Input is expected to be of size [bsz x seqlen].

r""" masked_lm_labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`, defaults to :obj:`None`):

r""" masked_lm_labels (:obj:`torch.LongTensor` of shape :obj:`(batch_size, sequence_length)`, `optional`, defaults to :obj:`None`):

Args: x (Tensor): input to the layer of shape `(seq_len, batch, embed_dim)` encoder_padding_mask (ByteTensor): binary

Args: input_ids (LongTensor): tokens in the source language of shape `(batch, src_len)` attention_mas

Input shape: Time(SeqLen) x Batch x Channel

Input is expected to be of size [bsz x seqlen].

Input is expected to be of size [bsz x seqlen].

Implements Figure 2

Pass the input (and mask) through each layer in turn.

Apply residual connection to any sublayer with the same size.

Follow Figure 1 (left) for connections.

Implements Figure 2

Implements Figure 2

Pass the input (and mask) through each layer in turn.

Apply residual connection to any sublayer with the same size.

Follow Figure 1 (left) for connections.

The same interface with RAT return: encodings with size: length * embed_size, alignment matrix

Returns the device of the tensor.

Generate mask from variable ``(batch_size,)`` which represents the sequence lengths of each batch element. Returns: torch.Tensor: ``(batch_si

Go from token-level labels to list of entities (start, end, class).

Load state from disk.

A numerically stable computation of logsumexp. This is mathematically equivalent to `tensor.exp().sum(dim, keep=keepdim).log()`. This function i

Given a token, return a vector embedding if token is in vocabulary. If token is not in the vocabulary, then return None.

``torch.nn.functional.log_softmax(vector)`` does not work if some elements of ``vector`` should be masked. This performs a log_softmax on just t

``torch.nn.functional.softmax(vector)`` does not work if some elements of ``vector`` should be masked. This performs a softmax on just the non-ma

Run prediction and return predictions and potential metrics. Depending on the dataset and your use case, your test dataset may conta

Run prediction and return predictions and potential metrics. Depending on the dataset and your use case, your test dataset may conta

Print a 2D tensor

If distributed is initialized print only on rank 0.