from-312-add-experiment-nane-and-trial-number-as-parameters

[david-thrower]

Task

• The trainer script container is set to automatically create a new experiment for each run, using a timestamp and the user's name, so there is never a collision in experiment naming or a run committed to the wrong experiment.

• We want to keep this as default behavior unless we intentionally set the run to a specific trial number or experiment:

• We want to add a parameter for experiment name and trial number, so that we can commit additional runs on an existing experiment.

• Add parameter EXPERIMENT_NAME and conditional use of the experiment name or the automated experiment naming.

• Add parameter TRAIL_NUMBER.

• Fix file nomenclature so that the mlflow files will be in the same directory if the experiment name is the same, but keep the trial number in the locations for the directly saved Keras artifacts (Model.save())

• Add the file path string for where Keras artifacts are directly saved as a parameter to log in mlflow, so the user can find the artifacts and keep track of which ones belong to which run.