Docker image
Your image runs one trial from a normal container command.
Introduction
Managed hyperparameter optimization for Dockerized workloads.
HyperOptimizer runs your container repeatedly with different --hpo-* values, collects hpo.metrics.* from stdout, and ranks trials in the dashboard. No SDK required.
How it works
Section titled “How it works”CLI inputs
HyperOptimizer injects --hpo-* parameter values at runtime (plain --* optional).
Metric output
Your process prints structured results to stdout.
Start here
Section titled “Start here”QuickstartBuild the smallest working trial: image, parameters, metrics, publish, experiment.
Publish an imagePush to your org Images registry so trials can pull your tag.
Create an experimentWizard: image, command, search space, objectives, machines, algorithm.
Plans and creditsPrepaid wallet, plan limits, machine rates, and top-up packs.
Any Docker workloadWrap Python, Rust, Julia, shell scripts, or any binary that runs in Docker.
FreqtradeConnect a trading strategy workload and optimize backtests.
References
Section titled “References”- CLI arguments — Flag naming,
--hpo-*vs plain modes - Container requirements — Entrypoints, Images registry, reproducibility
- Machines and limits — SKUs, prices, plan caps
- Troubleshooting — Missing metrics, registry, credits
- Algorithms — Bayesian default and other search modes