Slurm
PulseAugur coverage of Slurm — every cluster mentioning Slurm across labs, papers, and developer communities, ranked by signal.
5 day(s) with sentiment data
-
Slurm-web v7.1.0 released with improved job details and GPU metrics
Slurm-web, an open-source web dashboard for Slurm clusters, has released version 7.1.0. This maintenance release includes accurate job details for requeued jobs and reused job IDs. It also provides reliable GPU metrics …
-
Slurm Workload Manager optimizes NVIDIA GPU sharing for AI clusters
This item details how to deploy the Slurm Workload Manager to optimize the use of NVIDIA GPUs within AI research clusters. The tutorial explains how Slurm can efficiently share and isolate these graphics processing unit…
-
New research models multi-vendor MPI ensembles for heterogeneous computing
A new research paper explores extending predictive theories of communication in parallel computing to heterogeneous environments. The study details a novel approach using a multi-vendor MPI ensemble that combines NVIDIA…
-
Deep Reinforcement Learning Tested for HPC Job Scheduling
A new research paper evaluates the effectiveness of six Deep Reinforcement Learning (DRL) treatments for High-Performance Computing (HPC) job scheduling. The study compared Proximal Policy Optimization (PPO), Advantage …
-
Slurm-web launches enterprise support for HPC cluster monitoring
Slurm-web, an open-source web UI for monitoring Slurm HPC clusters and tracking jobs, is now offering enterprise support. This support includes guaranteed response times under an SLA and is provided directly by the deve…
-
AWS SageMaker HyperPod enables shared GPU cluster access for AI teams
AWS has introduced Amazon SageMaker HyperPod, a new service designed to help organizations manage large-scale GPU clusters for generative AI workloads. This service addresses the challenge of multiple teams needing shar…
-
AWS SageMaker HyperPod enhances ML team governance and development workflows
AWS has introduced new features for Amazon SageMaker HyperPod, enhancing its administration and governance capabilities for machine learning teams. These updates allow for better management of shared compute clusters, i…
-
AI-powered HPC cluster Hyperion predicts job times for research
Researchers at the University of South Carolina have developed Hyperion, a new high-performance computing (HPC) cluster designed for both scientific and humanities research. The system integrates two machine learning mo…
-
New Software Platform Streamlines AI Evaluation in Digital Pathology
Researchers have developed STEP, a new software platform designed to streamline the operational evaluation of AI models in digital pathology. This modular engine facilitates prospective silent trials, allowing AI perfor…
-
NVIDIA releases srt-slurm for easier inference service deployment
NVIDIA has released srt-slurm, an open-source orchestration layer designed to simplify the deployment of complex inference services on GPU clusters managed by Slurm. This tool, developed by the NVIDIA AI team, uses a YA…
-
Futurama Season 14 criticized as disappointing, with a nod to AI infrastructure
A review of Futurama Season 14 criticizes the new episodes as disappointing, suggesting Hulu should reconsider the show's future if quality does not improve. The review specifically calls out a lack of engaging content,…
-
Slurm-quota v3.0.0 released with enhanced auth, API, and RBAC
Slurm-quota has released version 3.0.0, introducing significant enhancements for managing CPU and GPU time allocations within the Slurm workload manager. This update includes robust authentication mechanisms such as Lig…
-
NVIDIA releases srt-slurm for distributed LLM serving benchmarks
NVIDIA has released srt-slurm, a framework designed to streamline the creation and validation of distributed LLM serving benchmarks. The tool, demonstrated using Google Colab for development, allows users to define clus…
-
Together AI boosts GPU cluster reliability with auto-repair and Slinky 1.0
Together AI has enhanced its GPU Clusters with new features focused on reliability and operational control for large-scale AI workloads. These updates include passive health checks to detect hardware and software degrad…
-
GPUScope tackles GPU compute waste outside Kubernetes
Managing AI infrastructure outside of Kubernetes presents a significant challenge in tracking GPU compute waste. Current methods often involve complex, manual setups using tools like NVIDIA's DCGM exporter and Prometheu…
-
Slurm-web v7.0.0 adds SSO, container support for HPC
Slurm-web has released version 7.0.0, introducing significant enhancements for High-Performance Computing (HPC) users and administrators. This update includes support for single sign-on via OpenID Connect, deployment op…
-
Fugaku-NEXT design report details hybrid AI/HPC scheduling and fast storage
A design report for Fugaku-NEXT, a successor to the Fugaku supercomputer, has been released, detailing ambitious plans for its architecture. The system will integrate Kubernetes for AI workloads alongside traditional HP…
-
Expanse launches to optimize GPU cluster utilization and cut waste
Expanse, a startup founded by Ismaeel, Eren, Yafet, and Nikodem, has launched a platform designed to optimize GPU cluster utilization. Their system analyzes job code, submission scripts, and hardware telemetry to predic…
-
Together AI launches self-service GPU clusters for AI development
Together AI has launched Together Instant Clusters, a new service providing readily available, self-service GPU clusters for AI development and deployment. This offering aims to simplify the complex process of setting u…
-
Hugging Face and AWS Detail Foundation Model Infrastructure
Hugging Face and AWS have collaborated to detail the infrastructure required for training and running large foundation models. The blog post outlines a layered architecture, emphasizing the interplay between AWS's compu…