huggingface-accelerate

Simplest distributed training API. 4 lines to add distributed support to any PyTorch script. Unified API for DeepSpeed/FSDP/Megatron/DDP. Automatic device placement, mixed precision (FP16/BF16/FP8). Interactive config, single launch command. HuggingFace ecosystem standard.

Quality

73%

Does it follow best practices?

Run evals on this skill

Adds up to 20 points to the overall score

View guide

Securityby

Low

Low-risk findings worth noting

Fix and improve this skill with Tessl

tessl review fix ./backend/cli/skills/ml-training/accelerate/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 real reference files, but it is somewhat redundant and lacks explicit validation checkpoints for risky distributed-training operations.

Suggestions

Add explicit verification checkpoints to the distributed launch and checkpoint workflows (e.g., 'verify all processes joined' / 'confirm checkpoint loaded on every rank before training resumes'), which would lift workflow_clarity above 3.

Consolidate the redundant Quick-start 'Convert PyTorch script' example with Workflow 1's before/after to remove repetition and tighten conciseness.

Trim or remove the 'Key advantages' list, which restates points already in the description and frontmatter.

Dimension	Reasoning	Score
Conciseness	Mostly efficient and code-forward, but the ~330-line body repeats material (the Quick-start 'Convert PyTorch script' before/after largely restates Workflow 1, and 'Key advantages' restate the description), so it could be tightened.	3 / 5
Actionability	Fully executable, copy-paste-ready code across prepare/backward, mixed precision, DeepSpeed, FSDP, and gradient accumulation, plus concrete per-setup launch commands covering the common cases.	5 / 5
Workflow Clarity	Workflows are clearly sequenced (install → convert → config → launch) but lack explicit validation/verification checkpoints for risky distributed/batch operations (e.g., confirming all ranks initialized, verifying checkpoint loaded), which caps workflow clarity at 3 per the feedback-loops note.	3 / 5
Progressive Disclosure	Clear overview in SKILL.md with well-signaled, one-level-deep references to three real files (megatron-integration.md, custom-plugins.md, performance.md), with content appropriately split and easy to navigate.	5 / 5
	Total	16 / 20 Passed

Description

75%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 specific and distinct with strong concrete capabilities and good trigger keywords, but it omits any explicit 'Use when...' guidance, which caps its completeness at 3.

Suggestions

Add an explicit 'Use when...' clause naming natural trigger phrases (e.g., 'Use when the user asks to distribute a PyTorch training script across multiple GPUs/nodes or mentions DeepSpeed, FSDP, DDP, or mixed precision').

Include common synonyms/extensions users might say (e.g., 'multi-GPU', 'multi-node', 'data parallel', 'sharding') to broaden trigger coverage from 4 to 5.

Drop or tighten 'HuggingFace ecosystem standard', which reads as borderline buzzword padding rather than a capability.

Dimension	Reasoning	Score
Specificity	Lists multiple concrete capabilities — 'add distributed support to any PyTorch script', 'Automatic device placement', 'mixed precision (FP16/BF16/FP8)', 'Interactive config, single launch command' — giving comprehensive coverage of specific actions.	5 / 5
Completeness	Provides a strong, multi-part 'what' but includes no 'Use when...' trigger clause, so per the rubric guideline a missing explicit 'when' caps completeness at 3.	3 / 5
Trigger Term Quality	Contains natural terms users say ('distributed training', 'PyTorch script', 'DeepSpeed', 'FSDP', 'mixed precision', 'FP16/BF16/FP8'), but lacks broader synonyms/extensions beyond the technical framework names.	4 / 5
Distinctiveness Conflict Risk	Targets a clear niche (HuggingFace Accelerate with named frameworks DeepSpeed/FSDP/Megatron/DDP) with distinct triggers and minimal overlap risk with other skills.	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/accelerate/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.