RNN_Sent140#

class fl_sim.models.RNN_Sent140(latent_size: int = 100, num_classes: int = 2, num_layers: int = 2, embedding: str | object = 'glove.6B.50d')[source]#

Bases: Module, CLFMixin, SizeMixin, DiffMixin

Stacked LSTM model for sentiment analysis on the Sent140 dataset.

Adapted from FedProx/flearn/models/sent140/stacked_lstm.py [1].

Parameters:

  • latent_size (int, default 100) – The number of features in the hidden state h.

  • num_classes (int, default 2) – The number of output classes.

  • num_layers (int, default 2) – The number of recurrent layers (LSTM).

  • embedding (str or GloveEmbedding, default “glove.6B.50d”) – The name of the pre-trained GloVe embedding to use or a GloveEmbedding object.

References

forward(input_seq: Tensor) → Tensor[source]#

Forward pass.

Parameters:

input_seq (torch.Tensor) – Shape (batch_size, seq_len), dtype torch.long.

Returns:

Shape (batch_size, num_classes), dtype torch.float32.

Return type:

torch.Tensor

pipeline(sentence: str) → int[source]#

Predict the sentiment of a sentence.

Parameters:

sentence (str) – The sentence to predict the sentiment of.

Returns:

The class index of the predicted sentiment.

Return type:

int