Overview

What is it?

The IVS AI infrastructure is built up a set of services that help users:

• Reserve computing resources: currently we have available a pool of 4x Nvidia V100 16G in a single server. These resources can be booked in 1x GPU slots.

• Manage your datasets: we offer 500GB of storage to manage your datasets and your results. This data can be shared across research groups and individuals. Data quota can be increased on request.

• Manage your AI workloads: launch your AI training workloads in containers with access to logs and system monitoring.

Why do I want it?

If you are a research scientist at IVS or have a collaboration agreement, you can request the use of the resources for AI training purposes.

• What is it good for?: The IVS AI infrastructure can handle AI training workloads in single GPU mode up to 16GB GPU of memory. This infrastructure is ideal for workloads that fit the amount of memory available and that would take long time in a consumer-grade GPU.

• What is it not good for?: AI training workloads that would exceed the amount of memory provided (16GB).

• What is it not meant for?: The server is a production server and should not be used for tasks such as development of algorithms or inference. These tasks require human interaction and make a poor use of the available resources.

Where should I go next?

• Getting Started: Get started with IVS AI Infrastructure