Run:ai Quickstart Guides

Below is a set of Quickstart documents. The purpose of these documents is to get you acquainted with an aspect of Run:ai in the simplest possible form.

Note

The Quickstart documents are based solely on the command-line interface. The same functionality can be achieved by using the Workloads User interface which allows for Workload submission and log viewing.

Follow the Quickstart documents below to learn more:

Training Quickstart documents:
- Unattended training sessions
- Distributed Training
Build Quickstart documents:
Inference
GPU Allocation documents:
- Using GPU Fractions
- Dynamic MIG
Scheduling documents:
- Over-Quota, Basic Fairness & Bin Packing
- Fairness

Most quickstarts rely on an image called gcr.io/run-ai-demo/quickstart. The image is based on TensorFlow Release 20-08. This TensorFlow image has minimal requirements for CUDA and NVIDIA Compute Capability.

If your GPUs do not meet these requirements, use gcr.io/run-ai-demo/quickstart:legacy instead.