Recursive function to encapsulate RNN calls. :param rnn: :param embedded_sequence_batch: :param lengths: :param hidden_tuple:

Get a yacs CfgNode object with default values for my_project.

Gets you a torch tensor with the embeddings in the indices given by self.vocab. Unknown (unseen) words are each mapped to a

Gets you a torch tensor with the embeddings in the indices given by self.vocab. Unknown (unseen) words are each mapped to a

:param batch_hidden_states: torch.Tensor(batch_size, seq_len, hidden_size) :return:

:param batch_hidden_states: torch.Tensor(batch_size, seq_len, hidden_size) :param batch_lengths: list(batch_size) :return:

:param embedded_sequence_batch: torch.Tensor(batch_size, seq_len) :param sequence_lengths: list(batch_size) :return:

:param packed_rnn_output: torch object :param indices: Variable(LongTensor) of indices to sort output :return:

:param sentences: :param token_function: :param min_count: :param add_padding: :param add_bos: :param

This function compute the KL divergency between gt localization and prediction. Parameters ---------- gt: pred_local: le