Name: huggingface-llm-trainer
Rating: 70.39999999999999 (1 reviews)
Author: huggingface

huggingface-llm-trainer

Train or fine-tune language and vision models using TRL (Transformer Reinforcement Learning) or Unsloth with 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, model selection/leaderboards and model persistence. Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup.

Quality

86%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Critical

Do not install without reviewing

Quality

Content

81%Weight 40%Scale 1-5

Reviews the quality of instructions and guidance provided to agents. Good implementation is clear, handles edge cases, and produces reliable results.

Highly actionable and well-sequenced content with strong validation feedback loops for an expensive, ephemeral workflow. Its main weakness is verbosity — recap and duplicated troubleshooting sections inflate the body and dilute the otherwise clean progressive-disclosure structure.

Suggestions

Remove the 'Key Takeaways' recap section — it re-states directives already covered above and adds length without new information.

Merge 'Common Failure Modes' and 'Troubleshooting' into one section (or defer entirely to references/troubleshooting.md) to eliminate the duplicated fix lists.

Trim the 'What is GGUF' bullet list to a single line or move it into references/gguf_conversion.md, since Claude already knows the GGUF format.

Dimension	Reasoning	Score
Conciseness	Mostly useful HF-Jobs-specific operational content, but the ~730-line body has clear tightening opportunities: a 'What is GGUF' explainer Claude already knows, a 10-item 'Key Takeaways' recap, and a 'Troubleshooting' section that duplicates 'Common Failure Modes'.	3 / 5
Actionability	Fully executable, copy-paste-ready guidance throughout: complete inline SFT training script with PEP 723 header, correct/incorrect CLI syntax blocks, hardware and cost tables, and a dataset-validation workflow with copy-paste mapping code.	5 / 5
Workflow Clarity	Multi-step processes are explicitly sequenced with validation checkpoints and feedback loops — the Dataset Validation workflow (inspect → check ✓ READY/✗ NEEDS MAPPING markers → apply mapping → launch) and the Hub-save verification checklist precede the expensive, ephemeral GPU job.	5 / 5
Progressive Disclosure	Clear one-level-deep references with a Resources section enumerating all 10 reference files and 8 scripts (all verified present), but the body inlines substantial detail (full examples, duplicated troubleshooting) and leaves references/reliability_principles.md unreferenced.	4 / 5
	Total	17 / 20 Passed

Description

92%Weight 40%Scale 1-5

Based on the skill's description, can an agent find and select it at the right time? Clear, specific descriptions lead to better discovery.

A strong, specific third-person description that crisply states capabilities and gives an explicit 'Use for ...' trigger clause. It is comprehensive and low-conflict; only minor natural-term synonyms are missing.

Dimension	Reasoning	Score
Specificity	Names multiple concrete actions — 'Train or fine-tune language and vision models', 'GGUF conversion', 'dataset preparation and validation', 'hardware selection', 'cost estimation', 'Trackio monitoring', 'Hub authentication', 'model persistence' — giving comprehensive coverage of what the skill does.	5 / 5
Completeness	Explicitly answers both what ('Train or fine-tune ... using TRL or Unsloth with Hugging Face Jobs') and when ('Use for tasks involving cloud GPU training, GGUF conversion, or when users mention training on Hugging Face Jobs without local GPU setup') with concrete trigger phrases.	5 / 5
Trigger Term Quality	Good coverage of natural terms ('cloud GPU training', 'GGUF conversion', 'fine-tune', 'training on Hugging Face Jobs') with synonyms (TRL, Unsloth), but a few natural variants users might say (LoRA, quantize, Ollama, .gguf) are absent.	4 / 5
Distinctiveness Conflict Risk	A clear niche (TRL/Unsloth training on Hugging Face Jobs) with distinct, specific triggers and minimal overlap with generic training or document skills.	5 / 5
	Total	19 / 20 Passed

Validation

93%

Warnings & errors only

Checks the skill against the spec for correct structure and formatting. All validation checks must pass before discovery and implementation can be scored.

Validation — 15 / 16 Passed

Validation for skill structure

Criteria	Description	Result
skill_md_line_count	SKILL.md is long (739 lines); consider splitting into references/ and linking	Warning

	Total	15 / 16 Passed

Repository: huggingface/skills
Path: skills/huggingface-llm-trainer/SKILL.md
Commit: f3186ef

Reviewed: 1 day ago

Table of Contents

Discovery Implementation Validation

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.