Data

Tools

Indexes

Applications

Experiments

Open Source Science

kleos

Complete GPU residency for ML.

https://github.com/osayamenja/kleos

Science Score: 44.0%

This score indicates how likely this project is to be science-related based on various indicators:

  • ✓
    CITATION.cff file
    Found CITATION.cff file
  • ✓
    codemeta.json file
    Found codemeta.json file
  • ✓
    .zenodo.json file
    Found .zenodo.json file
  • ○
    DOI references
  • ○
    Academic publication links
  • ○
    Academic email domains
  • ○
    Institutional organization owner
  • ○
    JOSS paper metadata
  • ○
    Scientific vocabulary similarity
    Low similarity (6.9%) to scientific vocabulary
Last synced: 6 months ago · JSON representation ·

Repository

Complete GPU residency for ML.

Basic Info
  • Host: GitHub
  • Owner: osayamenja
  • License: other
  • Language: Cuda
  • Default Branch: main
  • Homepage:
  • Size: 5.64 MB
Statistics
  • Stars: 37
  • Watchers: 4
  • Forks: 7
  • Open Issues: 2
  • Releases: 0
Created almost 2 years ago · Last pushed 7 months ago
Metadata Files
Readme License Citation

README.md

Kleos Conceptual Overview

Complete, efficient GPU residency for Distributed Machine Learning.

🌹Kleos: GPU-Resident Runtime for ML

Kleos is an ongoing research project exploring the design of GPU-resident infrastructure for distributed machine learning workloads. The goal is to eliminate CPU bottlenecks by fusing scheduling, communication, and compute directly on the GPU using lightweight, asynchronous primitives.

🎯 Vision and Methods

We aim to achieve complete, efficient GPU residency. We investigate optimizations that approach hardware peak performance for both distributed and single GPU workloads, where bulk-synchronous, CPU-driven orchestration is a limiting factor. To attain this vision, we employ kernel fusion enabled by (1) algorithmic innovations with strong theoretical footing and (2) principled systems design and implementation.

This repository represents a very early-stage release of Kleos infrastructure.

🗞️ News

  • June 5, 2025 — ⚡️Introducing FlashDMoE, a fused GPU kernel for distributed MoE execution.
    ➤ See this README for details, benchmarks, and usage.

⚖️ License

This project is licensed under the BSD 3-Clause License. See LICENSE for full terms.

Owner

  • Name: Osayamen Aimuyo
  • Login: osayamenja
  • Kind: user
  • Location: Dallas, Texas.

Citation (CITATION.cff) 

cff-version: 1.2.0
message: "If you use this software, please cite it as below."
type: software
authors:
- family-names: "Aimuyo"
  given-names: "Osayamen Jonathan"
  orcid: "https://orcid.org/0009-0007-3272-1063"
  email: "osayamen@stanford.edu"
title: "Kleos"
date-released: 2025-06-06
url: "https://github.com/osayamenja/Kleos"

GitHub Events

Total
  • Issues event: 3
  • Watch event: 24
  • Issue comment event: 4
  • Push event: 8
  • Public event: 1
  • Pull request review event: 2
  • Pull request review comment event: 1
  • Pull request event: 3
  • Fork event: 2
  • Create event: 2
Last Year
  • Issues event: 3
  • Watch event: 24
  • Issue comment event: 4
  • Push event: 8
  • Public event: 1
  • Pull request review event: 2
  • Pull request review comment event: 1
  • Pull request event: 3
  • Fork event: 2
  • Create event: 2