deepspeed
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
npx skills add Orchestra-Research/AI-Research-SKILLs --skill deepspeed --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
What it does
Provides comprehensive guidance for developing with DeepSpeed, focusing on distributed training concepts, ZeRO optimization stages, pipeline parallelism, mixed precision (FP16/BF16/FP8), 1-bit Adam, and sparse attention. It targets tasks involving working with DeepSpeed features, APIs, implementing DeepSpeed solutions, debugging related code, and learning best practices.
How it works
The skill outlines when to use it (triggered by working with DeepSpeed, asking about DeepSpeed features or APIs, implementing, debugging, or learning best practices). It then provides a structured reference and examples that illustrate DeepNVMe/I/O patterns and DeepSpeed APIs, including handle creation (aio_handle and gds_handle), blocking and non-blocking file writes, parallel I/O settings, pinning tensors, and tuning utilities (ds_nvme_tune). It enumerates APIs such as async_pread, sync_pwrite, async_pwrite, wait, new_pinned_device_tensor, pin_device_tensor, unpin_device_tensor, and related handle parameters (block_size, queue_depth, single_submit, overlap_events, intra_op_parallelism), and shows usage snippets and expected behaviors. The content is framed as practical instructions and code-style examples for performing I/O and tensor pinning in the context of DeepNVMe and DeepSpeed integration. It references the ds_nvme_tune tool and describes how to configure handles and interpret results to optimize throughput.
When to use it
Use this skill when you are:
- Working with deepspeed
- Asking about deepspeed features or APIs
- Implementing deepspeed solutions
- Debugging deepspeed code
- Learning deepspeed best practices
What it can touch
The skill references and implies use of external tools and libraries such as "deepspeed", "torch", "transformers", and "accelerate". It mentions specific APIs and classes like AsyncIOBuilder, aio_handle, GDSBuilder, gds_handle, and ds_nvme_tune, and describes functions (async_pread, sync_pwrite, wait, pin_device_tensor, etc.). However, it does not provide executable commands beyond illustrative Python code snippets and command-line examples. The interaction points are through the listed APIs and the ds_nvme_tune utility.
Caveats
The content is examples-driven and includes inline code snippets that assume appropriate environment setup (e.g., NVMe storage, libaio, CUDA tensors). It references performance tuning parameters and default values for aio_handle and gds_handle that are environment-sensitive. No guarantees of outcomes are stated beyond illustrative results in the snippets. The skill requires the listed dependencies (deepspeed, torch, transformers, accelerate) to be present.
# Deepspeed Skill Comprehensive assistance with deepspeed development, generated from official documentation. ## When to Use This Skill This skill should be triggered when: - Working with deepspeed - Asking about deepspeed features or APIs - Implementing deepspeed solutions - Debugging deepspeed code - Learning deepspeed best practices ## Quick Reference ### Common Patterns **Pattern 1:** DeepNVMe Contents Requirements Creating DeepNVMe Handles Using DeepNVMe Handles Blocking File Write Non-Blocking File Write Parallel File Write Pinned Tensors Putting it together Acknowledgements Appendix Advanced Handle Creation Performance Tuning DeepNVMe APIs General I/O APIs GDS-specific APIs Handle Settings APIs This tutorial will show how to use DeepNVMe for data transfers between persistent storage and tensors residing in host or device memory. DeepNVMe improves the performance and efficiency of I/O operations in Deep Learning applications through powerful optimizations built on Non-Volatile Memory Express (NVMe) Solid State Drives (SSDs), Linux Asynchronous I/O (libaio), and NVIDIA Magnum IOTM GPUDirect® Storage (GDS). Requirements Ensure your environment is properly configured to use Deep
- When to Use This Skill
- Quick Reference
- Common Patterns
- Example Code Patterns
- Reference Files
- Working with This Skill
- For Beginners
- For Specific Features
- For Code Examples
- Resources
- references/
- scripts/
- assets/
- Notes
What does the deepspeed skill do?
Expert guidance for distributed training with DeepSpeed - ZeRO optimization stages, pipeline parallelism, FP16/BF16/FP8, 1-bit Adam, sparse attention
How do I install it?
Run `npx skills add Orchestra-Research/AI-Research-SKILLs --skill deepspeed --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From Orchestra-Research/AI-Research-SKILLs, a repository with 11,391 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.