You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
NumPy uses the .base ndarray attribute to trace the views, as in
In [77]: a = np.arange(10)
In [78]: a[::2].base is a
Out[78]: True
In [79]: a[::2].copy().base is a
Out[79]: False
It's been suggested (#46 (comment)) to drop this, because PyTorch tracks views vs copies in a private Tensor._base attribute.
While there is no way to restore the pubic base from private _tensor._base, the question is whether we want/need to keep the .base on the ndarray wrapper level.
The text was updated successfully, but these errors were encountered:
NumPy uses the
.base
ndarray attribute to trace the views, as inIt's been suggested (#46 (comment)) to drop this, because PyTorch tracks views vs copies in a private
Tensor._base
attribute.While there is no way to restore the pubic
base
from private_tensor._base
, the question is whether we want/need to keep the.base
on the ndarray wrapper level.The text was updated successfully, but these errors were encountered: