[Bug]: Urgent! ValueError: Bad serialized model or bad model name

[janosh]

Email (Optional)

No response

Version

0.8.5

Which OS(es) are you using?

• MacOS
• Windows
• Linux

What happened?

matgl is holding up the whole MP stack with this error

ValueError: Bad serialized model or bad model name. It is possible that you have an older model cached. Please clear your cache by running python -c "import matgl; matgl.clear_cache()"

/opt/hostedtoolcache/Python/3.11.6/x64/lib/python3.11/site-packages/matgl/utils/io.py:213:

It's affecting pymatgen, atomate2, matcalc, emmet, ...
Please add tests for whatever broke so this doesn't happen again!

• Improve doc strings substitution_probability.py materialsproject/pymatgen#3477
• MACE Static/RelaxMakers default to loading mace_mp instead of test model materialsproject/atomate2#614
• Allow optimizer to be configured for MACE/GAP materialsproject/atomate2#615
• Add MACE to recognized UNIVERSAL_CALCULATORS matcalc#11

Code snippet

No response

Log output

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[shyuep]

Erm, I think the error message says it all? @kenko911 You need to change the model number when you make internal changes like the stress computation. Otherwise the loading of the models can be wrong.

[janosh]

How come there's no test that caught this?

[ucarh]

I had this error with python 3.12, downgrading it to 3.9 solved the issue.

[shyuep]

DGL (the underlying dependency) is not even available on py3.12. In general, we would recommend using Py3.10 (or at most one version below the latest Python, which usually has bugs as well as a lot of unsupported dependencies).