def __init__(self,
N_word,
N_h=768,
N_depth=2,
gpu=False,
use_ca=True,
trainable_emb=False):
super(SQLNet, self).__init__()
self.use_ca = use_ca
self.trainable_emb = trainable_emb
self.gpu = gpu
self.N_h = N_h
self.N_depth = N_depth
self.max_col_num = 45
self.max_tok_num = 200
self.SQL_TOK = [
'<UNK>', '<END>', 'WHERE', 'AND', 'OR', '==', '>', '<', '!=',
'<BEG>'
]
self.COND_OPS = ['>', '<', '==', '!=']
# Predict the number of selected columns
self.sel_num = SelNumPredictor(N_word, N_h, N_depth, use_ca=use_ca)
#Predict which columns are selected
self.sel_pred = SelPredictor(N_word,
N_h,
N_depth,
self.max_tok_num,
use_ca=use_ca)
#Predict aggregation functions of corresponding selected columns
self.agg_pred = AggPredictor(N_word, N_h, N_depth, use_ca=use_ca)
#Predict number of conditions, condition columns, condition operations and condition values
self.cond_pred = SQLNetCondPredictor(N_word, N_h, N_depth,
self.max_col_num,
self.max_tok_num, use_ca, gpu)
# Predict condition relationship, like 'and', 'or'
self.where_rela_pred = WhereRelationPredictor(N_word,
N_h,
N_depth,
use_ca=use_ca)
self.CE = nn.CrossEntropyLoss()
self.softmax = nn.Softmax(dim=-1)
self.log_softmax = nn.LogSoftmax()
self.bce_logit = nn.BCEWithLogitsLoss()
if gpu:
self.cuda()
def __init__(self, word_emb, N_word, N_h=100, N_depth=2,
gpu=False, trainable_emb=False):
super(Seq2SQL, self).__init__()
self.trainable_emb = trainable_emb
self.gpu = gpu
self.N_h = N_h
self.N_depth = N_depth
self.max_col_num = 45
self.max_tok_num = 200
self.SQL_TOK = ['<UNK>', '<END>', 'WHERE', 'AND',
'EQL', 'GT', 'LT', '<BEG>']
self.COND_OPS = ['EQL', 'GT', 'LT']
#Word embedding
if trainable_emb:
self.agg_embed_layer = WordEmbedding(word_emb, N_word, gpu,
self.SQL_TOK, our_model=False,
trainable=trainable_emb)
self.sel_embed_layer = WordEmbedding(word_emb, N_word, gpu,
self.SQL_TOK, our_model=False,
trainable=trainable_emb)
self.cond_embed_layer = WordEmbedding(word_emb, N_word, gpu,
self.SQL_TOK, our_model=False,
trainable=trainable_emb)
else:
self.embed_layer = WordEmbedding(word_emb, N_word, gpu,
self.SQL_TOK, our_model=False,
trainable=trainable_emb)
#Predict aggregator
self.agg_pred = AggPredictor(N_word, N_h, N_depth, use_ca=False)
#Predict selected column
self.sel_pred = SelPredictor(N_word, N_h, N_depth, self.max_tok_num,
use_ca=False)
#Predict number of cond
self.cond_pred = Seq2SQLCondPredictor(
N_word, N_h, N_depth, self.max_col_num, self.max_tok_num, gpu)
self.CE = nn.CrossEntropyLoss()
self.softmax = nn.Softmax()
self.log_softmax = nn.LogSoftmax()
self.bce_logit = nn.BCEWithLogitsLoss()
if gpu:
self.cuda()
def __init__(self, N_word, N_h=512, N_depth=1, gpu=False, use_table=False,
word_emb=None, trainable_emb=False, bert_path=None):
super(SQLNet, self).__init__()
self.trainable_emb = trainable_emb
self.sample_data = False
self.gpu = gpu
self.N_h = N_h
self.N_depth = N_depth
self.use_table = use_table
self.max_col_num = 50
self.max_tok_num = 200
self.COND_OPS = ['>', '<', '==', '!=']
# Word embedding
if N_word == 300:
self.embed_layer = WordEmbedding(word_emb, N_word, gpu, our_model=True, trainable=trainable_emb)
else:
self.embed_layer = BertEmbedding(N_word, gpu, our_model=True, bert_path=bert_path)
print('Using Pre-trained BERT as Embedding')
# Predict the number of selected columns
self.sel_num = SelNumPredictor(N_word, N_h, N_depth)
# Predict which columns are selected
self.sel_pred = SelPredictor(N_word, N_h, N_depth, self.max_tok_num)
# Predict aggregation functions of corresponding selected columns
self.agg_pred = AggPredictor(N_word, N_h, N_depth)
# Predict number of conditions, condition columns, condition operations and condition values
self.cond_pred = SQLNetCondPredictor(N_word, N_h, N_depth, self.max_col_num, self.max_tok_num,
gpu, self.embed_layer, use_table)
# Predict condition relationship, like 'and', 'or'
self.where_rela_pred = WhereRelationPredictor(N_word, N_h, N_depth)
self.CE = nn.CrossEntropyLoss()
self.softmax = nn.Softmax(dim=-1)
self.log_softmax = nn.LogSoftmax()
self.bce_logit = nn.BCEWithLogitsLoss()
if gpu:
self.to('cuda')
if self.use_table:
print("using table content for condition value prediction")