pytorch-lightning

High-level PyTorch framework with Trainer class, automatic distributed training (DDP/FSDP/DeepSpeed), callbacks system, and minimal boilerplate. Scales from laptop to supercomputer with same code. Use when you want clean training loops with built-in best practices.

Quality

75%

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

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-training/pytorch-lightning/SKILL.md

Quality

Content

72%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.

The body is highly actionable and well-structured with clean progressive disclosure to verified reference files. Its weaknesses are repeated boilerplate code across workflows and missing explicit validation checkpoints in the training workflows.

Suggestions

Deduplicate the LightningModule definition: define it once in Quick start and reuse/reference it in Workflows 1, 2, and 5 instead of re-pasting the training_step each time.

Add explicit validation checkpoints to the training workflows (e.g., after trainer.fit, instruct to confirm logs/metrics are being written and that loss is decreasing before declaring success).

Trim or move the 'Key advantages' and 'Resources' bullets that restate marketing-style claims (e.g., '1M+ downloads/month') to reduce token overhead.

Dimension	Reasoning	Score
Conciseness	Prose is lean, but the LightningModule/training_step pattern is redefined nearly verbatim across the Quick start and Workflows 1, 2, and 5, which is redundant and could be tightened by referencing the canonical definition.	3 / 5
Actionability	Provides fully executable, copy-paste-ready code covering the common cases (conversion, validation/testing, DDP, callbacks, LR scheduling) plus concrete troubleshooting snippets.	5 / 5
Workflow Clarity	The Quick start gives a clear 3-step sequence and workflows are ordered, but validation checkpoints are implicit rather than explicit (e.g., no 'verify logs/training started before proceeding' gate), and trainer.fit is a long batch operation lacking a verify step.	3 / 5
Progressive Disclosure	Clear overview with well-signaled one-level-deep references (callbacks.md, distributed.md, hyperparameter-tuning.md) that are real bundle files, each briefly described; content is appropriately split and easy to navigate.	5 / 5
	Total	16 / 20 Passed

Description

78%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.

The description is concrete and distinct, clearly answering both what the skill does and when to use it with named framework features. Its main weakness is the 'when' trigger being a single soft phrase rather than multiple concrete user scenarios.

Dimension	Reasoning	Score
Specificity	Lists several concrete capabilities (Trainer class, DDP/FSDP/DeepSpeed, callbacks, scaling laptop-to-supercomputer) rather than vague language, though it enumerates features more than a comprehensive list of discrete actions.	4 / 5
Completeness	Explicitly provides both the 'what' (high-level PyTorch framework with Trainer, distributed, callbacks) and a 'when' clause ('Use when you want clean training loops...'), but the trigger could enumerate more concrete user scenarios.	4 / 5
Trigger Term Quality	Includes natural phrasing ('clean training loops with built-in best practices') plus domain keywords (DDP/FSDP/DeepSpeed, callbacks), but misses common user variations like 'training script' or 'train a model'.	4 / 5
Distinctiveness Conflict Risk	Names the specific framework and strategies (PyTorch Lightning, DDP/FSDP/DeepSpeed, Trainer), carving a clear niche with minimal overlap risk against raw PyTorch or Accelerate.	5 / 5
	Total	17 / 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: synthetic-sciences/openscience
Path: backend/cli/skills/ml-training/pytorch-lightning/SKILL.md
Commit: fcd2e11

Reviewed: 2 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.