The unexpected behavior of min()

[hyperkai]

🐛 Describe the bug

min() with input and keepdim argument gets the error but min() with input and dim doesn't get error even though the error message doesn't have such a combination as shown below:

import torch

my_tensor = torch.tensor([0, 1, 2])

torch.min(input=my_tensor, keepdim=True) # Error

torch.min(input=my_tensor, dim=0)
# torch.return_types.min(
# values=tensor(0),
# indices=tensor(0))

TypeError: min() received an invalid combination of arguments - got (keepdim=bool, input=Tensor, ), but expected one of:
 * (Tensor input, *, Tensor out)
 * (Tensor input, Tensor other, *, Tensor out)
 * (Tensor input, int dim, bool keepdim, *, tuple of Tensors out)
 * (Tensor input, name dim, bool keepdim, *, tuple of Tensors out)

And following the error message above, I couldn't use min() with input and the name type of dim argument as shown below but the name type of dim argument is not explained in the doc of min():

import torch

my_tensor = torch.tensor([0, 1, 2])

torch.min(input=my_tensor, dim='zero') # RuntimeError: Name 'zero' not found in Tensor[None].

Versions

import torch

torch.__version__ # 2.2.1+cu121

cc @malfet

[janeyx99]

We would accept a PR making the typechecking consistent with what is supported.