Getting decimal places when multiplying two integers

1

I was wondering why this happens:

n = nn.Conv2d(1, 1, 2, stride=1)
n.weight = nn.Parameter(torch.from_numpy(np.random.randint(low=0, high=5, size=(1,1,2,2)).astype(np.float32)))
inp = Variable(torch.from_numpy(np.random.randint(low=0, high=5, size=(1, 1,3,3)))).float()

n.weight

Parameter containing:
(0 ,0 ,.,.) = 
  4  2
  0  1
[torch.FloatTensor of size 1x1x2x2]

inp

Variable containing:
(0 ,0 ,.,.) = 
  1  1  3
  2  3  1
  4  0  4
[torch.FloatTensor of size 1x1x3x3]

n(inp)

Variable containing:
(0 ,0 ,.,.) = 
   9.1881  11.1881
  14.1881  18.1881
[torch.FloatTensor of size 1x1x2x2]

Where are the decimal places coming from? Is there something that I’m not getting here?

2

Don’t worry silly mistake… I didn’t account for the bias…

3

Don’t forget limited precision of floats. Try this plain python for example.

1.0 - 0.8 - 0.2
# Result: -5.551115123125783e-17