def _create_tfrecord(self, sentences, task):
tf.logging.set_verbosity(tf.logging.INFO)
examples = []
i = 0
tokenizer = tokenization.FullTokenizer(vocab_file=self.vocab,
do_lower_case=self.lower)
for data in convert_sentences_to_features(sentences, self.data_dir,
tokenizer,
self.max_seq_length,
self.label_map):
examples.append(
single_example(i,
*data,
max_seq_length=self.max_seq_length,
mode=task,
data_dir=self.data_dir))
i += 1
sep = os.sep if os.sep in self.data_dir else "/"
if not os.path.exists(sep.join([self.data_dir, "tfrecord_data"])):
os.mkdir(sep.join([self.data_dir, "tfrecord_data"]))
output_file = sep.join([
self.data_dir, "tfrecord_data",
"{}_{}_{}.tfrecord".format(self.processor_data_name, task,
self.max_seq_length)
])
file_based_convert_examples_to_features(examples, output_file, task)
return len(examples)
def test_full_tokenizer(self):
vocab_tokens = [
"[UNK]", "[CLS]", "[SEP]", "want", "##want", "##ed", "wa", "un", "runn",
"##ing", ","
]
with tempfile.NamedTemporaryFile(delete=False) as vocab_writer:
vocab_writer.write("".join([x + "\n" for x in vocab_tokens]))
vocab_file = vocab_writer.name
tokenizer = tokenization.FullTokenizer(vocab_file)
os.unlink(vocab_file)
tokens = tokenizer.tokenize(u"UNwant\u00E9d,running")
self.assertAllEqual(tokens, ["un", "##want", "##ed", ",", "runn", "##ing"])
self.assertAllEqual(
tokenizer.convert_tokens_to_ids(tokens), [7, 4, 5, 10, 8, 9])
def _create_tfrecord(self, sentences, task, FLAGS):
tf.logging.set_verbosity(tf.logging.INFO)
sep = os.sep if os.sep in self.data_dir else "/"
top_token_num = None
if os.path.exists(sep.join([self.data_dir, "token_num.json"])):
with open(sep.join([self.data_dir, "token_num.json"]), "r") as f:
top_token_num = json.loads(f.read())
else:
top_token_num = {}
examples = []
i = 0
tokenizer = tokenization.FullTokenizer(vocab_file=self.vocab,
do_lower_case=self.lower)
# for data in convert_sentences_to_simple_features(sentences, self.data_dir, tokenizer, self.max_seq_length, self.label_map):
features = convert_sentences_to_features(
sentences, self.data_dir, tokenizer, self.max_seq_length,
self.label_map
) if FLAGS.data_mode == "combine" else convert_sentences_to_simple_features(
sentences, self.data_dir, tokenizer, self.max_seq_length,
self.label_map)
for data in features:
examples.append(
single_example(i,
*data,
label_map=self.label_map,
max_seq_length=self.max_seq_length,
mode=task,
data_dir=self.data_dir,
biobert=FLAGS.biobert))
get_token_num(top_token_num, data[2])
i += 1
if not os.path.exists(sep.join([self.data_dir, "tfrecord_data"])):
os.mkdir(sep.join([self.data_dir, "tfrecord_data"]))
output_file = sep.join([
self.data_dir, "tfrecord_data",
"{}_{}_{}.tfrecord".format(self.processor_data_name, task,
self.max_seq_length)
])
file_based_convert_examples_to_features(examples, output_file, task)
if not os.path.exists(sep.join([self.data_dir, "token_num.json"])):
with open(sep.join([self.data_dir, "token_num.json"]), "w") as f:
f.write(json.dumps(top_token_num))
return len(examples)
def initialize_config():
processors = {"newssentiment": NewsSentimentProcessor}
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
if not FLAGS.do_train and not FLAGS.do_eval and not FLAGS.do_predict:
raise ValueError(
"At least one of `do_train`, `do_eval` or `do_predict' must be True."
