TacoSkill LABTacoSkill LAB

The full-lifecycle AI skills platform.

DiscordFeedback

Product

  • SkillHub
  • Playground
  • Skill Create
  • SkillKit

Resources

  • Help Center
  • Privacy
  • Terms
  • About

Platforms

  • Claude Code
  • Cursor
  • Codex CLI
  • Gemini CLI
  • OpenCode

© 2026 TacoSkill LAB. All rights reserved.

TacoSkill LAB
TacoSkill LAB
HomeSkillHubCreatePlaygroundLeaderboardSkillKit
  1. Home
  2. /
  3. SkillHub
  4. /
  5. pytorch-fsdp
Improve

pytorch-fsdp

7.5

by davila7

0Views
129Favorites
173Upvotes
0Downvotes

Expert guidance for Fully Sharded Data Parallel training with PyTorch FSDP - parameter sharding, mixed precision, CPU offloading, FSDP2

distributed training

7.5

Rating

0

Installs

Machine Learning

Category

Quick Review

This skill provides comprehensive PyTorch FSDP guidance extracted from official documentation. The description clearly identifies when to use the skill (FSDP-related tasks), and the content includes detailed API documentation, initialization patterns, distributed communication primitives, and debugging guidance. The structure is reasonable with quick reference patterns, though the SKILL.md is somewhat cluttered with large documentation blocks that could be better organized into separate reference files. The skill offers moderate novelty - while it consolidates scattered FSDP documentation, much of this information could be retrieved through standard documentation queries, though having it pre-organized does save tokens and time for complex FSDP setups involving parameter sharding, mixed precision, and CPU offloading.

LLM Signals

Description coverage7
Task knowledge8
Structure7
Novelty6

GitHub Signals

18,073
1,635
132
71
Last commit 0 days ago

Publisher

davila7 logo
davila7

Skill Author

Related Skills

ml-pipelinepyvene-interventionsnnsight-remote-interpretability

Loading SKILL.md…

Try onlineView on GitHub

Publisher

davila7 logo
davila7

Skill Author

Related Skills

ml-pipeline

Jeffallan

6.4

pyvene-interventions

zechenzhangAGI

7.6

nnsight-remote-interpretability

zechenzhangAGI

7.0

mlflow

zechenzhangAGI

7.6
Try online