Skip to content

Skills suchen.../

Agent Skill Search Engine

Suche

Suche
Kategorien
Berufe

About

About
Privacy
Terms

© 2026 Skills Pool. Alle Rechte vorbehalten.

Hf Job Ops Playbook | Skills Pool

Skill-Datei

Hf Job Ops Playbook

Operate Hugging Face Jobs reliably for training and evaluation in this repository, with reproducible submission records, monitoring, and retry flow. Use when submitting, monitoring, or triaging HF Jobs.

Haruk1y0 Sterne01.03.2026

Beruf
Kategorien: Laborwerkzeuge

Skill-Inhalt

Use this skill for end-to-end Hugging Face Jobs operations in Mistral_Hackathon.

When To Use

User asks to submit training or eval jobs on Hugging Face.
User asks for job status, logs, failure triage, or rerun.
You need reproducible submission ledgers and job ids.

Preflight

Authenticate CLI:

hf auth whoami

Confirm required secrets are set:

HF_TOKEN
WANDB_API_KEY (if W&B logging is enabled)
MISTRAL_API_KEY (if Mistral API fallback is used by eval)

Run dry-run first:

SFT: node scripts/hf/submit_sft_job.mjs
Eval: node scripts/hf/submit_eval_job.mjs

Submit Workflow

Verwandte Skills

Schnellinstallation

Hf Job Ops Playbook

npx skillvault add Haruk1y/haruk1y-mistral-hackathon-skills-public-hf-job-ops-playbook-skill-md

Skill herunterladen Repository öffnen

Autor: Haruk1y
Sterne: 0
Aktualisiert: 01.03.2026
Beruf

Auf dieser Seite

Submit SFT job.

HF_JOB_SUBMIT=true node scripts/hf/submit_sft_job.mjs
Record is appended to artifacts/hf_jobs/submissions.jsonl.

Submit eval job.

HF_EVAL_JOB_SUBMIT=true node scripts/hf/submit_eval_job.mjs
Record is appended to artifacts/hf_jobs/eval_submissions.jsonl.

Capture job id from CLI and ledger.

If id parse fails, keep stdout_tail and stderr_tail for manual lookup.

Monitoring Workflow

Use HF CLI directly:

Running jobs: hf jobs ps --namespace Haruk1y
All jobs: hf jobs ps -a --namespace Haruk1y
Inspect: hf jobs inspect <job_id>
Logs: hf jobs logs -f <job_id>
Usage stats: hf jobs stats <job_id>

Failure Triage

Submission failure:

Check hf auth whoami.
Verify required secrets and env vars.
Re-run dry-run command and inspect built command string.

Runtime failure:

Inspect logs and classify as data, dependency, OOM, or auth issue.
Apply smallest config change needed and re-submit.

Post-run artifact missing:

Confirm push_to_hub and output model id settings.
Confirm local metrics files exist in output dir.

Repository Mapping

SFT submission: scripts/hf/submit_sft_job.mjs
Eval submission: scripts/hf/submit_eval_job.mjs
Training script: scripts/hf/train_sft_request_to_hidden_lm.py
Eval runner: scripts/wandb/weave_eval_runner.py
Submission ledgers: artifacts/hf_jobs/submissions.jsonl, artifacts/hf_jobs/eval_submissions.jsonl

References

references/job-ops-checklist.md

02

Preflight

03Submit Workflow

04Monitoring Workflow

05Failure Triage

06Repository Mapping

Data Scientists

Automation Audit Ops

Evidence-first automation inventory and overlap audit workflow for ECC. Use when the user wants to know which jobs, hooks, connectors, MCP servers, or wrappers are live, broken, redundant, or missing before fixing anything.

Github Qa Labels

Label GitHub issues and PRs found during QA testing. Use when organizing QA findings with proper labels.

storybookjs89.7k

Jupyter Notebook

Gamitin kapag humihiling ang gumagamit na gumawa, mag-scaffold, o mag-edit ng mga Jupyter notebook (`.ipynb`) para sa mga eksperimento, eksplorasyon, o tutorial; mas piliin ang mga naka-bundle na template at patakbuhin ang helper script na `new_notebook.py` upang makabuo ng malinis na panimulang notebook.

Tidb Integrationtest Recorder

Use when recording TiDB integration tests under tests/integrationtest and verifying regenerated result files stay minimal and correct.

Quality Nonconformance

Codified expertise for quality control, non-conformance investigation, root cause analysis, corrective action, and supplier quality management in regulated manufacturing.

Hugging Face Trackio

Track ML experiments with Trackio using Python logging, alerts, and CLI metric retrieval.

Data Scientists