Datasets:

---

uv-scripts

/

embeddings

like 2

Follow

uv scripts for HF Jobs 182

Recovery no longer requires knowing which rank failed:

• --resume --run-id <id>: waits for any still-in-flight workers, diffs the
  manifest against worker done-states in the bucket, re-runs exactly the
  missing ranks, then consolidates. Idempotent - safe to run repeatedly.
• Attached runs now auto-retry failed workers once before giving up, and the
  failure message points at --resume instead of per-rank instructions.
• README: recovery section + fleet dashboard link.

Tested on Jobs: 2-shard run launched with --no-wait, one worker canceled
mid-run, single --resume converged it to the exact 4,000-row output
(davanstrien/resume-smoke-test).

Co-Authored-By: Claude Fable 5 noreply@anthropic.com