Train your model on your best data.

Skyu scores every training data sample by its real impact on your model, so you can train on the best subset that improves your model.

Find the demonstrations that are hurting your policy.

Skyu scores every episode in a robot-learning dataset by how much it surprises your trained policy, flags the suspect tail, and writes a reproducible report. It runs offline, on your hardware.

$ pip install skyu-0.2.1-py3-none-any.whl

$ skyu score --dataset ./pusht --checkpoint ./ckpt/030000

scored 1,204 episodes · K=8 · seed 17

flagged 96 (top 8% + elbow)

→ report.html

report.html skyu score · 1,204 episodes · seed 17

Surprise score distribution per-episode mean loss, K=8

flag threshold · top 8%

0.00surprise →0.92

Episodes by rank known-bad episodes marked red

rank 1 · most surprisingrank 1,204

Placeholder data — 12 injected corruptions, all recovered in the flagged tail.

[ HOW IT WORKS ]

Point

Give Skyu a LeRobot-format dataset and a policy checkpoint. Nothing else — no labels, no cloud account, no instrumentation of your training run.

skyu score --dataset D --checkpoint C

Score

Every episode gets a surprise score: mean training loss over K seeded draws from your policy. Episodes the policy can't explain rise to the top.

surprise = mean loss over K draws

Review

The report flags the suspect tail — top-N% plus elbow detection — and every number in it carries the exact command that reproduces it.

→ report.html · single file