ScalingNmtEnDe

class opennmt.models.ScalingNmtEnDe(*args, **kwargs)[source]

Defines a big Transformer model using the En-De hyperparameters from https://arxiv.org/abs/1806.00187.

The architecture is equivalent to transformer_wmt_en_de_big in Fairseq.

Inherits from: opennmt.models.Transformer

Extended by:

auto_config(num_replicas=1)[source]

Returns automatic configuration values specific to this model.

Parameters

num_replicas – The number of synchronous model replicas used for the training.

Returns

A partial training configuration.