Results for “gpu-cluster”

Checks and installs NVIDIA driver, CUDA Toolkit, and NVIDIA Container Toolkit for GPU-accelerated Docker and Kubernetes hosts. Supports multiple Linux distributions with automated install and read-only check modes.

Plans, executes, and validates Google Kubernetes Engine (GKE) cluster upgrades and maintenance operations for both Standard and Autopilot clusters, producing upgrade plans, checklists, and runbooks with gcloud commands.

Runs batch processing and high-performance computing (HPC) workloads on Google Kubernetes Engine (GKE), including job queues, parallel processing, and MPI workloads.

Routes to specialized GKE sub-skills for cluster management, networking, security, scaling, and more on Google Kubernetes Engine.

Deploys and optimizes AI/ML inference workloads on GKE, using GPUs, TPUs, and model servers.

Provides guidance on enabling and optimizing GKE Cluster Autoscaler, including Node Auto Provisioning, troubleshooting scale-up/down issues, and best practices for capacity management.

Configures, optimizes, and troubleshoots GKE ComputeClasses for Spot VMs with on-demand fallback, GPU/TPU targeting, machine family selection, and zone colocation.

Hardens Google Kubernetes Engine (GKE) clusters with Workload Identity, Secret Manager, RBAC, Binary Authorization, Network Policies, and Pod Security Standards.

Configures GKE storage including PVCs, PersistentVolumes, Filestore, and GCS FUSE with best practices for production workloads.

Accelerate pandas workflows with GPU DataFrames using cuDF and dask-cuDF for ETL, joins, groupby, and large-scale data processing.

Reduces peak GPU memory in Megatron Bridge training by applying expandable segments, parallelism resizing, activation recompute, and CPU offloading constraints.

Submits TAO container jobs as single-pod Kubernetes Jobs with NVIDIA GPU scheduling on EKS, GKE, AKS, or on-prem clusters.

Plans and configures multi-tenancy on GKE, covering namespace isolation, RBAC planning, resource quotas, LimitRanges, network isolation, and cost allocation.

Launch distributed Megatron-LM training jobs on a SLURM cluster with a minimal sbatch skeleton, environment-variable setup for torch.distributed.run, CUDA_DEVICE_MAX_CONNECTIONS rules, container conventions, monitoring, and per-rank failure diagnosis.

Protects stateful GKE workloads by configuring backup plans, restore workflows, and disaster recovery using Backup for GKE.

Plans, configures, and manages GKE networking including private clusters, VPC-native configurations, Gateway API, DNS, ingress/egress, Dataplane V2, and IP planning.

Configures GKE workload reliability with PodDisruptionBudgets, health probes, topology spread constraints, and graceful shutdown patterns.

Launch, monitor, stop, and debug NeMo-RL recipes on a Kubernetes cluster using the nrl-k8s CLI, supporting ephemeral and long-lived RayCluster modes.