Removed the unnecessary softplus in _address_memory

Removed the softplus in the softmax:
```python
        s = F.softmax(F.softplus(s), dim=1)
```
softmax already constrains the values to (0, +inf), the softplus doesn't achieve much.

ntm/head.py

@@ -45,7 +45,7 @@ def _address_memory(self, k, β, g, s, γ, w_prev):
         k = k.clone()
         β = F.softplus(β)
         g = F.sigmoid(g)
-        s = F.softmax(F.softplus(s), dim=1)
+        s = F.softmax(s, dim=1)
         γ = 1 + F.softplus(γ)
 
         w = self.memory.address(k, β, g, s, γ, w_prev)