)
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
if FLAGS.max_seq_length > bert_config.max_position_embeddings:
raise ValueError(
"Cannot use sequence length %d because the BERT model "
"was only trained up to sequence length %d" %
(FLAGS.max_seq_length, bert_config.max_position_embeddings))
tf.gfile.MakeDirs(FLAGS.output_dir)
task_name = FLAGS.task_name.lower()
if task_name not in processors:
raise ValueError("Task not found: %s" % (task_name))
processor = processors[task_name]()
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
tpu_cluster_resolver = None
if FLAGS.use_tpu and FLAGS.tpu_name:
tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=FLAGS.master,
model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps,
tpu_config=tf.contrib.tpu.TPUConfig(
iterations_per_loop=FLAGS.iterations_per_loop,
num_shards=FLAGS.num_tpu_cores,
per_host_input_for_training=is_per_host))
return processor, tokenizer, bert_config, run_config
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
processors = {
"newssentiment": NewsSentimentProcessor,
}
tokenization.validate_case_matches_checkpoint(do_lower_case,
init_checkpoint)
bert_config = modeling.BertConfig.from_json_file(bert_config_file)
if max_seq_length > bert_config.max_position_embeddings:
raise ValueError(
"Cannot use sequence length %d because the BERT model "
"was only trained up to sequence length %d" %
(max_seq_length, bert_config.max_position_embeddings))
tf.gfile.MakeDirs(output_dir)
task_name1 = task_name.lower()
if task_name1 not in processors:
raise ValueError("Task not found: %s" % (task_name1))
processor = processors[task_name1]()
label_list = processor.get_labels()
tokenizer = tokenization.FullTokenizer(vocab_file=vocab_file,
do_lower_case=do_lower_case)
tpu_cluster_resolver = None
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=master,
model_dir=output_dir,
save_checkpoints_steps=save_checkpoints_steps,
tpu_config=tf.contrib.tpu.TPUConfig(
iterations_per_loop=iterations_per_loop,
num_shards=num_tpu_cores,
per_host_input_for_training=is_per_host))
num_train_steps = None
num_warmup_steps = None
model_fn = model_fn_builder(bert_config=bert_config,
num_labels=len(label_list),
init_checkpoint=init_checkpoint,
learning_rate=learning_rate,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
use_tpu=use_tpu,
use_one_hot_embeddings=use_tpu)
# If TPU is not available, this will fall back to normal Estimator on CPU
# or GPU.
estimator = tf.contrib.tpu.TPUEstimator(
use_tpu=use_tpu,
model_fn=model_fn,
config=run_config,
train_batch_size=train_batch_size,
eval_batch_size=eval_batch_size,
predict_batch_size=predict_batch_size)
if do_predict:
print('do_predict')
predict_examples = processor.get_test_examples(data_dir)
num_actual_predict_examples = len(predict_examples)
print('**num_actual_predict_examples', num_actual_predict_examples)
predict_file = os.path.join(output_dir, "predict.tf_record")
file_based_convert_examples_to_features(predict_examples, label_list,
max_seq_length, tokenizer,
predict_file)
tf.logging.info("***** Running prediction*****")
tf.logging.info(" Num examples = %d (%d actual, %d padding)",
len(predict_examples), num_actual_predict_examples,
len(predict_examples) - num_actual_predict_examples)
tf.logging.info(" Batch size = %d", predict_batch_size)
predict_drop_remainder = True if use_tpu else False
predict_input_fn = file_based_input_fn_builder(
input_file=predict_file,
seq_length=max_seq_length,
is_training=False,
drop_remainder=predict_drop_remainder)
print("***** Predict results *****")
result = estimator.predict(input_fn=predict_input_fn)
ids = _read_ids()
for (id, prediction) in tqdm(zip(ids, result)):
output = {}
# print(i+1,prediction["probabilities"][0])
if prediction["probabilities"][0] > 0.15:
output['sentiment'] = '负面'
else:
output['sentiment'] = '非负面'
output['sentiment_prob'] = str(prediction["probabilities"][0])
to_sql_com_name(json.dumps(output, ensure_ascii=False), id)
# print("\t".join(str(class_probability) for class_probability in probabilities) + "\n")
print('complete...')
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
processors = {"cola": ColaProcessor}
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
if not FLAGS.do_train and not FLAGS.do_eval and not FLAGS.do_predict:
raise ValueError(
"At least one of `do_train`, `do_eval` or `do_predict' must be True."
