peft-fine-tuning

Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods. Use when fine-tuning large models (7B-70B) with limited GPU memory, when you need to train <1% of parameters with minimal accuracy loss, or for multi-adapter serving. HuggingFace's official library integrated with transformers ecosystem.

Quality

83%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Passed

No findings from the security scan

Quality

Content

78%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.

A thorough, highly actionable PEFT reference with copy-paste-ready code across all common methods and well-structured progressive disclosure into reference files. Its main weakness is mild verbosity from explaining concepts Claude already knows.

Suggestions

Trim explanatory asides that restate well-known concepts (e.g., what rank and alpha control, 'capacity vs efficiency') and let inline code comments carry the minimal guidance.

Add an explicit validate-fix-retry loop for training issues (e.g., after print_trainable_parameters, branch on unexpected trainable% back to config), which would lift workflow clarity.

Consider moving the PEFT methods comparison detail or the per-architecture target_modules list into a reference file to keep SKILL.md a leaner overview.

Dimension	Reasoning	Score
Conciseness	The body is code- and table-heavy and mostly efficient, but includes some explanation of concepts Claude already knows (e.g., what rank/alpha do, 'capacity vs efficiency' commentary) and restates LoRA/QLoRA definitions that could be tightened.	3 / 5
Actionability	Provides fully executable, copy-paste-ready scripts for LoRA, QLoRA, IA3, Prefix Tuning, multi-adapter serving, merging, TRL, and vLLM with real imports and complete configs covering the common cases.	5 / 5
Workflow Clarity	The install-configure-train-save/merge-serve sequence is clear with verification checkpoints like 'model.print_trainable_parameters()' and 'Verify adapter is active', though explicit validate-fix-retry feedback loops are only lightly present.	4 / 5
Progressive Disclosure	SKILL.md serves as an overview with quick start and core methods, while advanced and troubleshooting content is split into clearly signaled, one-level-deep reference files (advanced-usage.md, troubleshooting.md) with descriptive link text.	5 / 5
	Total	17 / 20 Passed

Description

87%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 description that clearly states capabilities and provides explicit, concrete use-when triggers with quantified constraints. It is specific, well-triggered, and distinct from other skills.

Dimension	Reasoning	Score
Specificity	Names concrete methods (LoRA, QLoRA, '25+ methods') and specific actions (fine-tuning 7B-70B models, multi-adapter serving) with quantified metrics, but does not enumerate the full set of method-level actions, so it stops short of comprehensive coverage.	4 / 5
Completeness	Explicitly states both what it does ('Parameter-efficient fine-tuning for LLMs using LoRA, QLoRA, and 25+ methods') and when to use it via a concrete 'Use when...' clause with multiple specific triggers.	5 / 5
Trigger Term Quality	Includes natural user phrases such as 'fine-tuning large models', 'limited GPU memory', and 'multi-adapter serving' plus synonyms (LoRA, QLoRA), with only a few common variations missing.	4 / 5
Distinctiveness Conflict Risk	The PEFT/LoRA/QLoRA niche is sharply defined with distinct triggers, making it unlikely to be selected for unrelated skills.	5 / 5
	Total	18 / 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
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning

	Total	15 / 16 Passed

Repository: OpenRaiser/NanoResearch
Path: skills/vendor-ai-research/peft/SKILL.md
Commit: 9d3b440

Reviewed: 11 days 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.