GPU requirements for Deploy SDK

Hi there,

I am hoping to implement the Nvidia Clara AI Colon Tumor Segmentation Pipeline, connected to XNAT on a Linux server in the hospital I work in (I’m a medical doctor, not an IT professional, so apologies in advance).

What are the specs required for this kind of inference? I imagine there’s a specs doc somewhere which if someone could link to me, I’d greatly appreciate it.

All the best,

Reuben Schmidt

Hi Reuben,

Thanks for your interest in the Clara Deploy platform!

You can find detailed system requirements here:

At a high level, any GPU Pascal or newer will run the Deploy platform, including Pascal, Volta, Turing and Ampere families.

I just ran the Clara AI Colon Tumor Segmentation Pipeline on a T4 GPU using the bundled test DCM input data and observed about 14GB GPU memory utilization.

To be safe, you’ll probably want to look for something with a minimum of 16GB GPU memory, possibly more if you have larger datasets or will run multiple Deploy pipelines and models.