[Python] How to create background process with smallest GPU memory consumption

I am trying to create a simple script that will be reserving small amount of gpu memory on each GPU. Process will be running in the background and if you run nvidia-smi you can see that it consumes GPUs. However the simple code below consumes quite a lot, about 522mb, on each GPUs. Is it possible to rewrite the script and reduce gpu memory consumption? Ideally I want the process to only consume 1mb from each gpu?

import torch
import time

if torch.cuda.is_available():
n = torch.cuda.device_count()

print(f"Total cuda devices {n}")
vars = []
for i in range(n):
a = torch.tensor(True).to(f'cuda:{i}')
vars.append(a)


while True:
time.sleep(60)




Continue reading...