This section is to be updated and improved
This is a short version on how it works.
We consider that each model might be a standalone python script.
*1) Optional: preferably, set up a conda environment for your model. You can also add the ‘conda create’ command to our Dockerfile. *2) Add a configuration .yml so your stand alone is called correctly. Specify the name of the conda environment it should be run on. *3) If a extended list of hyperparamenter has to be passed, consider to stack them into a JSON file and add it into “experiments/modelname” *4) Confirm that your model can be trained and tested by issuing commands to run.py.
That’s it! Thank you for making it easy to everyone to play with a multitude of models!