class Model_LSTM(nn.Module):
def __init__(self, embedding_matrix, num_sub_targets):
super(Model, self).__init__()
self.embedding = Embedding(embedding_matrix)
self.encoder = Encoder(num_sub_targets)
def forward(self, x):
x = self.embedding(x)
out = self.encoder(x)
return out
crawl_matrix, unknown_words_crawl = build_matrix(
tokenizer.word_index, CRAWL_EMBEDDING_PATH
)
glove_matrix, unknown_words_glove = build_matrix(
tokenizer.word_index, GLOVE_EMBEDDING_PATH
)
>>> print("n unknown words (crawl): ", len(unknown_words_crawl))
>>> print("n unknown words (glove): ", len(unknown_words_glove))
n unknown words (crawl): 140854
n unknown words (glove): 143012
embedding_matrix = np.concatenate([crawl_matrix, glove_matrix], axis=-1)
>>> print(embedding_matrix.shape)
(416731, 600)