torch.Tensor.add, add_, etc. documentation doesn't match api.

[cpuhrsch]

📚 Documentation

For example torch.add says

add(value) → Tensor
    add(other, *, value=1) -> Tensor

    See torch.add()

But indeed the value flag doesn't exist.

>>> import torch
>>> a = torch.randn(1, 3)
>>> b = torch.randn(1, 3)
>>> a
tensor([[-0.5835, -0.1405, -1.8119]])
>>> b
tensor([[ 0.8623, -1.3727,  1.5434]])
>>> a.add(b)
tensor([[ 0.2789, -1.5132, -0.2685]])
>>> a.add(b, 1)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: add() takes 1 positional argument but 2 were given
>>> a.add(b, 2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: add() takes 1 positional argument but 2 were given
>>> a.add(b, valu=2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: add() got an unexpected keyword argument 'valu'
>>> a.add(b, value=2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: add() got an unexpected keyword argument 'value'
>>> a.add(b, 2)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: add() takes 1 positional argument but 2 were given

Instead one has to use alpha as a flag.

>>> import torch
>>> a = torch.randn(1, 3)
>>> b = torch.randn(1, 3)
>>> a.add(a, alpha=3.0)
tensor([[ 3.3190, -0.4248, -3.6609]])

There are many other instances of this in the documentation.

[gchanan]

Closing in favor of #19200.