SourKream

class multiAttentionRNN(Recurrent):
    '''
    Word by Word attention model 
    # Arguments
        output_dim: output_dimensions
        init: weight initialization function.
            Can be the name of an existing function (str),
            or a Theano function (see: [initializations](../initializations.md)).
        inner_init: initialization function of the inner cells.