Backend: Model Implementation

[AngelinaZhai]

Objective: Create definitions of classes for different types of text processing deep learning networks. We will be comparing the results generated by each model near the end and will talk about the comparison in our documentation near the end.

Requirements:

• Use Sigmoid activation function
• Make sure to understand all components within the model and be able to explain the rationales behind your design choices
  • Components present within network:
  • embedding layer
  • hidden state
  • model module (nn.rnn, nn.lstm, etc)
  • Sigmoid activation function (last layer)
• See tutorial 4 code for input and output parameters

Tasks:

• Design class objects for: (the difference between all models should just differ by one line of code, all built in within PyTorch nn)
  • [x] RNN
  • [x] BiRNN
  • [x] LSTM
  • [x] BiLSTM
  • [x] GRU
• For each of the above, make sure you have the following word embedding options (write them in comments)
  • [x] Glove
  • [x] Word2Vec
  • [x] No pre-trained embeddings

    self.emb = nn.Embedding(vocab_size, vector_size)