MultiHeadAttention parameter setting

[LXXiaogege]

Is the output linear layer parameter of the MultiHeadAttention class incorrectly set in mha.py file? in_features should be heads*d_k?

[LXXiaogege]

The get_positional_encoding method of position encoder generates an error when d_model is set to odd

[vpj]

Our implementation assumes that heads * d_k = d_model. Need to change that