DEV Community

Super Kai (Kazuya Ito)
Super Kai (Kazuya Ito)

Posted on • Updated on

round, ceil, floor and trunc in PyTorch

Buy Me a Coffee

round() can round the zero or more elements of a 0D or more D tensor up to 4 decimal places, getting the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • round() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int or float).
  • The 2nd argument with torch or the 1st argument with a tensor is decimals(Optional-Default:0-Type:int): Memos:
    • It is a decimal place.
    • You must use decimals=.
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
import torch

my_tensor = torch.tensor([4.735277, -3.370677, 1.888877, -7.888877])

torch.round(input=my_tensor)
my_tensor.round()
torch.round(input=my_tensor, decimals=0)
# tensor([5., -3., 2., -8.])

torch.round(input=my_tensor, decimals=1)
# tensor([4.7000, -3.4000, 1.9000, -7.9000])

torch.round(input=my_tensor, decimals=2)
# tensor([4.7400, -3.3700, 1.8900, -7.8900])

torch.round(input=my_tensor, decimals=3)
# tensor([4.7350, -3.3710, 1.8890, -7.8890])

torch.round(input=my_tensor, decimals=4)
torch.round(input=my_tensor, decimals=5)
torch.round(input=my_tensor, decimals=6)
etc...
# tensor([4.7353, -3.3707, 1.8889, -7.8889])

my_tensor = torch.tensor([4, -3, 1, -7])

torch.round(input=my_tensor)
# tensor([4, -3, 1, -7])
Enter fullscreen mode Exit fullscreen mode

ceil() can round up the zero or more elements of a 0D or more D tensor, getting the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • ceil() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int or float).
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
  • ceil() doesn't have decimals argument.
import torch

my_tensor = torch.tensor([4.735277, -3.370677, 1.888877, -7.888877])

torch.ceil(input=my_tensor)
my_tensor.ceil()
# tensor([5., -3., 2., -7.])

my_tensor = torch.tensor([4, -3, 1, -7])

torch.ceil(input=my_tensor)
# tensor([4, -3, 1, -7])
Enter fullscreen mode Exit fullscreen mode

floor() can round down the zero or more elements of a 0D or more D tensor, getting the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • floor() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int or float).
  • There is out argument with torch(Optional-Default:None:Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
  • floor() doesn't have decimals argument.
import torch

my_tensor = torch.tensor([4.735277, -3.370677, 1.888877, -7.888877])

torch.floor(input=my_tensor)
my_tensor.floor()
# tensor([4., -4., 1., -8.])

my_tensor = torch.tensor([4, -3, 1, -7])

torch.floor(input=my_tensor)
# tensor([4, -3, 1, -7])
Enter fullscreen mode Exit fullscreen mode

trunc() can truncate the decimal part of zero or more elements of a 0D or more D tensor, getting the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • trunc() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int or float).
  • There is out argument with torch(Optional-Default:None-Type:tensor): *Memos:
    • out= must be used.
    • My post explains out argument.
  • trunc() doesn't have decimals argument.
  • fix() is the alias of trunc().
import torch

my_tensor = torch.tensor([4.735277, -3.370677, 1.888877, -7.888877])

torch.trunc(input=my_tensor)
my_tensor.trunc()
# tensor([4., -3., 1., -7.])

my_tensor = torch.tensor([4, -3, 1, -7])

torch.trunc(input=my_tensor)
# tensor([4, -3, 1, -7])
Enter fullscreen mode Exit fullscreen mode

Top comments (0)