12

It would be nice to be able to change the hdf5 version matlab is using, or specify in my make file to use hdf5 v 1.8.12

Help!

Answer 1

[Durriya Doctor]

You can avoid the library conflict by running your Python functions in out-of-process mode. See https://www.mathworks.com/help/matlab/matlab_external/out-of-process-execution-of-python-functionality.html?searchHighlight=out%20of%20process%20python&s_tid=srchtitle

[Douglas Franz]

Thanks. This worked for me, although I'm hoping that the performance difference that is mentioned won't matter much.