Implement a schema for performing a grid search over a parameter range.
Will need to recursively call train.py with different (but similar) parameters (without saving model to disk) and store the cross-val score to determine which parameter set is best.
This function is intended to run with the sklearn-type models, but for parameters which can't use sklearn's grid search function because they are deeper parameters.
I think we could just do this in notebooks at the moment, but scripting could be useful. I'll make a notebook about how to tune learning rate and the code used in that can be repurposed for any gridsearch code.
Implement a schema for performing a grid search over a parameter range.
Will need to recursively call train.py with different (but similar) parameters (without saving model to disk) and store the cross-val score to determine which parameter set is best.
This function is intended to run with the sklearn-type models, but for parameters which can't use sklearn's grid search function because they are deeper parameters.