def scores(self, data_dir, fquery, freply, fqvocab, frvocab, init=False):
if not init:
self.init_model()
queries = data_helpers.load_file(data_dir, fquery)
replies = data_helpers.load_file(data_dir, freply)
data_size = len(queries)
qvocab = data_helpers.load_vocab(data_dir, fqvocab)
rvocab = data_helpers.load_vocab(data_dir, frvocab)
scores=[]
with self.session.as_default():
for query, reply in zip(queries, replies):
ql, qids = data_helpers.transform_to_id(qvocab, query,
self.qmax_length)
rl, rids = data_helpers.transform_to_id(rvocab, reply,
self.rmax_length)
feed_dict = self.make_input_feed([qids], [ql], [rids], [rl], training=False)
score = self.session.run(self.pos_score, feed_dict)
scores.append(score[0])
""" Debug
for i, s in enumerate(scores):
print(i,s)
"""
return scores
def get_scores(self, query_file, reply_file, query_vocab_file, reply_vocab_file, init=False):
if not init:
self.init_model()
queries = data_helpers.load_file(query_file)
replies = data_helpers.load_file(reply_file)
query_vocab = data_helpers.load_vocab(query_vocab_file)
reply_vocab = data_helpers.load_vocab(reply_vocab_file)
scores = []
logger.info('looping over query-reply pairs')
with self.session.as_default():
for query, reply in zip(queries, replies):
q_len, q_ids = data_helpers.transform_to_id(query_vocab, query, self.query_max_length)
r_len, r_ids = data_helpers.transform_to_id(reply_vocab, reply, self.reply_max_length)
feed_dict = self.make_input_feed([q_ids], [q_len], [r_ids], [r_len], training=False)
# When training=False there is no neg_score, so as pos_score.
score = self.session.run(self.score, feed_dict)
score = float(score[0])
scores.append(score)
return scores