RumourDetectionTwitterConfig

class RumourDetectionTwitterConfig(num_classes=4, max_vocab=15000, emb_dim=300, num_structure_index=5, include_key_structure=True, include_val_structure=True, word_module_version=4, post_module_version=3, train_word_emb=False, train_pos_emb=False, size=100, interval=10, include_time_interval=True, max_length=35, max_tweets=339, d_model=300, dropout_rate=0.3, ff_word=True, num_emb_layers_word=2, n_mha_layers_word=2, n_head_word=2, ff_post=True, num_emb_layers=2, n_mha_layers=12, n_head=2, d_feed_forward=600, gpu=False, gpu_idx=[0], main_gpu=[0], initializer_range=0.02, loss='cross_entropy', **kwargs)

This is the configuration class to store the configuration of RumourDetectionTwitterModel. It is used to instantiate a Rumour Detection model.

Parameters

• num_classes (int, optional) – number of classers predicted by the model. Defaults to 4.

• max_vocab (int, optional) – vocabulary size. Defaults to 15000.

• emb_dim (int, optional) – size of each token embedding vector. Defaults to 300.