Watching a Hyperparameter Search in the Browser
A seeded atune study searches a multimodal function live in your browser, one trial at a time. atune is a Rust-native hyperparameter optimization library, built for reinforcement learning. You describe the knobs, such as a learning rate or a network width, and atune proposes values, watches the results, and spends the remaining budget where the results are best. Its core compiles to WebAssembly, so the study below runs in your browser. Press Open preview to start it. The study minimizes the two-dimensional Rastrigin function, whose grid of local minima traps a naive search. Pick the Random, TPE or CMA-ES sampler and press Start: each trial lands on the heatmap, and a chart tracks the best value so far. Every run uses the seed 42 and 64 trials, so the same sampler repeats exactly. atune is in pre-release development, and nothing is on crates.io or PyPI yet. Its documentation describes the library, its Python package and its command-line tool.
