Steve Schmerler
Steve Schmerler
Some cleanup / update in `network/predictor.py` that I'd like to move out of feature branches into develop, in preparation of a larger PR later which includes one feature only.
Motivated by #497, here is a summary of the current resume workflow, along with questions and thoughts for improvement: MALA parallel Optuna usage is supposed to be as described [in...
This setting can be used to speed up calculations with the TEM, but is not documented.
We install all deps and the package for tests for each Python version (well, ok) but then again once more for building pages, because we use https://github.com/elcorto/sphinx-autodoc which needs an...
When using `parameters.running.during_training_metric`, the quantity (e.g. band energy) is used as validation loss, both in the text log and when fed to tensorboard, which is a bit confusing as it...
When using `during_training_metric`, the respective quantity is calculated in every epoch, which may be costly if `during_training_metric="total_energy"`. When using shuffled snapshots, adding the required `calculation_output_file` as in ```py data_handler.add_snapshot( "Be_snapshot_shuffled1.in.npy",...
When this is a required arg, it breaks `torch.save()` and `torch.load()`. But more importantly, it also breaks code that does `copy.deepcopy(model)`. One option is to make it an optional keyword...