Example notebook for DNN training

[rogerkuou]

After the model training split and the train-test split, we will have dask data bags with data-frames. These DataFrames can also be used in DNN trainings, as in the example python script.

Currently when training for large number of gridcells, we parallel the workflow by submitting each grid cell as a separate SLURM job, as this example. We can improve this by using Dask SLURMCluster instead of directly submit via SLURM. In this way one can reserve a continuous time on the HPC computation resources and get less chance to be interrupted by a SLURM queue.

Goal: have 1) an example Jupyter NB and 2) and example python script to:

1. load example data as Xarray following this example
2. Split data to 1) per grid cell per training and 2) train test dataset following this example (https://github.com/VegeWaterDynamics/motrainer/tree/generalization/exploration)
3. Train the splitted data following the example python script, parallel per grid using Dask SLURMCLUSTER

[rogerkuou]

Hi @SarahAlidoost, I assigned you here just to keep track of who is working on what. We can discuss details when you would like to start.

[rogerkuou]

solved by #97