DEV Community

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

Posted on

isclose and equal in PyTorch

#python #pytorch #isclose #equal

Buy Me a Coffee☕

*Memos:

isclose() can check if the zero or more elements of the 1st 0D or more D tensor are equal or nearly equal to the zero or more elements of the 2nd 0D or more D tensor element-wise, getting the 0D or more D tensor of zero or more elements as shown below:

*Memos:

  • isclose() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
  • The 2nd argument with torch or the 1st argument with a tensor is other(Required-Type:tensor of int, float, complex or bool).
  • The 3rd argument with torch or the 2nd argument with a tensor is rtol(Optional-Default:1e-05-Type:float).
  • The 4th argument with torch or the 3rd argument with a tensor is atol(Optional-Default:1e-08-Type:float).
  • The 5th argument with torch or the 4th argument with a tensor is equal_nan(Optional-Default:False-Type:bool): *Memos:
    • If it's True, nan and nan return True.
    • Basically, nan and nan return False.
  • The formula is |input - other| <= rtol x |other| + atol. 
import torch tensor1 = torch.tensor([1.00001001, 1.00000996, 1.00000995, torch.nan]) tensor2 = torch.tensor([1., 1., 1., torch.nan]) torch.isclose(input=tensor1, other=tensor2) torch.isclose(input=tensor1, other=tensor2, rtol=1e-05, atol=1e-08, equal_nan=False) # 0.00001 # 0.00000001 tensor1.isclose(other=tensor2) torch.isclose(input=tensor2, other=tensor1) # tensor([False, False, True, False])  torch.isclose(input=tensor1, other=tensor2, equal_nan=True) # tensor([False, False, True, True])  tensor1 = torch.tensor([[1.00001001, 1.00000996], [1.00000995, torch.nan]]) tensor2 = torch.tensor([[1., 1.], [1., torch.nan]]) torch.isclose(input=tensor1, other=tensor2) torch.isclose(input=tensor2, other=tensor1) # tensor([[False, False], # [True, False]])  tensor1 = torch.tensor([[[1.00001001], [1.00000996]], [[1.00000995], [torch.nan]]]) tensor2 = torch.tensor([[[1.], [1.]], [[1.], [torch.nan]]]) torch.isclose(input=tensor1, other=tensor2) torch.isclose(input=tensor2, other=tensor1) # tensor([[[False], [False]], # [[True], [False]]])  tensor1 = torch.tensor([[1.00001001, 1.00000996], [1.00000995, torch.nan]]) tensor2 = torch.tensor([1., 1.]) torch.isclose(input=tensor1, other=tensor2) torch.isclose(input=tensor2, other=tensor1) # tensor([[False, False], # [True, False]])  tensor1 = torch.tensor([[1.00001001, 1.00000996], [1.00000995, torch.nan]]) tensor2 = torch.tensor(1.) torch.isclose(input=tensor1, other=tensor2) torch.isclose(input=tensor2, other=tensor1) # tensor([[False, False], # [True, False]])  tensor1 = torch.tensor([0, 1, 2]) tensor2 = torch.tensor(1) torch.isclose(input=tensor1, other=tensor2) # tensor([False, True, False])  tensor1 = torch.tensor([0.+0.j, 1.+0.j, 2.+0.j]) tensor2 = torch.tensor(1.+0.j) torch.isclose(input=tensor1, other=tensor2) # tensor([False, True, False])  tensor1 = torch.tensor([False, True, False]) tensor2 = torch.tensor(True) torch.isclose(input=tensor1, other=tensor2) # tensor([False, True, False])
Enter fullscreen mode Exit fullscreen mode

equal() can check if two of 0D or more D tensors have the same size and elements, getting the scalar of a boolean value as shown below:

*Memos:

  • equal() can be used with torch or a tensor.
  • The 1st argument(input) with torch or using a tensor(Required-Type:tensor of int, float, complex or bool).
  • The 2nd argument with torch or the 1st argument with a tensor is other(Required-Type:tensor of int, float, complex or bool). 
import torch tensor1 = torch.tensor([5, 9, 3]) tensor2 = torch.tensor([5, 9, 3]) torch.equal(input=tensor1, other=tensor2) tensor1.equal(other=tensor2) torch.equal(input=tensor2, other=tensor1) # True  tensor1 = torch.tensor([5, 9, 3]) tensor2 = torch.tensor([7, 9, 3]) torch.equal(input=tensor1, other=tensor2) torch.equal(input=tensor2, other=tensor1) # False  tensor1 = torch.tensor([5, 9, 3]) tensor2 = torch.tensor([[5, 9, 3]]) torch.equal(input=tensor1, other=tensor2) torch.equal(input=tensor2, other=tensor1) # False  tensor1 = torch.tensor([5., 9., 3.]) tensor2 = torch.tensor([5.+0.j, 9.+0.j, 3.+0.j]) torch.equal(input=tensor1, other=tensor2) torch.equal(input=tensor2, other=tensor1) # True  tensor1 = torch.tensor([1.+0.j, 0.+0.j, 1.+0.j]) tensor2 = torch.tensor([True, False, True]) torch.equal(input=tensor1, other=tensor2) torch.equal(input=tensor2, other=tensor1) # True  tensor1 = torch.tensor([], dtype=torch.int64) tensor2 = torch.tensor([], dtype=torch.float32) torch.equal(input=tensor1, other=tensor2) torch.equal(input=tensor2, other=tensor1) # True
Enter fullscreen mode Exit fullscreen mode

Top comments (0)