Optuna¶

To be able to load your Optuna run into DeepCave, there are a few points to consider when running Optuna:

Save the Optuna study object as a Pickle file:
```
import pickle
with open("<path-to-save-study>/optuna_study.pkl", "wb") as f:
    pickle.dump(study, f)
```
Replace <path-to-save-study> with the directory where you want to store the Pickle file.
Optional: If you want to display the names of your objectives in DeepCAVE, you can set metric names in Optuna using the following command:
```
study.set_metric_names(["Accuracy", "Flops"])
```
This step is optional but recommended for better visualization and understanding of your study’s objectives in DeepCAVE.

Furthermore, to load an Optuna run into DeepCAVE, it is necessary to install Optuna, e.g. via:

pip install deepcave[optuna]

Warning

Loading Optuna runs with conditional search spaces or dynamic hyperparameter value ranges is not supported. This limitation arises because Optuna leverages a dynamic search space, which cannot be trivially converted into a static search space as used in DeepCAVE.

AMLTK

BOHB

DeepCAVE Documentation

Optuna¶