pytorch-patterns

PyTorch deep learning patterns and best practices for building robust, efficient, and reproducible training pipelines, model architectures, and data loading. Use when writing or reviewing PyTorch training loops, model architectures, or data loading, or when a run will not reproduce.

Quality

74%

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 ./skills/pytorch-patterns/SKILL.md

Quality

Content

61%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 content is a well-organized, actionable catalog of PyTorch patterns with strong copy-paste-ready examples. Its weaknesses are mild duplication that inflates length, illustrative (non-runnable) snippets in places, and the absence of any sequenced workflow or validation guidance. No external references are used, which is appropriate but leaves the file monolithic.

Suggestions

Deduplicate mixed-precision/AMP guidance (present in both the training loop and a standalone Performance section) and consolidate repeated optimizer.zero_grad(set_to_none=True) commentary to tighten conciseness.

Make illustrative snippets fully runnable by defining referenced symbols (e.g., MyModel, self.block1/block2, self.features) or clearly mark them as fragmentary to raise actionability.

For destructive/batch-adjacent operations like checkpoint save/load, add an explicit validate-then-proceed sequence (e.g., load -> verify keys match -> step) so the skill carries a concrete workflow with checkpoints.

Dimension	Reasoning	Score
Conciseness	The body is code-heavy and largely lean, but mixed-precision/AMP appears twice, optimizer.zero_grad(set_to_none=True) is repeated, and the verbose 'Bad' blocks plus the Quick Reference table restate inline guidance. It is mostly efficient with clear duplication that could be trimmed, fitting anchor 3 rather than the tighter anchor 4.	3 / 5
Actionability	Abundant copy-paste-ready functions (train_one_epoch, evaluate, ImageDataset, save_checkpoint, collate_fn, set_seed) cover common cases, but several snippets are illustrative with undefined symbols (MyModel(), self.block1, fragmentary self.features). It is mostly executable with minor gaps, fitting anchor 4 rather than the fully self-contained anchor 5.	4 / 5
Workflow Clarity	The skill is a categorized pattern reference rather than a sequenced workflow, with no validation checkpoints or feedback loops; the simple-skill exception does not apply since it is multi-topic and well over 50 lines. Structure is present but explicit workflow sequencing and checkpoints are missing, matching anchor 3.	3 / 5
Progressive Disclosure	No bundle files exist and all content is inline in a single ~390-line file, but section headers are clear and logically grouped (Core Principles, Model Architecture, Training Loop, Data Pipeline, etc.). It has good structure with minor organization gaps — advanced/performance sections could be split out — fitting anchor 4 rather than the reference-split anchor 5.	4 / 5
	Total	14 / 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.

The description is strong: it explicitly states both capabilities and trigger conditions with concrete, PyTorch-specific language. It is concise, third-person, and unlikely to conflict with unrelated skills. Minor keyword variants (CUDA, GPU memory) could broaden trigger coverage.

Dimension	Reasoning	Score
Specificity	Names concrete domains — 'training pipelines, model architectures, and data loading' — with several specific actions, though the framing 'robust, efficient, and reproducible' is somewhat abstract. It lists several specific capabilities with minor coverage gaps, fitting anchor 4 rather than the fully comprehensive anchor 5.	4 / 5
Completeness	It explicitly answers both what ('patterns and best practices for building... training pipelines, model architectures, and data loading') and when ('Use when writing or reviewing PyTorch training loops... or when a run will not reproduce') with concrete trigger phrases, matching anchor 5.	5 / 5
Trigger Term Quality	Natural user-facing terms appear — 'PyTorch training loops', 'model architectures', 'data loading', 'a run will not reproduce' — giving good keyword coverage. Common variants like CUDA, GPU memory, or debugging training are absent, so it sits at anchor 4 rather than 5.	4 / 5
Distinctiveness Conflict Risk	The PyTorch-specific niche with PyTorch-named triggers ('PyTorch training loops') gives a clear niche with minimal conflict risk against other skills, matching anchor 5.	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
metadata_version	'metadata.version' is missing	Warning

	Total	15 / 16 Passed

Repository: affaan-m/ECC
Path: skills/pytorch-patterns/SKILL.md
Commit: d8409a4

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.