Encoder

class opennmt.encoders.Encoder(*args, **kwargs)

Base class for encoders.

Inherits from: keras.src.engine.base_layer.Layer

Extended by:

• opennmt.encoders.ConvEncoder

• opennmt.encoders.LSTMEncoder

• opennmt.encoders.MeanEncoder

• opennmt.encoders.ParallelEncoder

• opennmt.encoders.PyramidalRNNEncoder

• opennmt.encoders.RNNEncoder

• opennmt.encoders.SelfAttentionEncoder

• opennmt.encoders.SequentialEncoder

build_mask(inputs, sequence_length=None, dtype=tf.bool)

Builds a boolean mask for inputs.

abstract call(inputs, sequence_length=None, training=None)

Encodes an input sequence.

Parameters

• inputs – The inputs to encode of shape \([B, T, ...]\).

• sequence_length – The length of each input with shape \([B]\).

• training – Run in training mode.

Returns

A tuple (outputs, state, sequence_length).

__call__(inputs, sequence_length=None, **kwargs)

Encodes an input sequence.

Parameters

• inputs – A 3D tf.Tensor or tf.RaggedTensor.

• sequence_length – A 1D tf.Tensor (optional if inputs is a tf.RaggedTensor).

• training – Run the encoder in training mode.

Returns

If inputs is a tf.Tensor, the encoder returns a tuple (outputs, state, sequence_length). If inputs is a tf.RaggedTensor, the encoder returns a tuple (outputs, state), where outputs is a tf.RaggedTensor.