implementing-llms-litgpt

Implements and trains LLMs using Lightning AI's LitGPT with 20+ pretrained architectures (Llama, Gemma, Phi, Qwen, Mistral). Use when need clean model implementations, educational understanding of architectures, or production fine-tuning with LoRA/QLoRA. Single-file implementations, no abstraction layers.

Quality

86%

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

72%

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

A highly actionable, well-structured skill body with good progressive disclosure into real reference files. Its weaknesses are repetition that hurts conciseness and missing validation checkpoints in the batch training workflows.

Suggestions

De-duplicate the litgpt finetune_lora flag block (shown three times) into a single canonical example, and collapse the overlapping Workflow 1 and Workflow 2 fine-tuning steps.

Add explicit validation/verification checkpoints to the training workflows (e.g., eval the checkpoint / confirm loss decreased before promoting it), not just the deploy workflow's local test step.

Trim the repeated LoRA rank guide and per-flag inline comments where they restate what the flags already name, to tighten token usage.

Dimension	Reasoning	Score
Conciseness	Mostly concrete commands rather than concept exposition, but the body is long and repeats material (the full litgpt finetune_lora flag block appears three times, and Workflow 1 overlaps Workflow 2).	2 / 3
Actionability	Throughout it gives executable bash and Python with specific flags and model names (litgpt download/finetune/pretrain, LLM.load/generate), copy-paste ready rather than pseudocode.	3 / 3
Workflow Clarity	Each workflow has a clear 4-step sequence, but the batch training workflows (fine-tune, pretrain) lack validation/verification checkpoints; per the rubric this caps workflow_clarity at 2.	2 / 3
Progressive Disclosure	Four real one-level-deep reference files (supported-models, training-recipes, distributed-training, custom-models) are clearly signaled via markdown links under Advanced topics, keeping the body an overview with detail split out.	3 / 3
	Total	10 / 12 Passed

Description

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.

A strong, well-targeted description that states concrete capabilities, gives explicit "Use when" triggers, and carves out a distinct niche. The only nit is a grammatical slip ("Use when need") that does not materially weaken the triggers.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete actions ("Implements and trains LLMs", "production fine-tuning with LoRA/QLoRA") rather than vague language, matching the multi-action anchor.	3 / 3
Completeness	Explicitly answers both what ("Implements and trains LLMs using ... LitGPT") and when ("Use when need clean model implementations, educational understanding ... or production fine-tuning").	3 / 3
Trigger Term Quality	Covers natural terms users would say (LoRA, QLoRA, fine-tuning, and named families Llama/Gemma/Phi/Qwen/Mistral), with only minor awkwardness in "educational understanding of architectures".	3 / 3
Distinctiveness Conflict Risk	LitGPT-specific niche with distinct, named-architecture and LoRA/QLoRA triggers makes it unlikely to fire for an unrelated skill.	3 / 3
	Total	12 / 12 Passed

Validation

87%

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 — 14 / 16 Passed

Validation for skill structure

Criteria	Description	Result
frontmatter_unknown_keys	Unknown frontmatter key(s) found; consider removing or moving to metadata	Warning
referenced_paths_exist	Referenced path issues: 2 missing	Warning

	Total	14 / 16 Passed

Repository: Orchestra-Research/AI-Research-SKILLs
Path: 01-model-architecture/litgpt/SKILL.md
Commit: 773a529

Reviewed: about 13 hours 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.