)
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
if FLAGS.max_seq_length > bert_config.max_position_embeddings:
raise ValueError(
"Cannot use sequence length %d because the BERT model "
"was only trained up to sequence length %d" %
(FLAGS.max_seq_length, bert_config.max_position_embeddings))
tf.gfile.MakeDirs(FLAGS.output_dir)
task_name = FLAGS.task_name.lower()
if task_name not in processors:
raise ValueError("Task not found: %s" % (task_name))
processor = processors[task_name]()
label_list = processor.get_labels()
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
tpu_cluster_resolver = None
if FLAGS.use_tpu and FLAGS.tpu_name:
tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=FLAGS.master,
model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps,
tpu_config=tf.contrib.tpu.TPUConfig(
iterations_per_loop=FLAGS.iterations_per_loop,
num_shards=FLAGS.num_tpu_cores,
per_host_input_for_training=is_per_host))
train_examples = None
num_train_steps = None
num_warmup_steps = None
if FLAGS.do_train:
train_examples = processor.get_train_examples(FLAGS.data_dir)
num_train_steps = int(
len(train_examples) / FLAGS.train_batch_size *
FLAGS.num_train_epochs)
num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)
model_fn = model_fn_builder(bert_config=bert_config,
num_labels=len(label_list),
init_checkpoint=FLAGS.init_checkpoint,
learning_rate=FLAGS.learning_rate,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
use_tpu=FLAGS.use_tpu,
use_one_hot_embeddings=FLAGS.use_tpu)
# If TPU is not available, this will fall back to normal Estimator on CPU
# or GPU.
estimator = tf.contrib.tpu.TPUEstimator(
use_tpu=FLAGS.use_tpu,
model_fn=model_fn,
config=run_config,
train_batch_size=FLAGS.train_batch_size,
eval_batch_size=FLAGS.eval_batch_size,
predict_batch_size=FLAGS.predict_batch_size)
if FLAGS.do_train:
train_file = os.path.join(FLAGS.output_dir, "train.tf_record")
file_based_convert_examples_to_features(train_examples, label_list,
FLAGS.max_seq_length,
tokenizer, train_file)
tf.logging.info("***** Running training *****")
tf.logging.info(" Num examples = %d", len(train_examples))
tf.logging.info(" Batch size = %d", FLAGS.train_batch_size)
tf.logging.info(" Num steps = %d", num_train_steps)
train_input_fn = file_based_input_fn_builder(
input_file=train_file,
seq_length=FLAGS.max_seq_length,
is_training=True,
drop_remainder=True)
estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)
if FLAGS.do_eval:
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
num_actual_eval_examples = len(eval_examples)
if FLAGS.use_tpu:
while len(eval_examples) % FLAGS.eval_batch_size != 0:
eval_examples.append(PaddingInputExample())
eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
file_based_convert_examples_to_features(eval_examples, label_list,
FLAGS.max_seq_length,
tokenizer, eval_file)
tf.logging.info("***** Running evaluation *****")
tf.logging.info(" Num examples = %d (%d actual, %d padding)",
len(eval_examples), num_actual_eval_examples,
len(eval_examples) - num_actual_eval_examples)
tf.logging.info(" Batch size = %d", FLAGS.eval_batch_size)
