Change the parameter type from int to float in torch.nn.Softplus (#120183)
Fixes #120175
1 The c_api uses the double
https://github.com/pytorch/pytorch/blob/f2cf0768d10df6eb578db734bf80f5281f7410ae/torch/csrc/api/include/torch/nn/options/activation.h#L501.
2 The type is also double in the test case
https://github.com/pytorch/pytorch/blob/f2cf0768d10df6eb578db734bf80f5281f7410ae/test/cpp/api/functional.cpp#L1788
3 With float parameter in python works perfectly fine
```
m = nn.Softplus(beta=0.1,threshold=1.2)
input = torch.randn(2)
output = m(input)
print(output)
tensor([7.3749, 7.6852])
```
Pull Request resolved: https://github.com/pytorch/pytorch/pull/120183
Approved by: https://github.com/mikaylagawarecki