ht.array creates a new DNDarray on the CPU when the arguments copy = False and split >= 0 are passed. This should not happen.
In addition, copy = None returns the same DNDarray which doesn't infer the device.
Code snippet triggering the error
import torch
import heat as ht
m,n = 40,20
a = torch.randn(m, n, dtype=torch.double, device='cuda')
b = ht.array(a, copy=False, split=0, dtype=ht.double)
print('a.device =', a.device)
print('b.device =', b.device)
print('b.split =', b.split)
What happened?
ht.array creates a new DNDarray on the CPU when the arguments copy = False and split >= 0 are passed. This should not happen. In addition, copy = None returns the same DNDarray which doesn't infer the device.
Code snippet triggering the error
Error message or erroneous outcome
Version
1.3.x
Python version
None
PyTorch version
None
MPI version
No response