Question about the model yaml config `accelerator_type_a100`

Hi, aviary community!
In scaling_config, I find the accelerator_type_a10 option of resources_per_worker can be 0.01. Does this mean the model will use only 1% of the GPU memory and SMs? I am not sure the meaning.

I have the same question, what is accelerator_type_a10 used for?