def __init__(self, opt, dicts):
self.layers = opt.layers
self.input_feed = opt.input_feed
input_size = opt.word_vec_size
if self.input_feed:
input_size += opt.rnn_size
super(TreeDecoder, self).__init__()
self.word_lut = nn.Embedding(dicts.size(), opt.word_vec_size, padding_idx=lib.Constants.PAD)
self.rnn = StackedLSTM(opt.layers, input_size, opt.rnn_size, opt.dropout)
if opt.has_attn:
self.attn = lib.GlobalAttention(opt.rnn_size)
self.dropout = nn.Dropout(opt.dropout)
self.hidden_size = opt.rnn_size
self.opt = opt
def __init__(self, opt, dicts):
self.layers = opt.layers
self.input_feed = opt.input_feed
input_size = opt.word_vec_size
if self.input_feed:
input_size += opt.rnn_size
super(TreeDecoder_W2V, self).__init__()
# self.word_lut = nn.Embedding(dicts.size(), opt.word_vec_size, padding_idx=lib.Constants.PAD)
self.embeddings = gensim.models.Word2Vec.load(opt.embedding_w2v + 'processed_all.train_xe.comment.gz')
self.rnn = StackedLSTM(opt.layers, input_size, opt.rnn_size, opt.dropout)
if opt.has_attn:
self.attn = lib.GlobalAttention(opt.rnn_size)
self.dropout = nn.Dropout(opt.dropout)
self.hidden_size = opt.rnn_size
self.opt = opt
self.dicts = dicts