PulseAugur
EN
LIVE 09:36:08

KOSMOS library accelerates LLM training on AMD GPUs by up to 1.83x

By PulseAugur Editorial · [1 sources] ·

A new library called KOSMOS has been developed to improve the efficiency of training large language models on GPUs. KOSMOS optimizes the communication and computation phases within transformer layers by using self-scheduling persistent kernels that adapt to data readiness. When evaluated on AMD Instinct MI350X and MI355X GPUs, KOSMOS demonstrated significant speedups, ranging from 1.10x to 1.83x, for models like Megatron-LM, DeepSeek-V3, and Qwen3-235B compared to existing methods. AI

IMPACT This library could significantly reduce LLM training times and costs, making large model development more accessible.

RANK_REASON The cluster describes a new library and its performance evaluation presented in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

KOSMOS library accelerates LLM training on AMD GPUs by up to 1.83x

How we ranked this

Signal score
16 / 100
Composite score across the factors below. Higher = stronger signal that this story matters right now.
Newsworthiness bucket
Tool
The cluster describes a new library and its performance evaluation presented in an arXiv paper. [lever_c_demoted from research: ic=1 ai=1.0]
Source corroboration
Single-source cluster
Only one publisher covered this so far. Single-source stories can still rank when the publisher is high-authority, but they lack cross-source corroboration.
Topics
infra, paper
Editorial topic classification. Feeds into how the story surfaces on /topic/<slug> hub pages and into the per-entity coverage mix.
AI-industry relevance
High
Clearly on-topic for AI-industry coverage.
Story freshness
Breaking (< 6h)
Fresh story with cross-source coverage still developing. Ranking may shift as more sources report.

Full methodology in our editorial standards.

COVERAGE [1]

  1. arXiv cs.AI TIER_1 English(EN) · Alexander Magro, Aristotle Martin, Ye Wang, Wen Chen ·

    KOSMOS: Self-scheduling Persistent Kernels for Communication-Compute Overlap

    arXiv:2611.07477v1 Announce Type: cross Abstract: Tensor and sequence parallelism (TP+SP) and expert parallelism (EP) put collectives on the critical path of transformer layers. State-of-the-art fused kernels mitigate this with fixed roles or schedules, but this rigidity leaves l…

RELATED ENTITIES

RELATED TOPICS