Args: base_target_emb: target emb tensor input_from_dec: output of decode conv encoder_out_top: the key m

Compute the context vector and the attention vectors. Args: key (FloatTensor): set of `key_len` key vector

Compute a distribution over the target dictionary extended by the dynamic dictionary implied by copying source words.

Args: scores (FloatTensor): ``(batch_size*tgt_len)`` x dynamic vocab size whose sum along dim 1 is less than or e

Embed inputs. Args: emb (FloatTensor): Sequence of word vectors ``(seq_len, batch_size, self.dim)`` s

Args: x (FloatTensor): input, ``(len, batch, 1, vec_feats)``. Returns: FloatTensor: embedded vecs ``(len, ba

Computes the embeddings for words and features. Args: source (LongTensor): index tensor ``(len, batch, nfeat)`` Returns:

Layer definition. Args: x: ``(batch_size, input_len, model_dim)`` Returns: (FloatTensor): Output ``(batch_si

output (FloatTensor): batch_size x n_classes target (LongTensor): batch_size

See :func:`EncoderBase.forward()`

See :func:`onmt.encoders.encoder.EncoderBase.forward()`

See :func:`onmt.encoders.encoder.EncoderBase.forward()`

Args: inputs (FloatTensor): ``(batch_size, src_len, model_dim)`` mask (LongTensor): ``(batch_size, 1, src_len)``

See :func:`EncoderBase.forward()`

Args: src (LongTensor): padded sequences of sparse indices ``(src_len, batch, nfeat)`` lengths (LongTe

See :func:`EncoderBase.forward()`

See :class:`onmt.modules.EncoderBase.forward()`

Args: tgt (LongTensor): sequences of padded tokens ``(tgt_len, batch, nfeats)``. memory_bank (FloatT

See :obj:`onmt.modules.RNNDecoderBase.forward()`

Extend _forward for (possibly) multiple decoder pass: 1. Always a default (future masked) decoder forward pass, 2. Possibly a second

Decode, possibly stepwise.

See :func:`onmt.decoders.decoder.DecoderBase.forward()`.

Compute a distribution over the target dictionary by averaging distributions from models in the ensemble. All models in the e

Forward propagate a `src` and `tgt` pair for training. Possible initialized with a beginning decoder state. Args: src (Te

Alternate constructor.

Alternate constructor.

Alternate constructor.

Alternate constructor.

Alternate constructor.

Alternate constructor.

Alternative constructor. Args: opt (argparse.Namespace): The parsed arguments.

Alternate constructor. Subclasses should override this method.

Alternate constructor.

Alternate constructor.

Alternate constructor.

Sort using the size of the image: (width, height).

Init decoder state.

Initialize decoder state.

See :obj:`RNNDecoderBase.init_state()`

DecodeStrategy subclasses should override :func:`initialize()`. `initialize` should be called before all actions. used to prepare nec

Initialize for decoding.

Returns the current sequence length.

Returns unmodified scores.

GNMT length re-ranking score. See "Google's Neural Machine Translation System" :cite:`wu2016google`.

Pass in needed options only when modify function definition.

?

Convert a sparse index_tensor into a batch of alignment matrix, with row normalize to the sum of 1 if set normalize. Args: index

In token batching scheme, the number of sequences is limited such that the total number of src/tgt tokens (including padding) in a batch

In token batching scheme, the number of sequences is limited such that the total number of src/tgt tokens (including padding) in a batch

These options are passed to the construction of the model. Be careful with these as they will be used during translation.

Learning rate schedule described in https://arxiv.org/pdf/1706.03762.pdf.

Learning rate schedule optimized for huge batches

Turn a batch of examples that use this field into a Variable. If the field has ``include_lengths=True``, a tensor of lengths will be

Turn a batch of examples that use this field into a Variable. If the field has ``include_lengths=True``, a tensor of lengths will be i

Pad a batch of examples to the length of the longest example. Args: minibatch (List[torch.FloatTensor]): A list of audio data,

Pad a batch of examples to the length of the longest example. Args: minibatch (List[torch.FloatTensor]): A list of audio data,

Parse Pharaoh alignment into [[<src>, <tgt>], ...]