When one gpu has multiple cpu, can it run concurrently on python?

[WYF10086]

I tried code like this:
device = 'cuda'
model = whisper.load_model('medium', device=device)
model.share_memory()
pool = torch.multiprocessing.Pool(2)
queue = torch.multiprocessing.Manager().Queue()
start = datetime.now()
for file_name in file_list:
pool.apply_async(worker,(model,file_name,queue))

Now there are two questions：

1. According to my assumption, will this operation increase the number of audio conversions per second, or for this purpose, how to do it?
2. if not model.share_memory(), the following worker function will not be entered, so I search and found that this might be needed, but it will have an error ：NotImplementedError: Cannot access storage of SparseTensorImpl

Can anyone help me with this answer? Thank you.