Tools for Optuna, MLflow and the integration of both

HPOflow

Tools for Optuna, MLflow and the integration of both.

The main components are:

• hpoflow.OptunaMLflow:
  
  A wrapper to use Optuna and log to MLflow at the same time.
• hpoflow.OptunaMLflowCallback:
  
  Class inheriting from transformers.TrainerCallback that integrates with OptunaMLflow
  to send the logs to MLflow and Optuna during model training.
• hpoflow.SignificanceRepeatedTrainingPruner:
  
  An Optuna pruner
  to use statistical significance (a t-test which serves as a heuristic) to stop
  unpromising trials early, avoiding unnecessary repeated training during cross validation.

Installation

HPOflow is available at the Python Package Index (PyPI).
It can be installed with pip:

$ pip install hpoflow

Some additional dependencies might be necessary.

To use hpoflow.optuna_mlflow.OptunaMLflow:

$ pip install mlflow GitPython

To use hpoflow.optuna_transformers.OptunaMLflowCallback:

$ pip install mlflow GitPython transformers

GitHub

https://github.com/telekom/HPOflow