nzw0301 / optuna-wandb

Example codes in the medium post titled "Optuna meets Weights and Biases."

Optuna meets Weights and Biases

This repo provides the examples codes used in the medium post titled Optuna meets Weights and Biases.

Installation

pip

pip install wandb optuna scikit-learn torch torchvision plotly

conda

conda env create -f environment.yml
conda activate optuna-wandb

---

Updated 🚀 [11-Aug-2022]: Add as_multirun=True example to make part-1 simpler

In forthcoming optuna v3, optuna's wandb callback provides as_multirun option to trace an objective function optimised by iterative way, e.g., stochastic gradient descent. Thanks to this feature, we can combine optuna and wandb more easily.

optuna-wandb's Issues

Plotting Sweep from Optuna Trials

On the blog post image there is a hyperparameter sweep plot. How can we create that after feeding the optuna study into wandb?

Add example of `as_multirun=True`