Exploration: Hyper Paremeter Tuning

Hyperparameter Tuning is a step during model creation where we optimise the parameters passed to the algorithm used to create the model. This is often a process that takes a long time, since you train the model over and over again, and pipelines help with that. Beyond creation, this could be part of the CI/CD pipeline for Machine Learning, any time a new commit is added a pipeline can be triggered to compute the optimal parameters.

Deliverable:

A repository that creates with a pipeline that finds the best hyperparameter set giving a certain amount of iterations, and posts the results to the MR

Repo

https://gitlab.com/gitlab-org/incubation-engineering/mlops/hyperparameter-tuning-exploration

Examples:

• https://wandb.ai/site/articles/running-hyperparameter-sweeps-to-pick-the-best-model-using-w-b
• https://cloud.google.com/blog/products/ai-machine-learning/hyperparameter-tuning-using-tpus-cloud-ml-engine