hugging-face-jobs
Run workloads on Hugging Face Jobs with managed CPUs, GPUs, TPUs, secrets, and Hub persistence.
Profile →npx skills add sickn33/agentic-awesome-skills --skill hugging-face-jobs --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
It enables executing Python workloads on Hugging Face Jobs using fully managed cloud resources (CPUs, GPUs, and TPUs) and optionally persists results to the Hugging Face Hub. It highlights common use cases such as data processing, batch inference, experiments, model training, synthetic data generation, development/testing, and scheduled jobs.
How it works
- Requires using the hf_jobs MCP tool to submit jobs, either with hf_jobs("uv", {...}) or hf_jobs("run", {...}), passing the script as a string. The script content is provided as the value of the script field.
- Ensures authentication by requiring HF_TOKEN via secrets for Hub interactions. Tokens can be provided via secrets {"HF_TOKEN": "$HF_TOKEN"} or retrieved within the script. When using HfApi().run_uv_job(), the real token must be passed.
- After submission, the agent must provide job details: job ID, monitoring URL, estimated time, and the option to request status checks later.
- Sets default timeouts (e.g., 30m) and allows longer runs when needed.
- Supports two main approaches: UV scripts (inline dependencies declared in the script header) and Docker-based jobs, with examples for both.
- Provides guidance on token handling, including automatic replacement of the HF_TOKEN placeholder and security considerations.
- Includes examples for pushing results to the Hub, external storage options, and API-based result delivery.
When to use it
Use this skill when you want to run Python workloads on cloud infrastructure without local GPU/TPU setup, process data at scale, run batch inference or experiments, schedule recurring tasks, or persist results to the Hugging Face Hub. It is suitable for CPU, GPU, and TPU workloads and supports both inline scripts and Docker-based execution.
What it can touch
- Tools: hf_jobs() MCP tool (to submit jobs) and related API usage with secret HF_TOKEN for Hub access.
- Scripts: script content is provided inline; the skill documents how to embed dependencies and code within the script string or supply a URL. When using the MCP tool, the script is passed as a string rather than saved to local files unless explicitly requested by the user.
- Secrets: HF_TOKEN is required for Hub interactions; token handling methods are demonstrated (automatic replacement, get_token, and environment variable usage).
Caveats
- The environment is ephemeral; results must be persisted to avoid data loss.
- Jobs require Hugging Face account with an appropriate plan (paid) and valid tokens with proper permissions.
- There are explicit constraints on how tokens are supplied to different APIs (automatic replacement vs. hard-coded tokens).
- The skill emphasizes not saving scripts to local files unless requested and recommends using secrets for tokens instead of environment variables where possible.
# Running Workloads on Hugging Face Jobs ## Overview Run any workload on fully managed Hugging Face infrastructure. No local setup required—jobs run on cloud CPUs, GPUs, or TPUs and can persist results to the Hugging Face Hub. **Common use cases:** - **Data Processing** - Transform, filter, or analyze large datasets - **Batch Inference** - Run inference on thousands of samples - **Experiments & Benchmarks** - Reproducible ML experiments - **Model Training** - Fine-tune models (see `model-trainer` skill for TRL-specific training) - **Synthetic Data Generation** - Generate datasets using LLMs - **Development & Testing** - Test code without local GPU setup - **Scheduled Jobs** - Automate recurring tasks **For model training specifically:** See the `model-trainer` skill for TRL-based training workflows. ## When to Use This Skill Use this skill when users want to: - Run Python workloads on cloud infrastructure - Execute jobs without local GPU/TPU setup - Process data at scale - Run batch inference or experiments - Schedule recurring tasks - Use GPUs/TPUs for any workload - Persist results to the Hugging Face Hub ## Key Directives When assisting with jobs: 1. **ALWAYS use `hf_jobs()` MCP
- Overview
- When to Use This Skill
- Key Directives
- Prerequisites Checklist
- ✅ Account & Authentication
- ✅ Token Usage (See Token Usage section for details)
- Token Usage Guide
- Understanding Tokens
- When Tokens Are Required
- How to Provide Tokens to Jobs
- Using Tokens in Scripts
- Token Verification
- Common Token Issues
- Token Security Best Practices
hf jobs uv run my_script.py --flavor cpu-basic --timeout 30m
hf jobs uv run --image vllm/vllm-openai:latest --flavor a10g-large inference.py
hf jobs uv run ./scripts/foo.py -- --your --args
hf jobs run python:3.12 python -c "print('Hello from HF Jobs!')"
hf jobs run hf.co/spaces/lhoestq/duckdb duckdb -c "SELECT 'Hello!'"
hf jobs ps # List jobs
hf jobs logs <job-id> # View logs
hf jobs cancel <job-id> # Cancel jobWhat does the hugging-face-jobs skill do?
Run workloads on Hugging Face Jobs with managed CPUs, GPUs, TPUs, secrets, and Hub persistence.
How do I install it?
Run `npx skills add sickn33/agentic-awesome-skills --skill hugging-face-jobs --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From sickn33/agentic-awesome-skills, a repository with 44,414 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.