RstPointerSegmenterConfig

class RstPointerSegmenterConfig(word_dim=1024, hidden_dim=64, dropout_prob=0.2, use_bilstm=True, num_rnn_layers=6, rnn_type='GRU', is_batch_norm=True, elmo_size='Large', **kwargs)[source]

This is the configuration class to store the configuration of a RstPointerSegmenterModel. It is used to instantiate a discourse segmenter pointer network according to the specified arguments, defining the model architecture.

Configuration objects inherit from PretrainedConfig and can be used to control the model outputs. Read the documentation from PretrainedConfig for more information.

Parameters

  • word_dim (int, defaults to 1024) – Word embedding dimension size.

  • hidden_dim (int, defaults to 64) – Hidden dimension zie.

  • dropout_prob (float, defaults to 0.2) – Dropout probability.

  • use_bilstm (bool, defaults to True) – Whether to use bilstm layer.

  • num_rnn_layers (int, defaults to 6) – Number of RNN layers.

  • rnn_type (str, defaults to “GRU”) – RNN type. Supported choices: [“LSTM”, “GRU”].

  • is_batch_norm (bool, defaults to True) – Whether to use batch normalization.

  • elmo_size (bool, defaults to “Large”) – Elmo size. Supported choices: [“Large”, “Medium”, “Small”].

Example:

from sgnlp.models.rst_pointer import RstPointerSegmenterConfig

# Initialize with default values
config = RstPointerSegmenterConfig()