pytorch-fsdp

1---2name: pytorch-fsdp3description: Fully sharded data-parallel training for large models.4version: 1.0.05author: Orchestra Research6license: MIT7dependencies: [torch>=2.0, transformers]8platforms: [linux, macos]9metadata:10  hermes:11    tags: [Distributed Training, PyTorch, FSDP, Data Parallel, Sharding, Mixed Precision, CPU Offloading, FSDP2, Large-Scale Training]12 13---14 15# Pytorch-Fsdp Skill16 17Assistance with pytorch-fsdp development, generated from official documentation.18 19## When to Use This Skill20 21This skill should be triggered when:22- Working with pytorch-fsdp23- Asking about pytorch-fsdp features or APIs24- Implementing pytorch-fsdp solutions25- Debugging pytorch-fsdp code26- Learning pytorch-fsdp best practices27 28## Quick Reference29 30The full common-patterns catalog (~157k chars of runnable FSDP snippets) lives in31`references/common-patterns.md` — load it with `read_file` when you need wrapping,32sharding-strategy, checkpoint, or mixed-precision examples. Start there rather than33reconstructing FSDP incantations from memory.34 35## Reference Files36 37This skill includes comprehensive documentation in `references/`:38 39- **other.md** - Other documentation40 41Use `view` to read specific reference files when detailed information is needed.42 43## Working with This Skill44 45### For Beginners46Start with the getting_started or tutorials reference files for foundational concepts.47 48### For Specific Features49Use the appropriate category reference file (api, guides, etc.) for detailed information.50 51### For Code Examples52The quick reference section above contains common patterns extracted from the official docs.53 54## Resources55 56### references/57Organized documentation extracted from official sources. These files contain:58- Detailed explanations59- Code examples with language annotations60- Links to original documentation61- Table of contents for quick navigation62 63### scripts/64Add helper scripts here for common automation tasks.65 66### assets/67Add templates, boilerplate, or example projects here.68 69## Notes70 71- This skill was automatically generated from official documentation72- Reference files preserve the structure and examples from source docs73- Code examples include language detection for better syntax highlighting74- Quick reference patterns are extracted from common usage examples in the docs75 76## Updating77 78To refresh this skill with updated documentation:791. Re-run the scraper with the same configuration802. The skill will be rebuilt with the latest information81 82 83