char_decoder.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

"""
CS224N 2018-19: Homework 5
"""

import torch
import torch.nn as nn

class CharDecoder(nn.Module):
    def __init__(self, hidden_size, char_embedding_size=50, target_vocab=None):
        """ Init Character Decoder.

        @param hidden_size (int): Hidden size of the decoder LSTM
        @param char_embedding_size (int): dimensionality of character embeddings
        @param target_vocab (VocabEntry): vocabulary for the target language. See vocab.py for documentation.
        """
        ### YOUR CODE HERE for part 2a
        ### TODO - Initialize as an nn.Module.
        ###      - Initialize the following variables:
        ###        self.charDecoder: LSTM. Please use nn.LSTM() to construct this.
        ###        self.char_output_projection: Linear layer, called W_{dec} and b_{dec} in the PDF
        ###        self.decoderCharEmb: Embedding matrix of character embeddings
        ###        self.target_vocab: vocabulary for the target language
        ###
        ### Hint: - Use target_vocab.char2id to access the character vocabulary for the target language.
        ###       - Set the padding_idx argument of the embedding matrix.
        ###       - Create a new Embedding layer. Do not reuse embeddings created in Part 1 of this assignment.
        super(CharDecoder, self).__init__()

        self.hidden_size = hidden_size
        self.char_embedding_size = char_embedding_size
        self.target_vocab = target_vocab

        self.decoderCharEmb = nn.Embedding(len(self.target_vocab.char2id), self.char_embedding_size, padding_idx = self.target_vocab.char2id['<pad>'])
        self.charDecoder = nn.LSTM(self.char_embedding_size, self.hidden_size)
        self.char_output_projection = nn.Linear(self.hidden_size, len(self.target_vocab.char2id))
        ### END YOUR CODE

    
    def forward(self, input, dec_hidden=None):
        """ Forward pass of character decoder.

        @param input: tensor of integers, shape (length, batch)
        @param dec_hidden: internal state of the LSTM before reading the input characters. A tuple of two tensors of shape (1, batch, hidden_size)

        @returns scores: called s_t in the PDF, shape (length, batch, self.vocab_size)
        @returns dec_hidden: internal state of the LSTM after reading the input characters. A tuple of two tensors of shape (1, batch, hidden_size)
        """
        ### YOUR CODE HERE for part 2b
        ### TODO - Implement the forward pass of the character decoder.

        char_embeddings = self.decoderCharEmb(input.permute(1,0)).permute(1,0,2)
        hidden_states, dec_hidden = self.charDecoder(char_embeddings, dec_hidden)
        scores = self.char_output_projection(hidden_states)

        return scores, dec_hidden


    def train_forward(self, char_sequence, dec_hidden=None):
        """ Forward computation during training.

        @param char_sequence: tensor of integers, shape (length, batch). Note that "length" here and in forward() need not be the same.
        @param dec_hidden: initial internal state of the LSTM, obtained from the output of the word-level decoder. A tuple of two tensors of shape (1, batch, hidden_size)

        @returns The cross-entropy loss, computed as the *sum* of cross-entropy losses of all the words in the batch.
        """
        ### YOUR CODE HERE for part 2c
        ### TODO - Implement training forward pass.
        ###
        ### Hint: - Make sure padding characters do not contribute to the cross-entropy loss.
        ###       - char_sequence corresponds to the sequence x_1 ... x_{n+1} from the handout (e.g., <START>,m,u,s,i,c,<END>).

        char_input = char_sequence[:-1,:]

        scores, _ = self.forward(char_input, dec_hidden)
        scores = scores.permute(1,2,0)

        target = char_sequence[1:,:].permute(1,0)

        loss = nn.functional.cross_entropy(scores, target, ignore_index=self.target_vocab.char2id['<pad>'], reduction='sum')

        return loss
        ### END YOUR CODE

    def decode_greedy(self, initialStates, device, max_length=21):
        """ Greedy decoding
        @param initialStates: initial internal state of the LSTM, a tuple of two tensors of size (1, batch, hidden_size)
        @param device: torch.device (indicates whether the model is on CPU or GPU)
        @param max_length: maximum length of words to decode

        @returns decodedWords: a list (of length batch) of strings, each of which has length <= max_length.
                              The decoded strings should NOT contain the start-of-word and end-of-word characters.
        """

        ### YOUR CODE HERE for part 2d
        ### TODO - Implement greedy decoding.
        ### Hints:
        ###      - Use target_vocab.char2id and target_vocab.id2char to convert between integers and characters
        ###      - Use torch.tensor(..., device=device) to turn a list of character indices into a tensor.
        ###      - We use curly brackets as start-of-word and end-of-word characters. That is, use the character '{' for <START> and '}' for <END>.
        ###        Their indices are self.target_vocab.start_of_word and self.target_vocab.end_of_word, respectively.

        _, bs, _ = initialStates[0].shape

        output_words = [["", False] for _ in range(bs)]

        current_chars = torch.ones([1, bs], dtype = torch.long, device = device) * self.target_vocab.start_of_word

        for t in range(max_length):
            scores, initialStates = self.forward(current_chars, initialStates)
            current_chars = torch.argmax(scores, dim=-1)
        
            for word_idx, char_idx in enumerate(current_chars.squeeze(0).detach()):
                if output_words[word_idx][1] == False:
                    if char_idx == self.target_vocab.end_of_word:
                        output_words[word_idx][1] = True
                    else:
                        output_words[word_idx][0] += self.target_vocab.id2char[char_idx.item()]

        output_words = [ word for word, _ in output_words]
        
        return output_words
        ### END YOUR CODE