# This tells the estimator to run through the entire set.
eval_steps = None
if FLAGS.use_tpu:
assert len(eval_examples) % FLAGS.eval_batch_size == 0
eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)
eval_drop_remainder = True if FLAGS.use_tpu else False
eval_input_fn = file_based_input_fn_builder(
input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=eval_drop_remainder)
result = estimator.evaluate(input_fn=eval_input_fn, steps=eval_steps)
output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
with tf.gfile.GFile(output_eval_file, "w") as writer:
tf.logging.info("***** Eval results *****")
for key in sorted(result.keys()):
tf.logging.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
if FLAGS.do_predict:
predict_examples = processor.get_test_examples(FLAGS.data_dir)
num_actual_predict_examples = len(predict_examples)
if FLAGS.use_tpu:
while len(predict_examples) % FLAGS.predict_batch_size != 0:
predict_examples.append(PaddingInputExample())
predict_file = os.path.join(FLAGS.output_dir, "predict.tf_record")
file_based_convert_examples_to_features(predict_examples, label_list,
FLAGS.max_seq_length,
tokenizer, predict_file)
tf.logging.info("***** Running prediction*****")
tf.logging.info(" Num examples = %d (%d actual, %d padding)",
len(predict_examples), num_actual_predict_examples,
len(predict_examples) - num_actual_predict_examples)
tf.logging.info(" Batch size = %d", FLAGS.predict_batch_size)
predict_drop_remainder = True if FLAGS.use_tpu else False
predict_input_fn = file_based_input_fn_builder(
input_file=predict_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=predict_drop_remainder)
result = estimator.predict(input_fn=predict_input_fn)
output_predict_file = os.path.join(FLAGS.output_dir,
"test_results.tsv")
with tf.gfile.GFile(output_predict_file, "w") as writer:
num_written_lines = 0
tf.logging.info("***** Predict results *****")
for (i, prediction) in enumerate(result):
probabilities = prediction["probabilities"]
if i >= num_actual_predict_examples:
break
output_line = "\t".join(
str(class_probability)
for class_probability in probabilities) + "\n"
writer.write(output_line)
num_written_lines += 1
assert num_written_lines == num_actual_predict_examples
def main(_):
# tf.logging.set_verbosity(logger.info)
processors = {
"zhejiang": ZhejiangProcesser,
}
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
if not FLAGS.do_train and not FLAGS.do_eval and not FLAGS.do_predict:
raise ValueError(
"At least one of `do_train`, `do_eval` or `do_predict' must be True."
)
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
# fine tuning 句子的长度不能比 pre train 长
if FLAGS.max_seq_length > bert_config.max_position_embeddings:
raise ValueError(
"Cannot use sequence length %d because the BERT model was only trained up to sequence length %d"
% (FLAGS.max_seq_length, bert_config.max_position_embeddings))
# 创建输出文件夹
tf.gfile.MakeDirs(FLAGS.output_dir)
# 必须为每一task定义一个processor
task_name = FLAGS.task_name.lower()
if task_name not in processors:
raise ValueError("Task not found: %s" % (task_name))
# 选取指定的数据
logger.info("数据集合是: %s" % task_name)
processor = processors[task_name]()
# 事先定义好分类类别的label集合
label_list = processor.get_labels()
logger.info(label_list)
# 加载词典，并将词典编号处理
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
# 忽略tpu
tpu_cluster_resolver = None
if FLAGS.use_tpu and FLAGS.tpu_name:
tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
tpu_config = tf.contrib.tpu.TPUConfig(
iterations_per_loop=FLAGS.iterations_per_loop,
num_shards=FLAGS.num_tpu_cores,
per_host_input_for_training=is_per_host)
# RunConfig的设置
run_config = tf.contrib.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=FLAGS.master,
model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps,
tpu_config=tpu_config)
train_examples = None
num_train_steps = None
num_warmup_steps = None
# 获取训练的数据
if FLAGS.do_train:
train_examples = processor.get_train_examples(FLAGS.data_dir)
num_train_steps = int(
len(train_examples) / FLAGS.train_batch_size *
FLAGS.num_train_epochs)
# 学习速率的确定？？？
num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)
# 获取函数
model_fn = model_fn_builder(bert_config=bert_config,
num_labels=len(label_list),
init_checkpoint=FLAGS.init_checkpoint,
learning_rate=FLAGS.learning_rate,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
use_tpu=FLAGS.use_tpu,
use_one_hot_embeddings=FLAGS.use_tpu)
