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

pytorch-fsdp

6.4

by zechenzhangAGI

76Favorites
117Upvotes
0Downvotes

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

distributed training

6.4

Rating

0

Installs

Machine Learning

Category

Quick Review

This skill provides comprehensive PyTorch FSDP documentation and guidance extracted from official sources. The description clearly indicates it offers expert guidance for FSDP features (parameter sharding, mixed precision, CPU offloading, FSDP2), making it actionable for a CLI agent. Task knowledge is strong, with extensive reference material covering initialization, configuration patterns, and distributed training concepts, though the patterns shown are primarily general torch.distributed examples rather than FSDP-specific. Structure is logical with clear organization of reference files and quick patterns, though SKILL.md could be more concise given it contains lengthy torch.distributed documentation that could be better summarized with references to detailed files. Novelty is moderate - while FSDP documentation consolidation is useful, much content is standard distributed training material that overlaps with general PyTorch knowledge, and the patterns shown don't demonstrate particularly complex FSDP-specific configurations that would be difficult for an agent to construct independently.

LLM Signals

Description coverage7
Task knowledge8
Structure7
Novelty6

GitHub Signals

891
74
19
2
Last commit 0 days ago

Publisher

zechenzhangAGI

zechenzhangAGI

Skill Author

Related Skills

ml-pipelinesparse-autoencoder-traininghuggingface-accelerate

Loading SKILL.md…

Try onlineView on GitHub

Publisher

zechenzhangAGI avatar
zechenzhangAGI

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