Participant GuideTroubleshooting
Troubleshooting — When something breaks
Sourced from whest-starterkit @
aaa3882.
Troubleshooting — When something breaks
If your run errored, your score regressed, or your local and remote scores disagree, start here.
| Doc | When to read |
|---|---|
| common-participant-errors.md | Symptom → cause → fix-now → verify, for the most common failures (wrong shape, NaN/Inf, exceeded budget, signature mismatches, import errors, numeric blow-up in deep networks). |
| faq.md | Quick answers — can I use scipy? what is residual_wall_time_limit? why does my submission score worse than my local run? |
➡️ Where to look next
- Want the tiered procedure for "estimator runs but something feels wrong"? → How-to: debugging checklist.
- Suspect drift between machines? → Reference:
whest doctor. - Need to interpret
per_mlp[i].error,budget_exhausted,time_exhaustedfields? → Reference: score report fields.