Gpu Communication Library In Meta

Presenter(s): James Hongyi Zeng, Senior Engineering Manager,

In this video, we break down NCCL (

Benjamin Glick Pouya Kousha, Arnav Goel (

Want to scale beyond the limits of a single

In this AI Research Roundup episode, Alex discusses the paper: 'Collective

Check out IBM Cloud for

Code and Slides: https://github.com/cuda-mode/lectures/tree/main/lecture_017.

RDMA (Remote Direct Memory Access) is the secret sauce behind fast

RSC is also estimated to be 9x faster, at running the

AI clusters are difficult to manage. There are multiple hardware and software elements to coordinate and constant updates that ...

Scaling beyond a single

What is CUDA? And how does parallel computing on the

NCCL: High-Speed Inter-GPU Communication for Large-Scale Training - Sylvain Jeaugey, NVIDIA

https://arxiv.org/pdf/2510.20171 The NCCLX collective

Zhiyi Hu, Siyuan Shen, Tommaso Bonato (ETH Zurich), Sylvain Jeaugey (

Sharing is Caring – Fractional

ML Performance research paper reading group session 1 meeting (2024/11/29). This was an intro session covering prerequisite ...

Speaker: Jeff Hammond.