DeepSeed Files

Deep learning optimization library making distributed training easy

This is an exact mirror of the DeepSeed project, hosted at https://github.com/microsoft/DeepSpeed. SourceForge is not affiliated with DeepSeed. For more information, see the SourceForge Open Source Mirror Directory.

The interactive file manager requires Javascript. Please enable it or use sftp or scp.
You may still browse the files here.

Name	Modified	Size	InfoDownloads / Week
Parent folder
README.md	2025-11-04	1.4 kB	0
v0.18.2 Patch Release source code.tar.gz	2025-11-04	215.2 MB	1
v0.18.2 Patch Release source code.zip	2025-11-04	216.3 MB	1
Totals: 3 Items		431.5 MB	2

What's Changed

Update version after 0.18.1 release by @loadams in https://github.com/deepspeedai/DeepSpeed/pull/7647
Deduplicate fp32 weights under torch autocast and ZeRO3 by @eternalNight in https://github.com/deepspeedai/DeepSpeed/pull/7651
ulysses mpu: additional api by @stas00 in https://github.com/deepspeedai/DeepSpeed/pull/7649
ALST/UlyssesSP: more intuitive API wrt variable seqlen by @stas00 in https://github.com/deepspeedai/DeepSpeed/pull/7656
Fix misplaced overflow handling return in fused_optimizer.py by @rraminen in https://github.com/deepspeedai/DeepSpeed/pull/7645
[bug]: fixed comm_dtype in extra_large_param_to_reduce by @therealnaveenkamal in https://github.com/deepspeedai/DeepSpeed/pull/7660
UlyssesSP: TiledMLP doc - recomputes forward twice by @stas00 in https://github.com/deepspeedai/DeepSpeed/pull/7664
resolved a 0-dim tensor slicing bug from _get_state_without_padding by @therealnaveenkamal in https://github.com/deepspeedai/DeepSpeed/pull/7659
Fix typo in pytorch-profiler.md documentation by @kunheek in https://github.com/deepspeedai/DeepSpeed/pull/7652
README refresh by @sfc-gh-truwase in https://github.com/deepspeedai/DeepSpeed/pull/7668

New Contributors

@kunheek made their first contribution in https://github.com/deepspeedai/DeepSpeed/pull/7652

Full Changelog: https://github.com/deepspeedai/DeepSpeed/compare/v0.18.1...v0.18.2

Source: README.md, updated 2025-11-04

Other Useful Business Software

Gen AI apps are built with MongoDB Atlas Icon

Gen AI apps are built with MongoDB Atlas

The database for AI-powered applications.

MongoDB Atlas is the developer-friendly database used to build, scale, and run gen AI and LLM-powered apps—without needing a separate vector database. Atlas offers built-in vector search, global availability across 115+ regions, and flexible document modeling. Start building AI apps faster, all in one place.

Start Free

Level Up Your Cyber Defense with External Threat Management Icon

Level Up Your Cyber Defense with External Threat Management

See every risk before it hits. From exposed data to dark web chatter. All in one unified view.

Move beyond alerts. Gain full visibility, context, and control over your external attack surface to stay ahead of every threat.

Try for Free

Recommended Projects

DeepSpeed
DeepSpeed is an easy-to-use deep learning optimization software suite that enables unprecedented scale and speed for Deep Learning Training and Inference. With DeepSpeed you can: 1. Train/Inference dense or sparse models with billions or trillions of parameters 2. Achieve excellent system...
Ludwig AI
Declarative deep learning framework built for scale and efficiency. Ludwig is a low-code framework for building custom AI models like LLMs and other deep neural networks. Declarative YAML configuration file is all you need to train a state-of-the-art LLM on your data. Support for multi-task and...
Megatron
Megatron is a large, powerful transformer developed by the Applied Deep Learning Research team at NVIDIA. This repository is for ongoing research on training large transformer language models at scale. We developed efficient, model-parallel (tensor, sequence, and pipeline), and multi-node...
DeepSpeed MII
MII makes low-latency and high-throughput inference possible, powered by DeepSpeed. The Deep Learning (DL) open-source community has seen tremendous growth in the last few months. Incredibly powerful text generation models such as the Bloom 176B, or image generation model such as Stable...
Ray
Modern workloads like deep learning and hyperparameter tuning are compute-intensive and require distributed or parallel execution. Ray makes it effortless to parallelize single machine code — go from a single CPU to multi-core, multi-GPU or multi-node with minimal code changes. Accelerate your...