[Feature] Support Nvidia A100 accelerator_type

[vicyap]

Search before asking

• I had searched in the issues and found no similar feature requirement.

Description

I could not find Nvidia's A100 as an accelerator type in https://github.com/ray-project/ray/blob/master/python/ray/util/accelerators/accelerators.py. I'm curious if this is simply missing or if there's a reason it's not included.

Use case

AWS offers the A100 in the p4d.24xlarge instance type. We use that instance type along with p3 (Nvidia V100). I would like to add A100 so code is more uniform.

from ray.util.accelerators import NVIDIA_TESLA_A100

@ray.remote(num_gpus=1, accelerator_type=NVIDIA_TESLA_A100)
def train():
    ....

Related issues

I found #12413, but that issues is trying to make use of the A100's MIG (multi-instance gpu) feature. That feature would be cool but will take longer to merge.

Are you willing to submit a PR?

• Yes I am willing to submit a PR!