# If TPU is not available, this will fall back to normal Estimator on CPU
# or GPU.
estimator = tf.contrib.tpu.TPUEstimator(
use_tpu=FLAGS.use_tpu,
model_fn=model_fn,
config=run_config,
train_batch_size=FLAGS.train_batch_size,
eval_batch_size=FLAGS.eval_batch_size,
predict_batch_size=FLAGS.predict_batch_size)
logger.info("是否加载训练数据：")
logger.info(FLAGS.do_train)
if FLAGS.do_train:
train_file = os.path.join(FLAGS.output_dir, "train.tf_record")
file_based_convert_examples_to_features(train_examples, label_list,
FLAGS.max_seq_length,
tokenizer, train_file)
logger.info("***** Running training *****")
logger.info(" Num examples = %d", len(train_examples))
logger.info(" Batch size = %d", FLAGS.train_batch_size)
logger.info(" Num steps = %d", num_train_steps)
train_input_fn = file_based_input_fn_builder(
input_file=train_file,
seq_length=FLAGS.max_seq_length,
is_training=True,
drop_remainder=True)
estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)
logger.info("是否加载DEV数据：")
logger.info(FLAGS.do_eval)
if FLAGS.do_eval:
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
num_actual_eval_examples = len(eval_examples)
if FLAGS.use_tpu:
# TPU requires a fixed batch size for all batches, therefore the number
# of examples must be a multiple of the batch size, or else examples
# will get dropped. So we pad with fake examples which are ignored
# later on. These do NOT count towards the metric (all tf.metrics
# support a per-instance weight, and these get a weight of 0.0).
while len(eval_examples) % FLAGS.eval_batch_size != 0:
eval_examples.append(PaddingInputExample())
eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
file_based_convert_examples_to_features(eval_examples, label_list,
FLAGS.max_seq_length,
tokenizer, eval_file)
logger.info("***** Running evaluation *****")
logger.info(" Num examples = %d (%d actual, %d padding)",
len(eval_examples), num_actual_eval_examples,
len(eval_examples) - num_actual_eval_examples)
logger.info(" Batch size = %d", FLAGS.eval_batch_size)
# This tells the estimator to run through the entire set.
eval_steps = None
# However, if running eval on the TPU, you will need to specify the
# number of steps.
if FLAGS.use_tpu:
assert len(eval_examples) % FLAGS.eval_batch_size == 0
eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)
eval_drop_remainder = True if FLAGS.use_tpu else False
eval_input_fn = file_based_input_fn_builder(
input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=eval_drop_remainder)
result = estimator.evaluate(input_fn=eval_input_fn, steps=eval_steps)
output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
with tf.gfile.GFile(output_eval_file, "w") as writer:
logger.info("***** Eval results *****")
for key in sorted(result.keys()):
logger.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
logger.info("是否加载TEST数据：")
logger.info(FLAGS.do_eval)
if FLAGS.do_predict:
predict_examples = processor.get_test_examples(FLAGS.data_dir)
num_actual_predict_examples = len(predict_examples)
if FLAGS.use_tpu:
# TPU requires a fixed batch size for all batches, therefore the number
# of examples must be a multiple of the batch size, or else examples
# will get dropped. So we pad with fake examples which are ignored
# later on.
while len(predict_examples) % FLAGS.predict_batch_size != 0:
predict_examples.append(PaddingInputExample())
predict_file = os.path.join(FLAGS.output_dir, "predict.tf_record")
file_based_convert_examples_to_features(predict_examples, label_list,
FLAGS.max_seq_length,
tokenizer, predict_file)
logger.info("***** Running prediction*****")
logger.info(" Num examples = %d (%d actual, %d padding)",
len(predict_examples), num_actual_predict_examples,
len(predict_examples) - num_actual_predict_examples)
logger.info(" Batch size = %d", FLAGS.predict_batch_size)
predict_drop_remainder = True if FLAGS.use_tpu else False
predict_input_fn = file_based_input_fn_builder(
input_file=predict_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=predict_drop_remainder)
result = estimator.predict(input_fn=predict_input_fn)
output_predict_file = os.path.join(FLAGS.output_dir,
"test_results.tsv")
with tf.gfile.GFile(output_predict_file, "w") as writer:
num_written_lines = 0
logger.info("***** Predict results *****")
for (i, prediction) in enumerate(result):
probabilities = prediction["probabilities"]
if i >= num_actual_predict_examples:
break
output_line = "\t".join(
str(class_probability)
for class_probability in probabilities) + "\n"
writer.write(output_line)
num_written_lines += 1
assert num_written_lines == num_actual_predict_examples
os.path.join(bert_dir, 'bert_config.json'))
(total_loss, logits,
pred_ids) = create_model_2(bert_config=bert_config,
is_training=False,
input_ids=input_ids_p,
input_mask=input_mask_p,
segment_ids=None,
label_ids=None,
num_labels=num_labels,
use_one_hot_embeddings=False)
saver = tf.train.Saver()
saver.restore(sess, tf.train.latest_checkpoint(model_dir))
tokenizer = tokenization.FullTokenizer(vocab_file=os.path.join(
bert_dir, 'vocab.txt'),
do_lower_case=True)
app = Flask(__name__)
@app.route('/ner_predict_service', methods=['GET'])
def ner_predict_service():
"""
do online prediction. each time make prediction for one instance.
you can change to a batch if you want.
