GPU Machine Learning From Laptop to Cloud
summary
Step-by-step workflow that builds a containerized Keras/TensorFlow pipeline from laptop to GKE using Docker + NVIDIA runtime, CUDA drivers and Minikube (VM-driver none or KVM2), with an orchestrator pod that downloads data to a read‑only PVC and other pods that run training jobs. Claims 10–20x speedup on GPU vs CPU, notes secure‑boot and version conflicts, and shows hyperparameter sweeping by launching multiple jobs (one GPU queues tasks). Demo produces TensorBoard‑style traces and saved model weights; mostly works though Minikube GPU limits and cert hassles appear.
AI-written summary · about
chapters (15)
- 0:00Intro
- 0:32Agenda
- 2:30Laptop Setup
- 7:59Docker Setup
- 9:40Running Docker
- 11:05Running in MiniChip
- 16:48Bringing data in
- 17:11Persistent volume
- 18:01Orchestrator
- 19:00Deploy
- 19:34Endpoint
- 25:05Job Builder
- 26:54Data Refresh
- 30:05Post JSON
- 30:30Summary
related talks
38mAll You Need to Know to Build Your GPU Machine Learning Cloud [B]
34mBuilding GPU-Accelerated Workflows with TensorFlow and Kubernetes [I]
38mKubernetes for AI Workloads – What Works and What… Doesn’t
34mGPU as a Service Over K8s: Drive Productivity and Increase Utilization
26mAutomating GPU Infrastructure for Kubernetes
39m✨ GPU in My Containers!
34mDemocratizing Machine Learning on Kubernetes [I]
31mKubernetes For GPU Powered Machine Learning Workloads In Academia