Tests if the embeddings works as expected args: opt: set of options sourceL: Length of generated input sente

Tests if the encoder works as expected args: opt: set of options sourceL: Length of generated input sentence

Args: input (LongTensor): len x batch x nfeat. lengths (LongTensor): batch hidden: Initial hidden state.

See EncoderBase.forward() for description of args and returns.

See EncoderBase.forward() for description of args and returns.

Forward through the decoder. Args: input (LongTensor): a sequence of input tokens tensors

See EncoderBase.forward() for description of args and returns.

Forward through the TransformerDecoder. Args: input (LongTensor): a sequence of input tokens tensors

Return the embeddings for words, and features if there are any. Args: input (LongTensor): len x batch x nfeat Ret

Computes p(w) = p(z=1) p_{copy}(w|z=0) + p(z=0) * p_{softmax}(w|z=0)

input (FloatTensor): batch x tgt_len x dim: decoder's rnn's output. context (FloatTensor): batch x src_len x dim: src hidden states

It's like Luong Attetion. Conv attention takes a key matrix, a value matrix and a query vector. Attention weight is calculate

See EncoderBase.forward() for description of args and returns.

Forward through the CNNDecoder. Args: input (LongTensor): a sequence of input tokens tensors

Conditional import of torch image libs.

Args: model_opt: the option loaded from checkpoint. fields: `Field` objects for the model. gpu(bool): whether to use gpu.

Args: batch (Variable): a batch of source or target data. side (str): for source or for target. Returns: A sequence o

See base class for args description.

See base class for args description.

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

Creates a ntmmodel with a custom opt function. Forwards a testbatch anc checks output size. Args: opt: Namespace

Step1: exact match context(temp) in temp_query_dict, if matched, refilling randomly and go to step3 Step2: Find the most similar context(temp

This prepare the data with following steps Step1: Tokenizing Step2: Filter the meaningless data Step3: Remove duplicate Step4: Bu

Repeat beam_size times along batch dimension.

Repeat beam_size times along batch dimension.

Sort in reverse size order