DefaultCPUAllocator: not enough memory in matrix multiplication broadcasting about pytorch HOT 2 OPEN

What are the strides of A and B?

from pytorch.

A is the data I saved (attached here A.zip) and B is just a zero tensor created by torch.zeros. Running this code would give

A = torch.load('A.pt')
print(A.stride())
print(A.contiguous().stride())
print(A.clone().stride())

B = torch.zeros([1, 2048, 2048, 1], dtype=torch.complex64)
print(B.stride())

(4194304, 2048, 1)
(4194304, 2048, 1)
(4194304, 2048, 1)
(4194304, 2048, 1, 1)

from pytorch.