models.rst

fairseq.models

Models

A Model defines the neural network's forward() method and encapsulates all of the learnable parameters in the network. Each model also provides a set of named architectures that define the precise network configuration (e.g., embedding dimension, number of layers, etc.).

Both the model type and architecture are selected via the --arch command-line argument. Once selected, a model may expose additional command-line arguments for further configuration.

Note

All fairseq Models extend BaseFairseqModel, which in turn extends torch.nn.Module. Thus any fairseq Model can be used as a stand-alone Module in other PyTorch code.

Convolutional Neural Networks (CNN)

fairseq.models.fconv

fairseq.models.fconv.FConvModel

fairseq.models.fconv.FConvEncoder

fairseq.models.fconv.FConvDecoder

Long Short-Term Memory (LSTM) networks

fairseq.models.lstm

fairseq.models.lstm.LSTMModel

fairseq.models.lstm.LSTMEncoder

fairseq.models.lstm.LSTMDecoder

Transformer (self-attention) networks

fairseq.models.transformer

fairseq.models.transformer.TransformerModel

fairseq.models.transformer.TransformerEncoder

fairseq.models.transformer.TransformerEncoderLayer

fairseq.models.transformer.TransformerDecoder

fairseq.models.transformer.TransformerDecoderLayer

Adding new models

fairseq.models

fairseq.models.register_model

fairseq.models.register_model_architecture

fairseq.models.BaseFairseqModel

fairseq.models.FairseqEncoderDecoderModel

fairseq.models.FairseqEncoderModel

fairseq.models.FairseqLanguageModel

fairseq.models.FairseqMultiModel

fairseq.models.FairseqEncoder

fairseq.models.CompositeEncoder

fairseq.models.FairseqDecoder

Incremental decoding

fairseq.models.FairseqIncrementalDecoder