TacoSkill LAB
TacoSkill LAB
HomeSkillHubCreatePlaygroundSkillKit
© 2026 TacoSkill LAB
AboutPrivacyTerms
  1. Home
  2. /
  3. SkillHub
  4. /
  5. pytorch-fsdp
Improve

pytorch-fsdp

7.5

by davila7

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

davila7

Skill Author

Related Skills

ml-pipelinesparse-autoencoder-traininghuggingface-accelerate

Loading SKILL.md…

Try onlineView on GitHub

Publisher

davila7 avatar
davila7

Skill Author

Related Skills

ml-pipeline

Jeffallan

6.4

sparse-autoencoder-training

zechenzhangAGI

7.6

huggingface-accelerate

zechenzhangAGI

7.6

moe-training

zechenzhangAGI

7.6
Try online