Skip to main content

Categories Trending Langs Tags Collections Compare

Product

Categories
Trending
Tags
Collections
New Repos
Comparisons

Resources

Submit a Repo
About
Newsletter
Privacy Policy
Terms
Contact

Popular Tags

#react
#typescript
#python
#machine-learning
#nextjs

sourcevana

Discover, Download, Deploy — Open Source Made Easy.

© 2026 Sourcevana. Made for developers.

deepspeedai/DeepSpeed — Its purpose is to efficiently train and deploy very lar | Sourcevana

HomeML Training & InferenceDeepSpeed

DeepSpeed

Its purpose is to efficiently train and deploy very large deep learning models across multiple GPUs.

Verified Snapshot (207.4 MB)

Python42.4K4.8KUpdated 16d agoFeatured

Quick Overview

What is this?

DeepSpeed is a Python library that scales deep learning training and inference for large models by handling distributed execution. It offers features like ZeRO++ for memory optimization and SuperOffload for CPU offloading, which users integrate into their PyTorch training scripts after installing via `pip install deepspeed`.

What problem does it solve?

Its purpose is to efficiently train and deploy very large deep learning models across multiple GPUs.

Who should use it?

It's for machine learning practitioners and researchers working with or building large-scale deep learning models that need distributed training and inference.

Setup difficulty:Medium

Pros

Provides ZeRO++ for significantly reducing GPU memory usage during training.
Includes SuperOffload to efficiently offload computation and model parameters to CPU memory.

Handles data, model, and pipeline parallelism strategies for models with billions of parameters.

Cons

Adds significant configuration and debugging complexity for distributed training setups.
Requires a PyTorch backend, limiting direct use with other deep learning frameworks.

Scores

Trust Score

75

Star reputation (15%)93

Star velocity 7d (15%)0

Commit recency (15%)80

Fork ratio (10%)38

Issue ratio (10%)70

Contributor signal (10%)100

README quality (5%)100

License (5%)100

Homepage/demo (5%)100

Docs URL (5%)0

Topic count (5%)100

Maintenance

68

Commit frequency95

Issue management25

Documentation85

Popularity

67

Stars100

Forks100

Growth trend10

Star History

Snapshot Versions

Version	Commit	Size	Downloads	Date
latestLatest	HEAD	207.4 MB	4	1mo ago

Alternatives

tensorflow

tensorflow

An Open Source Machine Learning Framework for Everyone

C++195.4K75.3K15d ago

stable-diffusion-webui

AUTOMATIC1111

Stable Diffusion web UI

Python163.4K30.4K3mo ago

transformers

huggingface

🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.

Python161.2K33.4K15d ago

pytorch

pytorch

Tensors and Dynamic neural networks in Python with strong GPU acceleration

Python100.3K27.9K15d ago

LLMs-from-scratch

rasbt

Implement a ChatGPT-like LLM in PyTorch from scratch, step by step

Jupyter Notebook96.5K14.8K15d ago

opencv

opencv

Open Source Computer Vision Library

C++87.7K56.6K15d ago

Community Reviews

Loading reviews...

README

Repository image

Repository image

Latest News

[2026/03] DeepSpeed Team gave a tutorial at ASPLOS 2026 titled "Building Efficient Large-Scale Model Systems with DeepSpeed: From Open-Source Foundations to Emerging Research"
[2026/03] Our SuperOffload work received an Honorable Mention for the ASPLOS 2026 Best Paper Award
[2025/12] DeepSpeed Core API updates: PyTorch-style backward and low-precision master states
[2025/11] DeepSpeed ZeRO++ powers large-scale distillation training of LLMs for Recommendation Systems at LinkedIn
[2025/10] We hosted the Ray x DeepSpeed Meetup at Anyscale. We shared our most recent work on SuperOffload, ZenFlow, Muon Optimizer Support, Arctic Long Sequence Training and DeepCompile. Please find the meetup slides here.
[2025/10] SuperOffload: Unleashing the Power of Large-Scale LLM Training on Superchips

Trust Score

75

Sourcevana Trust · 75/100

License

Apache-2.0

Languages

Python74.1%

C++17.0%

Cuda8.0%

Other0.8%

Topics

billion-parameters compression data-parallelism deep-learning gpu inference machine-learning mixture-of-experts model-parallelism pipeline-parallelism pytorch trillion-parameters+1 more

AddedMay 4, 2026

UpdatedJun 18, 2026

Last commit16d ago

Browse more in ML Training & Inference View all repos by deepspeedai

Embed Trust Badge

README.md preview:

Sourcevana Trust · 75/100

[![Sourcevana Trust](https://sourcevana.com/api/badge/deepspeedai-deepspeed)](https://sourcevana.com/repo/deepspeedai-deepspeed)

Paste this into your README.md

Embed Widget

<iframe src="https://sourcevana.com/embed/deepspeedai-deepspeed" width="480" height="120" frameborder="0"></iframe>

Embed this repo card on any website or blog