What is the difference between assiging a float value and a tensor value?

[Endtroducing1993]

Hi Team,

Their are two random tensors that I created:

Tensor A:
tensor([[0.2054, 0.9767, 0.3130, 0.2153],
[0.0492, 0.5223, 0.7216, 0.6107],
[0.5989, 0.1208, 0.0331, 0.5088]])

Tensor B:
tensor([[0.9559, 0.7885, 0.2089, 0.4351],
[0.1314, 0.2588, 0.5905, 0.7723],
[0.9142, 0.0409, 0.8343, 0.1474]])

Now If I assign value 0.9559 to tensor_B in the first row and first column i.e. tensor_B[0,0]=0.2054
then I get the below results:
tensor([[0.9559, 0.9767, 0.3130, 0.2153],
[0.0492, 0.5223, 0.7216, 0.6107],
[0.5989, 0.1208, 0.0331, 0.5088]])

But when I equate them I get all False. I have checked the type also they both are type float32.
tensor_A==tensor_B
tensor([[ False, False, False, False],
[False, False, False, False],
[False, False, False, False]])

But if I do tensor_B[0,0]=tensor_A[0,0]
We get the below result:
tensor([[ True, False, False, False],
[False, False, False, False],
[False, False, False, False]])

Request you to please suggest why are we getting True in one case but False in other. Where the datatype as both torch.float32

[Perian-Yan]

Hi,
I guess the randomly generated tensor is float32. So 0.2054 is just a rough number, and when you assign it to tensor_B[0,0], it's not equal to the exact number.