def __init__(self, vocab_num, embed_dim=100, bigram_vocab_num=None, bigram_embed_dim=100, num_bigram_per_char=None,
hidden_size=200, bidirectional=True, embed_drop_p=0.2, num_layers=1, tag_size=4):
"""
默认使用BMES的标注方式
:param vocab_num:
:param embed_dim:
:param bigram_vocab_num:
:param bigram_embed_dim:
:param num_bigram_per_char:
:param hidden_size:
:param bidirectional:
:param embed_drop_p:
:param num_layers:
:param tag_size:
"""
super(CWSBiLSTMCRF, self).__init__()
self.tag_size = tag_size
self.encoder_model = CWSBiLSTMEncoder(vocab_num, embed_dim, bigram_vocab_num, bigram_embed_dim, num_bigram_per_char,
hidden_size, bidirectional, embed_drop_p, num_layers)
size_layer = [hidden_size, 200, tag_size]
self.decoder_model = MLP(size_layer)
allowed_trans = allowed_transitions({0:'b', 1:'m', 2:'e', 3:'s'}, encoding_type='bmes')
self.crf = ConditionalRandomField(num_tags=tag_size, include_start_end_trans=False,
allowed_transitions=allowed_trans)
def __init__(self, args=None):
super(SELF_ATTENTION_YELP_CLASSIFICATION, self).__init__()
self.embedding = Embedding((len(word2index), embeding_size))
self.lstm = LSTM(input_size=embeding_size,
hidden_size=lstm_hidden_size,
bidirectional=True)
self.attention = SelfAttention(lstm_hidden_size * 2,
dim=attention_unit,
num_vec=attention_hops)
self.mlp = MLP(
size_layer=[lstm_hidden_size * 2 * attention_hops, nfc, class_num])
def __init__(self, vocab_num, embed_dim=100, bigram_vocab_num=None, bigram_embed_dim=100, num_bigram_per_char=None,
hidden_size=200, bidirectional=True, embed_drop_p=None, num_layers=1, tag_size=2):
super(CWSBiLSTMSegApp, self).__init__()
self.tag_size = tag_size
self.encoder_model = CWSBiLSTMEncoder(vocab_num, embed_dim, bigram_vocab_num, bigram_embed_dim, num_bigram_per_char,
hidden_size, bidirectional, embed_drop_p, num_layers)
size_layer = [hidden_size, 200, tag_size]
self.decoder_model = MLP(size_layer)
def __init__(self, init_embed,
num_classes,
hidden_dim=256,
num_layers=1,
attention_unit=256,
attention_hops=1,
nfc=128):
super(BiLSTM_SELF_ATTENTION,self).__init__()
self.embed = get_embeddings(init_embed)
self.lstm = LSTM(input_size=self.embed.embedding_dim, hidden_size=hidden_dim, num_layers=num_layers, bidirectional=True)
self.attention = SelfAttention(input_size=hidden_dim * 2 , attention_unit=attention_unit, attention_hops=attention_hops)
self.mlp = MLP(size_layer=[hidden_dim* 2*attention_hops, nfc, num_classes])
def __init__(self,
init_embed,
num_classes,
hidden_dim=256,
num_layers=1,
nfc=128):
super(BiLSTMSentiment, self).__init__()
self.embed = encoder.Embedding(init_embed)
self.lstm = LSTM(input_size=self.embed.embedding_dim,
hidden_size=hidden_dim,
num_layers=num_layers,
bidirectional=True)
self.mlp = MLP(size_layer=[hidden_dim * 2, nfc, num_classes])
def __init__(self,
init_embed,
num_classes,
hidden_dim=256,
num_layers=1,
nfc=128,
wdrop=0.5):
super(AWDLSTMSentiment, self).__init__()
self.embed = get_embeddings(init_embed)
self.lstm = LSTM(input_size=self.embed.embedding_dim,
hidden_size=hidden_dim,
num_layers=num_layers,
bidirectional=True,
wdrop=wdrop)
self.mlp = MLP(size_layer=[hidden_dim * 2, nfc, num_classes])