Spike: Exploring the Need for a Dedicated GPU-Test Project/Image

schwesig commented 2 months ago

Exploring the Need for a Dedicated GPU-Test Project/Image

We've been testing new GPUs using existing projects, but for future tests, including adding or removing GPUs, a standardized test project/image might be more effective.

Objective:

Discuss the creation of a dedicated GPU test project/image for consistent and efficient GPU testing across our infrastructure.

Key Points:

Advantages: Quick deployment and reliable performance benchmarking.
Scope: What kind of tests should it include?
Implementation: What tools and libraries are essential? How do we integrate this with our current setup?

Let's evaluate if a GPU test project/image can streamline our scaling needs.

schwesig commented 2 months ago

connected before and after issues:

joachimweyl commented 2 months ago

@schwesig please add an estimate for this spike.

computate commented 2 months ago

@schwesig I have an existing GPU testing image that I have used before, and we can add any additional features to it. https://github.com/nerc-images/cuda-jupyter-tensorflow

nerc-project / operations