#!/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.charDecoder = nn.LSTM(char_embedding_size,hidden_size,batch_first=True) #bias = True self.char_output_projection = nn.Linear(hidden_size,len(target_vocab.char2id)) self.decoderCharEmb = nn.Embedding(len(target_vocab.char2id),char_embedding_size,padding_idx=target_vocab.char2id['']) self.target_vocab = target_vocab ### 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. #print('size of input is',input.size()) input = input.permute(1,0).contiguous() ip_embedding=self.decoderCharEmb(input)# F.embedding(source_padded, self.model_embeddings.source.weight) #X = nn.utils.rnn.pack_padded_sequence(src_padded_embedding,source_lengths) #ip_embedding = ip_embedding.permute(1,0,2).contiguous() output,(h_n,c_n) = self.charDecoder(ip_embedding,dec_hidden) #print('shape of hidden is',h_n.size()) s_t = self.char_output_projection(output) #print('shape of logits is',s_t.size()) s_t = s_t.permute(1,0,2).contiguous() return s_t,(h_n,c_n) ### END YOUR CODE 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., ,m,u,s,i,c,). input = char_sequence[:-1,:] output = char_sequence[1:,:] #print(input) #print(output) target = output.reshape(-1) #print('shape of target',target.shape) s_t,(h_n,c_n) = self.forward(input,dec_hidden) #print('shape of s_t',s_t.shape) s_t_shape = s_t.shape s_t_re = s_t.reshape(-1,s_t.shape[2]) #print('shape of s_t_re',s_t_re.shape) loss = nn.CrossEntropyLoss(ignore_index=self.target_vocab.char2id[''],reduction='sum') return loss(s_t_re,target) ### 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 and '}' for . ### Their indices are self.target_vocab.start_of_word and self.target_vocab.end_of_word, respectively. decodedWords = [] current_char = self.target_vocab.start_of_word start_tensor = torch.tensor([current_char],device=device) #print('size of start_tensor is',start_tensor.shape) batch_size = initialStates[0].shape[1] start_batch = start_tensor.repeat(batch_size,1) #print('size of start_batch is',start_batch.shape) embed_current_char = self.decoderCharEmb(start_batch) #print('size of embed_current_char is',embed_current_char.shape) h_n,c_n = initialStates output_word = torch.zeros((batch_size,1),dtype=torch.long,device=device) for t in range(0,max_length): #h_n,c_n = self.charDecoder(embed_current_char,(h_n,c_n)) # s_t,(h_n,c_n) = self.forward(embed_current_char,(h_n,c_n)) #print('shape of embed_current_char is',embed_current_char.shape) output,(h_n,c_n) = self.charDecoder(embed_current_char,(h_n,c_n)) s_t = self.char_output_projection(output) #print(s_t.shape) st_smax = nn.Softmax(dim=2)(s_t) p_next = st_smax.argmax(2) current_char = p_next embed_current_char = self.decoderCharEmb(current_char) #decodedWords.append(self.target_vocab.id2char[current_char]) #print('*** size of current_char is',current_char.size()) output_word = torch.cat((output_word,current_char),1) #Convert output_word tensor to list and each element to char and put together in decodedWords out_list = output_word.tolist() out_list = [[self.target_vocab.id2char[x] for x in ilist[1:]] for ilist in out_list] decodedWords = [] for string_list in out_list: stringer = '' for char in string_list: if char!='}': stringer = stringer+char else: break decodedWords.append(stringer) return decodedWords ### END YOUR CODE