hugging-face-model-trainer
This skill should be used when users want to train or fine-tune language models using TRL (Transformer Reinforcement Learning) on Hugging Face Jobs infrastructure. Covers SFT, DPO, GRPO and reward modeling training methods, plus GGUF conversion for local deployment. Includes guidance on the TRL Jobs package, UV scripts with PEP 723 format, dataset preparation and validation, hardware selection, cost estimation, Trackio monitoring, Hub authentication, and model persistence. Should be invoked for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.
94
92%
Does it follow best practices?
Impact
99%
1.65xAverage score across 3 eval scenarios
Advisory
Suggest reviewing before use
Quality
Discovery
100%Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.
This is a strong skill description that excels across all dimensions. It provides comprehensive coverage of specific capabilities, includes natural trigger terms users would actually use, explicitly states both what the skill does and when to invoke it, and carves out a distinct niche that minimizes conflict risk with other skills.
| Dimension | Reasoning | Score |
|---|---|---|
Specificity | Lists multiple specific concrete actions: 'train or fine-tune language models', 'SFT, DPO, GRPO and reward modeling training methods', 'GGUF conversion', 'dataset preparation and validation', 'hardware selection', 'cost estimation', 'Trackio monitoring', 'Hub authentication', and 'model persistence'. | 3 / 3 |
Completeness | Clearly answers both what (training methods, GGUF conversion, dataset prep, monitoring, etc.) AND when ('when users want to train or fine-tune', 'Should be invoked for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs'). | 3 / 3 |
Trigger Term Quality | Includes strong natural keywords users would say: 'train', 'fine-tune', 'language models', 'TRL', 'Hugging Face Jobs', 'SFT', 'DPO', 'GRPO', 'GGUF conversion', 'cloud GPU training', 'local deployment'. Good coverage of both technical terms and natural phrases. | 3 / 3 |
Distinctiveness Conflict Risk | Highly distinctive with clear niche: specifically targets TRL on Hugging Face Jobs infrastructure, not generic ML training. The combination of 'Hugging Face Jobs', 'TRL', specific training methods (SFT, DPO, GRPO), and GGUF conversion creates a unique fingerprint unlikely to conflict with other skills. | 3 / 3 |
Total | 12 / 12 Passed |
Implementation
85%Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.
This is a high-quality, production-ready skill for TRL training on Hugging Face Jobs. It excels in actionability with complete, executable code examples and clear workflow guidance with validation checkpoints. The progressive disclosure is well-implemented with appropriate separation of concerns. Minor verbosity issues exist with some repeated warnings and explanatory content that could be tightened.
| Dimension | Reasoning | Score |
|---|---|---|
Conciseness | The skill is comprehensive but includes some redundancy (e.g., repeated warnings about ephemeral environments, multiple mentions of timeout defaults). Some sections like 'When to Use This Skill' explain things Claude could infer. However, most content is necessary for this complex topic. | 2 / 3 |
Actionability | Excellent executable examples throughout - complete Python code blocks with hf_jobs() calls, proper PEP 723 headers, CLI commands with correct syntax. The training scripts are copy-paste ready with all required parameters clearly shown. | 3 / 3 |
Workflow Clarity | Clear multi-step workflows with explicit validation checkpoints. The Prerequisites Checklist, Dataset Validation workflow (validate → check markers → apply mapping → train), and Verification Checklist before submission demonstrate proper sequencing with feedback loops for error recovery. | 3 / 3 |
Progressive Disclosure | Well-structured with clear overview in SKILL.md and one-level-deep references to detailed materials (references/training_methods.md, references/gguf_conversion.md, scripts/). Navigation is clear with a comprehensive Resources section organizing internal references, scripts, and external links. | 3 / 3 |
Total | 11 / 12 Passed |
Validation
90%Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.
Validation — 10 / 11 Passed
Validation for skill structure
| Criteria | Description | Result |
|---|---|---|
skill_md_line_count | SKILL.md is long (718 lines); consider splitting into references/ and linking | Warning |
Total | 10 / 11 Passed | |
- Repository
- huggingface/skills
- Commit
73246ad
- Reviewed
Table of Contents
Is this your skill?
If you maintain this skill, you can claim it as your own. Once claimed, you can manage eval scenarios, bundle related skills, attach documentation or rules, and ensure cross-agent compatibility.