Championship setup
Choose the model, dataset, search style, and challenge rules.
Competition limits
Championship arena
Trials race across checkpoints for performance, stability, and efficiency.
Resource dashboard
Hyperparameter workshop
Tune values manually or let the selected search strategy propose them.
Trial estimate
Trial history
Compare experiments, duplicate settings, and protect promising configurations.
| Trial | Model | Score | Test | Gap | Latency | Size | CPU | GPU | Time | Status | |
|---|---|---|---|---|---|---|---|---|---|---|---|
| No trials yet. | |||||||||||
Live leaderboard
Championship events
Performance laboratory
Inspect score progression, costs, learning behaviour, trade-offs, and diagnostic metrics.
Achievements
Accessibility and controls
Keyboard: T starts a trial, R randomises, B restores the best, Space pauses.
Best model report
How to play
1. Set constraints
Select a model and dataset. Decide how much time, compute, memory, and latency you can spend.
2. Tune intelligently
Adjust parameters manually or use strategy suggestions. Watch estimated costs before launching each experiment.
3. Win efficiently
Maximise validation performance while preserving resources. Strong generalisation, speed, simplicity, and stability raise the championship score.