:param line: a list. element is: [dummy_label,text_a,text_b]
:return:
"""
def convert(line):
feature = convert_single_example(0, line, label_list, max_seq_length,
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
processors = {"ner": NerProcessor}
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
if not FLAGS.do_train and not FLAGS.do_eval and not FLAGS.do_predict:
raise ValueError(
"At least one of `do_train`, `do_eval` or `do_predict' must be True."
)
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
if FLAGS.max_seq_length > bert_config.max_position_embeddings:
raise ValueError(
"Cannot use sequence length %d because the BERT model "
"was only trained up to sequence length %d" %
(FLAGS.max_seq_length, bert_config.max_position_embeddings))
tf.gfile.MakeDirs(FLAGS.output_dir)
task_name = FLAGS.task_name.lower()
if task_name not in processors:
raise ValueError("Task not found: %s" % (task_name))
processor = processors[task_name]()
label_list = processor.get_labels(FLAGS.data_dir, "ner")
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
tpu_cluster_resolver = None
if FLAGS.use_tpu and FLAGS.tpu_name:
tpu_cluster_resolver = tf.contrib.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name, zone=FLAGS.tpu_zone, project=FLAGS.gcp_project)
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=FLAGS.master,
model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps,
tpu_config=tf.contrib.tpu.TPUConfig(
iterations_per_loop=FLAGS.iterations_per_loop,
num_shards=FLAGS.num_tpu_cores,
per_host_input_for_training=is_per_host))
train_examples = None
num_train_steps = None
num_warmup_steps = None
if FLAGS.do_train:
train_examples = processor.get_train_examples(FLAGS.data_dir)
num_train_steps = int(
len(train_examples) / FLAGS.train_batch_size *
FLAGS.num_train_epochs)
num_warmup_steps = int(num_train_steps * FLAGS.warmup_proportion)
model_fn = model_fn_builder(bert_config=bert_config,
num_labels=len(label_list),
init_checkpoint=FLAGS.init_checkpoint,
learning_rate=FLAGS.learning_rate,
num_train_steps=num_train_steps,
num_warmup_steps=num_warmup_steps,
use_tpu=FLAGS.use_tpu,
use_one_hot_embeddings=FLAGS.use_tpu)
# If TPU is not available, this will fall back to normal Estimator on CPU
# or GPU.
estimator = tf.contrib.tpu.TPUEstimator(
use_tpu=FLAGS.use_tpu,
model_fn=model_fn,
config=run_config,
train_batch_size=FLAGS.train_batch_size,
eval_batch_size=FLAGS.eval_batch_size,
predict_batch_size=FLAGS.predict_batch_size)
if FLAGS.do_train:
train_file = os.path.join(FLAGS.output_dir, "train.tf_record")
file_based_convert_examples_to_features(train_examples, label_list,
FLAGS.max_seq_length,
tokenizer, train_file)
tf.logging.info("***** Running training *****")
tf.logging.info(" Num examples = %d", len(train_examples))
tf.logging.info(" Batch size = %d", FLAGS.train_batch_size)
tf.logging.info(" Num steps = %d", num_train_steps)
train_input_fn = file_based_input_fn_builder(
input_file=train_file,
seq_length=FLAGS.max_seq_length,
is_training=True,
drop_remainder=True)
estimator.train(input_fn=train_input_fn, max_steps=num_train_steps)
if FLAGS.do_eval:
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
num_actual_eval_examples = len(eval_examples)
if FLAGS.use_tpu:
