Bartosz Brzoza

Disabling JSHint

1

Is there a way to disable JSHint? Currently the preview doesn't work when there are no semicolons on the ends of the lines.

Multiple GPUs using DataParallel

5

This allows for utilization of multiple GPUs for training of MALA models. It is done with DataParallel, which in contrast to DistributedDataParallel does not require multiprocessing. Simply use ```python parameters.running.num_gpus...

Zero validation data loss during hyperparameter optimization

1

When running TPE hyperparameter optimization with ```python parameters.hyperparameters.hyper_opt_method = "optuna" parameters.running.after_before_training_metric = "band_energy" ... ``` validation data loss (during training) is always 0 ``` Epoch 0: validation data loss: 0.000e+00,...

This PR includes logging overhaul and makes energy error calculations uniform (in units of meV/atom).

This solves #573