Multiscale Demand Forecasting
If some packages have been updated (pyforecaster), run
pip install -r /home/elettra/remote_projects/V2G4/requirements.txt --force-reinstall --upgrade
this force a re-installation of all packages in the requirements.txt file, including jax[cuda12_local]
which must be
installed in a specific sequence (especially after jax)
NOTE: It could be needed to run the following:
pip install --upgrade "jax[cuda12_local]==0.4.23" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
Seems like the version 0.4.23 is the only one that works with the current version of jaxlib, see here
YoshitakaMo/localcolabfold#212
pip install --upgrade "jax[cuda12_local]==0.4.26" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
and optax == 0.2.2 now works