mygrad.Tensor.null_grad

Tensor.null_grad(*, _clear_view_info: bool = False) → Tensor

Sets this tensor’s gradient to be None.

This operation is performed in-place, but a reference to the tensor is returned in order to permit mapping semantics.

Also removes any base reference from disconnected views.

Returns:

self

Examples

>>> import  mygrad as mg
>>> x = mg.Tensor(2.)
>>> (x ** 2).backward()
>>> x.grad
array(4.)
>>> x.null_grad()  # returns a reference of `x`
Tensor(2.0)
>>> x.grad is None
True