min(), max(), argmin() and argmax() in PyTorch

*Memos:

min() can get the one or more minimum values of a 0D or more D tensor as shown below:

*Memos:

min() can be called both from torch and a tensor.
Setting a dimension(dim) to the 2nd argument with torch or the 1st argument with a tensor gets zero or more 1st minimum values and the indices of them.
Only zero or more integers, floating-point numbers or boolean values can be used so zero or more complex numbers cannot be used.

import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.min(my_tensor)
my_tensor.min()
# tensor(3)

torch.min(my_tensor, 0)
my_tensor.min(0)
torch.min(my_tensor, -2)
my_tensor.min(-2)
# torch.return_types.min(
# values=tensor([3, 4, 3, 3]),
# indices=tensor([2, 0, 1, 2]))

torch.min(my_tensor, 1)
my_tensor.min(1)
torch.min(my_tensor, -1)
my_tensor.min(-1)
# torch.return_types.min(
# values=tensor([4, 3, 3]),
# indices=tensor([1, 2, 0]))

my_tensor = torch.tensor([[5., 4., 7., 7.], [6., True, 3., 5.], [3., 8., False, 3.]])
torch.min(my_tensor, 0)
my_tensor.min(0)
torch.min(my_tensor, -2)
my_tensor.min(-2)
# torch.return_types.min(
# values=tensor([3., 1., 0., 3.]),
# indices=tensor([2, 1, 2, 2]))

Enter fullscreen mode
Exit fullscreen mode

max() can get the one or more maximum values of a 0D or more D tensor as shown below:

*Memos:

max() can be called both from torch and a tensor.
Setting a dimension(dim) to the 2nd argument with torch or the 1st argument with a tensor gets zero or more 1st maximum values and the indices of them.
Only zero or more integers, floating-point numbers or boolean values can be used so zero or more complex numbers cannot be used.

import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.max(my_tensor)
my_tensor.max()
# tensor(9)

torch.max(my_tensor, 0)
my_tensor.max(0)
torch.max(my_tensor, -2)
my_tensor.max(-2)
# torch.return_types.max(
# values=tensor([6, 8, 9, 7]),
# indices=tensor([1, 2, 2, 0]))

torch.max(my_tensor, 1)
my_tensor.max(1)
torch.max(my_tensor, -1)
my_tensor.max(-1)
# torch.return_types.max(
# values=tensor([7, 6, 9]),
# indices=tensor([2, 0, 2]))

my_tensor = torch.tensor([[5., 4., 7., 7.], [6., True, 3., 5.], [3., 8., False, 3.]])
torch.max(my_tensor, 0)
my_tensor.max(0)
torch.max(my_tensor, -2)
my_tensor.max(-2)
# torch.return_types.max(
# values=tensor([6., 8., 7., 7.]),
# indices=tensor([1, 2, 0, 0]))

Enter fullscreen mode
Exit fullscreen mode

argmin() can get the indices of the 1st minimum values of a 0D or more D tensor as shown below:

*Memos:

argmin() can be called both from torch and a tensor.
The 2nd argument with torch or the 1st argument with a tensor is a dimension(dim).
Zero or more integers, floating-point numbers or boolean values can be used so zero or more complex numbers cannot be used except the 1D tensor of a complex number with dim=0 or dim=-1.

import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.argmin(my_tensor)
my_tensor.argmin()
# tensor(6)

torch.argmin(my_tensor, 0)
my_tensor.argmin(0)
torch.argmin(my_tensor, -2)
my_tensor.argmin(-2)
# tensor([2, 0, 1, 2])

torch.argmin(my_tensor, 1)
my_tensor.argmin(1) torch.argmin(my_tensor, -1)
my_tensor.argmin(-1) # tensor([1, 2, 0])

my_tensor = torch.tensor([[5., 4., 7., 7.], [6., True, 3., 5.], [3., 8., False, 3.]])
torch.argmin(my_tensor, 0)
my_tensor.argmin(0)
torch.argmin(my_tensor, -2)
my_tensor.argmin(-2)
# tensor([2, 1, 2, 2])

my_tensor = torch.tensor([5+7j])
torch.argmin(my_tensor, 0)
my_tensor.argmin(0)
torch.argmin(my_tensor, -1)
my_tensor.argmin(-1)
# tensor(0)

Enter fullscreen mode
Exit fullscreen mode

argmax() can get the indices of the 1st maximum values of a 0D or more D tensor:

*Memos:

argmax() can be called both from torch and a tensor.
The 2nd argument with torch or the 1st argument with a tensor is a dimension(dim).
Zero or more integers, floating-point numbers or boolean values can be used so zero or more complex numbers cannot be used except the 1D tensor of a complex number with dim=0 or dim=-1.

import torch my_tensor = torch.tensor([[5, 4, 7, 7], [6, 5, 3, 5], [3, 8, 9, 3]])
torch.argmax(my_tensor)
my_tensor.argmax()
# tensor(10)

torch.argmax(my_tensor, 0)
my_tensor.argmax(0)
torch.argmax(my_tensor, -2)
my_tensor.argmax(-2)
# tensor([1, 2, 2, 0])

torch.argmax(my_tensor, 1)
my_tensor.argmax(1)
torch.argmax(my_tensor, -1)
my_tensor.argmax(-1)
# tensor([2, 0, 2])

my_tensor = torch.tensor([[5., 4., 7., 7.], [6., True, 3., 5.], [3., 8., False, 3.]])
torch.argmax(my_tensor, 0)
my_tensor.argmax(0)
torch.argmax(my_tensor, -2)
my_tensor.argmax(-2)
# tensor([1, 2, 0, 0])

my_tensor = torch.tensor([5+7j])
torch.argmax(my_tensor, 0)
my_tensor.argmax(0)
torch.argmax(my_tensor, -1)
my_tensor.argmax(-1)
# tensor(0)

Enter fullscreen mode
Exit fullscreen mode

Discover more from Coursity

Subscribe to get the latest posts sent to your email.

min(), max(), argmin() and argmax() in PyTorch

Like this:

Related

Discover more from Coursity

Leave a Comment Cancel Reply