def __init__(self,
config,
word_emb_layer,
char_emb_layer,
n_class,
use_cuda=False):
super(Model, self).__init__()
self.use_cuda = use_cuda
self.config = config
if config['token_embedder']['name'].lower() == 'cnn':
self.token_embedder = ConvTokenEmbedder(config, word_emb_layer,
char_emb_layer, use_cuda)
elif config['token_embedder']['name'].lower() == 'lstm':
self.token_embedder = LstmTokenEmbedder(config, word_emb_layer,
char_emb_layer, use_cuda)
if config['encoder']['name'].lower() == 'elmo':
self.encoder = ElmobiLm(config, use_cuda)
elif config['encoder']['name'].lower() == 'lstm':
self.encoder = LstmbiLm(config, use_cuda)
self.output_dim = config['encoder']['projection_dim']
if config['classifier']['name'].lower() == 'softmax':
self.classify_layer = SoftmaxLayer(self.output_dim, n_class)
elif config['classifier']['name'].lower() == 'cnn_softmax':
self.classify_layer = CNNSoftmaxLayer(
self.token_embedder, self.output_dim, n_class,
config['classifier']['n_samples'],
config['classifier']['corr_dim'], use_cuda)
elif config['classifier']['name'].lower() == 'sampled_softmax':
self.classify_layer = SampledSoftmaxLayer(
self.output_dim, n_class, config['classifier']['n_samples'],
use_cuda)
def __init__(self, config, word_emb_layer, char_emb_layer, use_cuda=False):
super(Model, self).__init__()
self.use_cuda = use_cuda
self.config = config
if config['token_embedder']['name'].lower() == 'cnn':
self.token_embedder = ConvTokenEmbedder(config, word_emb_layer, char_emb_layer, use_cuda)
elif config['token_embedder']['name'].lower() == 'lstm':
self.token_embedder = LstmTokenEmbedder(config, word_emb_layer, char_emb_layer, use_cuda)
if config['encoder']['name'].lower() == 'elmo':
self.encoder = ElmobiLm(config, use_cuda)
elif config['encoder']['name'].lower() == 'lstm':
self.encoder = LstmbiLm(config, use_cuda)
self.output_dim = config['encoder']['projection_dim']