def train_model(self, max_epoch=30):
evaluator = Evaluator(true_labels=self.test_labels,
sentences=self.test_sentence_words_input,
position_mt=self.test_position_mt,
position_me=self.test_position_ma,
correction_factor=self.correction_factor)
log = open("../log/" + self.name + ".txt", 'a+', encoding='utf-8')
for i in range(max_epoch):
self.model.fit(
{
'sentence_word': self.train_sentence_words_input,
# 'sentence_entity_type': self.train_sentence_entity_inputs,
'position_t': self.train_position_t,
'position_a': self.train_position_a,
},
self.train_labels,
epochs=1,
batch_size=256,
verbose=1)
print("# -- test set --- #")
results = self.model.predict(
{
'sentence_word': self.test_sentence_words_input,
# 'sentence_entity_type': self.test_sentence_entity_inputs,
'position_t': self.test_position_t,
'position_a': self.test_position_a,
},
batch_size=128,
verbose=0)
print("--------------epoch " + str(i + 1) +
" ---------------------")
macro_f1, micro_F1, p, r = evaluator.get_f1(predictions=results,
epoch=i + 1)
log.write("epoch: " + str(i + 1) + " " + str(p) + " " + str(r) +
" " + str(micro_F1) + "\n")
if (i + 1) % 5 == 0:
print("current max macro_F1 score: " +
str(evaluator.max_macro_F1 * 100))
print("max macro_F1 is gained in epoch " +
str(evaluator.max_macro_F1_epoch))
print("current max micro_F1 score: " +
str(evaluator.max_micro_F1 * 100))
print("max micro_F1 is gained in epoch " +
str(evaluator.max_micro_F1_epoch))
log.write("current max macro_F1 score: " +
str(evaluator.max_macro_F1 * 100) + "\n")
log.write("max macro_F1 is gained in epoch " +
str(evaluator.max_macro_F1_epoch) + "\n")
log.write("current max micro_F1 score: " +
str(evaluator.max_micro_F1 * 100) + "\n")
log.write("max micro_F1 is gained in epoch " +
str(evaluator.max_micro_F1_epoch) + "\n")
print(
"------------------------------------------------------------")
log.close()
position_ids = example_reader.get_position_ids(max_len=82)
example_reader.convert_position(position_inputs=train_positions, position_ids=position_ids)
example_reader.convert_position(position_inputs=test_positions, position_ids=position_ids)
train_aspects = example_reader.pad_aspect_index(train_aspect_text_inputs.tolist(), max_length=9)
test_aspects = example_reader.pad_aspect_index(test_aspect_text_inputs.tolist(), max_length=9)
# ---------------------------------------------------------------------
for i in range(5):
model = m.build_model(max_len=82,
aspect_max_len=9,
embedding_matrix=embedding_matrix,
position_embedding_matrix=position_matrix,
class_num=3,
num_words=4582) # 5144 4582 // # 1523 // 1172
evaluator = Evaluator(true_labels=test_true_labels, sentences=test_sentence_inputs, aspects=test_aspect_text_inputs)
epoch = 1
while epoch <= 80:
model = m.train_model(sentence_inputs=train_sentence_inputs,
position_inputs=train_positions,
aspect_input=train_aspects,
labels=train_aspect_labels,
model=model)
results = m.get_predict(sentence_inputs=test_sentence_inputs,
position_inputs=test_positions,
aspect_input=test_aspects,
model=model)
print("\n--------------epoch " + str(epoch) + " ---------------------")
F, acc = evaluator.get_macro_f1(predictions=results, epoch=epoch)
if epoch % 5 == 0:
print("current max F1 score: " + str(evaluator.max_F1))
def train_model(self, max_epoch=30):
e1 = Evaluator(true_labels=self.test_labels,
sentences=self.test_word_inputs,
index_ids=self.index_ids,
val=False)
if self.use_development_set:
e2 = Evaluator(true_labels=self.dev_labels,
sentences=self.dev_word_inputs,
index_ids=self.index_ids)
log = open("../log/" + self.name + ".txt", 'a+', encoding='utf-8')
for i in range(max_epoch):
print("====== epoch " + str(i + 1) + " ======")
# -------- if use development set, the early stop will be introduced.
if self.use_development_set:
self.model.fit(
{
'sentence_input': self.train_word_inputs,
'entity_type_input': self.train_entity_inputs
},
self.train_labels,
epochs=1,
batch_size=32,
validation_data=([
self.dev_word_inputs, self.dev_entity_inputs
], self.dev_labels),
verbose=2)
print("# -- develop set --- #")
results = self.model.predict(
{
'sentence_input': self.dev_word_inputs,
'entity_type_input': self.dev_entity_inputs
},
batch_size=64,
verbose=0)
results = e2.get_true_label(label=results)
results = e2.process_bie(sen_label=results)
f1, _, _ = e2.get_true_prf(results, epoch=i + 1)
if f1 < 0:
break
else:
self.model.fit(
{
'sentence_input': self.train_word_inputs,
'entity_type_input': self.train_entity_inputs
},
self.train_labels,
epochs=1,
batch_size=32,
verbose=2)
print("# -- test set --- #")
results = self.model.predict(
{
'sentence_input': self.test_word_inputs,
'entity_type_input': self.test_entity_inputs
},
batch_size=64,
verbose=0)
results = e1.get_true_label(label=results)
results = e1.process_bie(sen_label=results)
f1, p1, r1 = e1.get_true_prf(results, epoch=i + 1)
log.write("epoch:{} p:{} r:{} f:{}\n".format(i + 1, p1, r1, f1))
log.close()