DeepSpeed
@DeepSpeedAI
Official account for DeepSpeed, a library that enables unprecedented scale and speed for deep learning training + inference. 日本語 : @DeepSpeedAI_JP
You might like
UIUC, AnyScale, and Snowflake significantly enhanced LLM offloading for the Superchip era!
🚀 SuperOffload: Unleashing the Power of Large-Scale LLM Training on Superchips Superchips like the NVIDIA GH200 offer tightly coupled GPU-CPU architectures for AI workloads. But most existing offloading techniques were designed for traditional PCIe-based systems. Are we truly…
🚨Meetup Alert🚨 Join us for @raydistributed × @DeepSpeedAI Meetup: AI at Scale, including talks from researchers and engineers at @LinkedIn, @anyscalecompute and @Snowflake. Learn how leading AI teams are scaling efficiently with Ray’s distributed framework and DeepSpeed’s…
Step into the future of AI at #PyTorchCon 2025, Oct 22–23 in San Francisco 🔥 Join the DeepSpeed keynote and technical talks. Register: events.linuxfoundation.org/pytorch-confer… + Oct 21 co-located events: Measuring Intelligence, Open Agent & AI Infra Summits / Startup Showcase & PyTorch Training
The @DeepSpeedAI would like to thank @modal for sponsoring our gpus for CI. This is an amazing contribution to our AI-democratizing open source project. github.com/deepspeedai/De… The Modal team is outstanding in their amazing support - speed, expertise and a human experience!
ZenFlow is a massive improvement to DeepSpeed Offloading. Courtesy of an excellent collaboration among University of Virginia, UC Merced, Argonne National Laboratory, Microsoft, and Snowflake.
Introducing #ZenFlow: No Compromising Speed for #LLM Training w/ Offloading 5× faster LLM training with offloading 85% less GPU stalls 2× lower I/O overhead 🚀 Blog: hubs.la/Q03DJ6GJ0 🚀 Try ZenFlow and experience 5× faster training with offloading: hubs.la/Q03DJ6Vb0
Kudos to Xinyu for giving an excellent presentation of DeepSpeed Universal Checkpointing (UCP) paper at USENIX ATC 2015.
📢 Yesterday at USENIX ATC 2025, Xinyu Lian from UIUC SSAIL Lab presented our paper on Universal Checkpointing (UCP). UCP is a new distributed checkpointing system designed for today's large-scale DNN training, where models often use complex forms of parallelism, including data,…
My first project at @Snowflake AI Research is complete! I present to you Arctic Long Sequence Training (ALST) Paper: arxiv.org/abs/2506.13996 Blog: snowflake.com/en/engineering… ALST is a set of modular, open-source techniques that enable training on sequences up to 15 million…
Improved DeepNVMe: Affordable I/O Scaling for AI - Faster I/O with PCIe Gen5 - 20x faster model checkpointing - Low-budget SGLang inference via NVMe offloading - Pinned memory for CPU-only workloads - Zero-copy tensor type casting Blog: tinyurl.com/yanbrjy9
PyTorch Foundation has expanded into an umbrella foundation. @vllm_project and @DeepSpeedAI have been accepted as hosted projects, advancing community-driven AI across the full lifecycle. Supporting quotes provided by the following members: @AMD, @Arm, @AWS, @Google, @Huawei,…
Come hear all the exciting DeepSpeed updates at the upcoming PyTorch Day France 2025 DeepSpeed – Efficient Training Scalability for Deep Learning Models - sched.co/21nyy @sched
Introducing 🚀DeepCompile🚀: compiler-based distributed training optimizations. - Automatic parallelization & profile-guided optimizations - Enable ZeRO1, ZeRO3, Offloading, etc. via compiler passes - 1.2X-7X speedups over manual ZeRO1/ZeRO3/Offloading tinyurl.com/8cys28xk
AutoTP + ZeRO Training for HF Models - Enhance HF post-training with larger models, batches, & contexts - 4x faster LLAMA3 fine-tuning with TP=2 vs TP=1 - No code changes needed Blog: tinyurl.com/5n8nfs2w
1/4⚡️nanoton now supports DoMiNo with intra-layer communication overlapping, achieving 60% communication hiding for tensor parallelism (TP) in both the forward and backward passes while maintaining the same training loss.
