Is it possible to support Poly BCEloss ?

[yiyixuxu]

Hi:

I added it, check it out

# Example of PolyBCELoss
from PolyLoss import PolyBCELoss
loss = PolyBCELoss()
B, H, W = 2, 3, 3
input = torch.rand(B, H, W, requires_grad=True)
target = torch.empty(B,H,W).random_(2)
output = loss(input, target)
output.backward()

[zyxu1996]

Kindly thank you for supporting PolyBCELoss. This is exactly the same way I write it, so I think this is a correct implementation.