Closed alex000kim closed 11 months ago
Should we do this like in a new branch where we refactor to use Hydra or by default?
Not sure, tbh. Maybe a separate branch where we do a grid search over the learning rate and the batch size?
Not sure, tbh. Maybe a separate branch where we do a grid search over the learning rate and the batch size?
If it's only about grid-search, I already used it on the tune-architecture
branch. Not sure how to showcase this as it is hidden in the generate.sh
. I have tried adding it to the description in the P.R. like: https://github.com/iterative/example-get-started-experiments/pull/8
Maybe we should have a section e.g. "Things to try next" to the README and list hyperparam tuning with Hydra as one of the items?
This is already covered in https://dvc.org/doc/start/experiments/experiment-pipelines#hyperparameter-tuning
Need to showcase that DVC experiments support Hydra composition: https://dvc.org/doc/user-guide/experiment-management/hydra-composition