🚀 Excited to introduce DeepSpeed, a deep learning optimization library from @Microsoft! It simplifies distributed training and inference, making AI scaling more efficient and cost-effective. Learn more 👉 hubs.la/Q0351DJC0 #DeepSpeed #AI #OpenSource #LFAIData
🚀Introducing Ulysses-Offload🚀 - Unlock the power of long context LLM training and finetuning with our latest system optimizations - Train LLaMA3-8B on 2M tokens context using 4xA100-80GB - Achieve over 55% MFU Blog: shorturl.at/Spx6Y Tutorial: shorturl.at/bAWu5
Introducing Domino: a novel zero-cost communication tensor parallelism (TP) training engine for both single node and multi-node settings. - Near-complete communication hiding - Novel multi-node scalable TP solution Blog: github.com/microsoft/Deep…
Great to see the amazing DeepSpeed optimizations from @Guanhua_Wang_, Heyang Qin, @toh_tana, @QuentinAnthon15, and @samadejacobs presented by @ammar_awan at MUG '24.
Dr. Ammar Ahmad Awan from Microsoft DeepSpeed giving a presentation at MUG '24 over Trillion-parameter LLMs and optimization with MVAPICH. @OSUengineering @Microsoft @OhTechCo @mvapich @MSFTDeepSpeed @MSFTDeepSpeedJP #MUG24 #MPI #AI #LLM #DeepSpeed
Announcing that DeepSpeed now runs natively on Windows. This exciting combination unlocks DeepSpeed optimizations to Windows users and empowers more people and organizations with AI innovations. - HF Inference & Finetuning - LoRA - CPU Offload Blog: shorturl.at/a7TF8
💡Check out Comet’s latest integration with DeepSpeed, a deep learning optimization library! 🤝With the @MSFTDeepSpeed + @Cometml integration automatically start logging training metrics generated by DeepSpeed. Try the quick-start Colab to get started: colab.research.google.com/github/comet-m…
Introducing DeepNVMe, a suite of optimizations for fast and efficient I/O operations in DL applications. - POSIX-style APIs - Direct HBM/NVMe xfers via NVIDIA GDS - Cheap Inference scaling via NVMe-Offload Blog: shorturl.at/l7Oue @Azure @NVIDIADC #FMS24 #GPUDirect
United States Trends
- 1. D’Angelo 39.3K posts
- 2. Happy Birthday Charlie 93.8K posts
- 3. Brown Sugar 3,498 posts
- 4. #BornOfStarlightHeeseung 63.6K posts
- 5. Black Messiah 1,012 posts
- 6. #csm217 2,227 posts
- 7. Alex Jones 21.3K posts
- 8. Drew Struzan N/A
- 9. #tuesdayvibe 5,370 posts
- 10. Voodoo 4,449 posts
- 11. Sandy Hook 7,147 posts
- 12. #PortfolioDay 6,064 posts
- 13. Powell 22K posts
- 14. Pentagon 86.7K posts
- 15. #NationalDessertDay N/A
- 16. George Floyd 6,599 posts
- 17. Rest in Power 6,867 posts
- 18. Cheryl Hines 1,748 posts
- 19. Monad 223K posts
- 20. Masuda 2,635 posts
You might like
-
Reka
@RekaAILabs -
lmarena.ai
@arena -
Jan Leike
@janleike -
Tri Dao
@tri_dao -
Susan Zhang
@suchenzang -
Hadi Salman
@hadisalmanX -
Harrison Chase
@hwchase17 -
Jerry Liu
@jerryjliu0 -
Wenhu Chen
@WenhuChen -
LlamaIndex 🦙
@llama_index -
Yi Tay
@YiTayML -
Georgi Gerganov
@ggerganov -
Julien Chaumond
@julien_c -
Lior Alexander
@LiorOnAI -
Zhuohan Li
@zhuohan123
Something went wrong.
Something went wrong.