# TPU requires a fixed batch size for all batches, therefore the number
# of examples must be a multiple of the batch size, or else examples
# will get dropped. So we pad with fake examples which are ignored
# later on. These do NOT count towards the metric (all tf.metrics
# support a per-instance weight, and these get a weight of 0.0).
while len(eval_examples) % FLAGS.eval_batch_size != 0:
eval_examples.append(PaddingInputExample())
eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
file_based_convert_examples_to_features(eval_examples, label_list,
FLAGS.max_seq_length,
tokenizer, eval_file)
tf.logging.info("***** Running evaluation *****")
tf.logging.info(" Num examples = %d (%d actual, %d padding)",
len(eval_examples), num_actual_eval_examples,
len(eval_examples) - num_actual_eval_examples)
tf.logging.info(" Batch size = %d", FLAGS.eval_batch_size)
# This tells the estimator to run through the entire set.
eval_steps = None
# However, if running eval on the TPU, you will need to specify the
# number of steps.
if FLAGS.use_tpu:
assert len(eval_examples) % FLAGS.eval_batch_size == 0
eval_steps = int(len(eval_examples) // FLAGS.eval_batch_size)
eval_drop_remainder = True if FLAGS.use_tpu else False
eval_input_fn = file_based_input_fn_builder(
input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=eval_drop_remainder)
result = estimator.evaluate(input_fn=eval_input_fn, steps=eval_steps)
output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
with tf.gfile.GFile(output_eval_file, "w") as writer:
tf.logging.info("***** Eval results *****")
for key in sorted(result.keys()):
tf.logging.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
if FLAGS.do_predict:
predict_examples = processor.get_test_examples(FLAGS.data_dir)
num_actual_predict_examples = len(predict_examples)
label_map_inv = {}
for (i, label) in enumerate(label_list):
label_map_inv[i] = label
if FLAGS.use_tpu:
# TPU requires a fixed batch size for all batches, therefore the number
# of examples must be a multiple of the batch size, or else examples
# will get dropped. So we pad with fake examples which are ignored
# later on.
while len(predict_examples) % FLAGS.predict_batch_size != 0:
predict_examples.append(PaddingInputExample())
predict_file = os.path.join(FLAGS.output_dir, "predict.tf_record")
file_based_convert_examples_to_features(predict_examples, label_list,
FLAGS.max_seq_length,
tokenizer, predict_file)
tf.logging.info("***** Running prediction*****")
tf.logging.info(" Num examples = %d (%d actual, %d padding)",
len(predict_examples), num_actual_predict_examples,
len(predict_examples) - num_actual_predict_examples)
tf.logging.info(" Batch size = %d", FLAGS.predict_batch_size)
predict_drop_remainder = True if FLAGS.use_tpu else False
predict_input_fn = file_based_input_fn_builder(
input_file=predict_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=predict_drop_remainder)
result = estimator.predict(input_fn=predict_input_fn)
output_predict_file = os.path.join(FLAGS.output_dir,
"test_results.tsv")
predict_features = convert_examples_to_features(
predict_examples, label_list, FLAGS.max_seq_length, tokenizer)
predict_label_ids = []
true_label_ids = []
with tf.gfile.GFile(output_predict_file, "w") as writer:
num_written_lines = 0
tf.logging.info("***** Predict results *****")
for (i, prediction) in enumerate(result):
pred_ids = prediction["predict"]
if i >= num_actual_predict_examples:
break
feature = predict_features[i]
# predict_label_ids.extend([pred_id for pred_id in pred_ids])
predict_label_ids.extend(pred_ids)
true_label_ids.extend(feature.label_ids)
f_tokens = tokenizer.convert_ids_to_tokens(feature.input_ids)
text_line = "\t".join(str(token) for token in f_tokens) + "\n"
writer.write(text_line)
label_line = "\t".join(
label_map_inv[label_id]
for label_id in feature.label_ids) + "\n"
writer.write(label_line)
output_line = "\t".join(label_map_inv[pred_id]
for pred_id in pred_ids) + "\n"
writer.write(output_line)
writer.write("\n")
num_written_lines += 1
assert num_written_lines == num_actual_predict_examples
from sklearn.metrics import classification_report
tf.logging.info("***** Prediction result *****")
opt_labels = [1, 2, 3, 4, 5, 6, 7, 9, 10, 11, 12, 13, 14, 15]
opt_label_list = label_list[1:8] + label_list[9:16]
tf.logging.info(
classification_report(true_label_ids,
predict_label_ids,
labels=opt_labels,
target_names=opt_label_list))
