Do we get NVIDIA GPU to run daisi?

Do we get Nvidia GPU hardware to deploy apps which require NVIDIA CUDA support to run properly?
Suppose Stable Diffusion requires CUDA to operate.

Hi @kali : we are actively working on providing access to GPUs, it is currently in testing on our side. It will be available to the community in the coming days.
We will make an announcement as soon as it is available!

2 Likes

Hi @kali : GPUs are now available on the Daisi platform! Check the doc here: Accessing GPUs on the Daisi platform - Daisi documentation

Thanks for your response I’ll desperately check for it and test some projects on it.

Hi I have checked the documentation part of usage of gpu in daisy can you mention want is the meaning of 0 and 1 in os.environ[“CUDA_VISIBLE_DEVICES”] = “0”,“1”

Each node has two GPUs, that are indexed 0 and 1. Your code will run on one node, and you can specify on which GPU it will run. This is the way CUDA works

So we can use any of the nodes either zero or one right. One more thing, the use cases part https://doc.daisi.io/use-cases/0.ml-model/ has incomplete documentation . That specifies how to deploy a model with daisi .

Yes you can choose 0 or 1
The doc is still in progress, we will update this chapter in the coming days.

gotit thanks for the response

I don’t think that we really need to make multiple daisy to use cuda service in our apps instead we can do the save in one go you can check it here https://app.daisi.io/daisies/kali/checkDiffusion/app I haven’t made multiple changes related to ui to understand it better.

It’s only for testing so please don’t delete it saying it’s a duplicate. I’ll update it later after your response.