Gpu wise memory allocation

Ray Libraries (Data, Train, Tune, Serve) Ray Tune
1

Hi,
I have the same question, that was asked earlier (https://github.com/ray-project/ray/discussions/11696), but since there has not been any activity for more than a month - I am asking again here.

Hi.

I am wondering if it is possible to configure memory fractions gpu-wise for more flexible usages on my gpus with different memory sizes. For example,

AS-IS

resources_per_trial={"cpu": 3, "gpu": 0.5}

WHAT-I-WANT

resources_per_trial={"cpu": 3, "gpu": gpu_wise_fraction}
gpu_wise_fraction = {
   gpu1: 0.5,
   gpu2: 0.1,
   gpu3: 0.3,
}

Thanks in advance.