Stance-Detection-in-Web-and-Social-Media/TAN

networks.py

class LSTM_TAN(nn.Module):

def __init__(self,version,embedding_dim, hidden_dim, vocab_size, n_targets,embedding_matrix,dropout = 0.5):

Args

1. version - str - one of ["lstm,"tan-","tan"]
2. embedding_dim - int - dimension of word_embedding
3. hidden_dim - int - dimension of LSTM hidden state
4. vocab_size - int - number of words in the vocabulary
5. n_targets - int - number of dataset classes
6. embedding_matrix - numpy array dtype=float - word embedding matrix
7. dropout - The dropout to be applied before on the final hidden state(lstm)/attention-weighted hidden state (tan-,tan)

Returns A torch.nn.module object for the specified version

def forward(self, sentence, target,verbose=False)

Args

1. sentence - a numpy array of shape [1xN] and dtype int, where N is the length of the input sentence and each entry is the corresponding index of the word in the embedding_matrix
2. target - a numpy array of shape [1xM] and dtype int, where M is the length of the target and each entry is the corresponding index of the word in the `embedding_matrix

Returns

1. target_scores - a torch float Tensor of shape [1xn_targets], where N is the number of dataset classes. This is the log likelihood probabilities of all the classes

Running the code

To run the code you have to run get_training_plots.py or early_stopping_training.py
To change models, call the classes with specific versions as mentioned above