device error in SinusoidalPositionalEmbedding (#746)

Summary:
Not sure if I'm doing something wrong elsewhere, but I had a device error in `SinusoidalPositionalEmbedding` when running on GPU > 0 because the weights were on a different device than the input.
Pull Request resolved: https://github.com/pytorch/fairseq/pull/746

Differential Revision: D15547217

Pulled By: myleott

fbshipit-source-id: 37849d895ce483c14615fdb4ace8a8c4fb05b568

fairseq/modules/sinusoidal_positional_embedding.py

@@ -65,7 +65,7 @@ class SinusoidalPositionalEmbedding(nn.Module):
                 self.embedding_dim,
                 self.padding_idx,
             )
-        self.weights = self.weights.type_as(self._float_tensor)
+        self.weights = self.weights.to(self._float_tensor)
 
         if incremental_state is not None:
             # positions is the same for every token when decoding a single step