from etaler import et
import numpy as np
t = et.ones((4, 4))
a = np.array(t)
print(shape) # (4, 4, 1, 1..... ,1)
This is likely cause by Etaler don't have a concept of a 0D tensor (i.e. scalar). Etaler always treat scalars as 1D, length 1 tensors. Will need fix from the C++ side.
TL;DR
This is likely cause by Etaler don't have a concept of a 0D tensor (i.e. scalar). Etaler always treat scalars as 1D, length 1 tensors. Will need fix from the C++ side.