#(total_loss, logits, trans, pred_ids) = \
(total_loss, per_example_loss, logits,
probabilities) = create_model(bert_config=bert_config,
is_training=False,
input_ids=input_ids_p,
input_mask=input_mask_p,
segment_ids=None,
labels=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.route('/class_predict_service', methods=['GET', 'POST'])
def class_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,
tokenizer)
input_ids = np.reshape([feature.input_ids],
encoding='utf-8')
text_f = open(os.path.join(os.path.join(DATA_OUTPUT_DIR, file_set_type),
"text.txt"),
"w",
encoding='utf-8')
token_in_f = open(os.path.join(
os.path.join(DATA_OUTPUT_DIR, file_set_type), "token_in.txt"),
"w",
encoding='utf-8')
token_in_not_UNK_f = open(os.path.join(
os.path.join(DATA_OUTPUT_DIR, file_set_type), "token_in_not_UNK.txt"),
"w",
encoding='utf-8')
# Processing
bert_tokenizer = tokenization.FullTokenizer(
vocab_file=vocab_file, do_lower_case=True) # 初始化 bert_token 工具
# feature
text = '\n'.join(df_data.item)
text_tokened = df_data.item.apply(bert_tokenizer.tokenize)
text_tokened = '\n'.join([' '.join(row) for row in text_tokened])
text_tokened_not_UNK = df_data.item.apply(bert_tokenizer.tokenize_not_UNK)
text_tokened_not_UNK = '\n'.join(
[' '.join(row) for row in text_tokened_not_UNK])
# label only choose first 3 lables: 高中 学科 一级知识点
# if you want all labels
# just remove list slice
predicate_list = df_data.labels.apply(lambda x: x.split())
predicate_list_str = '\n'.join([' '.join(row) for row in predicate_list])
print(f'datasize: {len(df_data)}')
text_f.write(text)
def predict_bert(m_type, data_path, train_data_file, test_data_file):
parameter_setting(m_type=m_type,
data_path=data_path,
train_file=train_data_file,
test_file=test_data_file)
checkpoint_file = tf.train.latest_checkpoint(FLAGS.model_output_dir)
# FLAGS.model_output_dir = None
precessor = DomainClf_Processor()
label_list = precessor.get_labels()
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=None,
master=FLAGS.master,
model_dir=FLAGS.model_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),
session_config=config)
model_fn = model_fn_builder(bert_config=bert_config,
num_labels=len(label_list),
init_checkpoint=checkpoint_file,
learning_rate=FLAGS.learning_rate,
num_train_steps=10000,
num_warmup_steps=1000,
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)
predict_examples = precessor.get_test_examples(FLAGS.data_dir)
num_actual_predict_examples = len(predict_examples)
predict_file = os.path.join(
FLAGS.model_output_dir,
"predict.tf_record") # overwrite the latest file
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.model_output_dir,
FLAGS.test_file.split('.')[0] + ".tsv")
output_label_file = os.path.join(
FLAGS.model_output_dir,
FLAGS.test_file.split('.')[0] + "_labels.tsv")
f = open(output_label_file, "w")
s = set()
result_list = []
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"]
assert len(probabilities) == len(label_list)
if i >= num_actual_predict_examples:
break
output_line = "\t".join(
str(class_probability)
for class_probability in probabilities) + "\n"
result_list.append(output_line.strip())
writer.write(output_line)
num_written_lines += 1
# labels
lbl_id = np.argmax(np.asarray(probabilities))
f.write(label_list[lbl_id] + "\n")
s.update([label_list[lbl_id]])
write_list_to_file(
os.path.join(base_dir, 'result/stacking/bert',
str(FLAGS.train_file.split('.')[0]),
FLAGS.test_file.split('.')[0] + '.tsv'), result_list)
def main(_):
tf.logging.set_verbosity(tf.logging.INFO)
processors = {
"baidu_95": Baidu_95_Multi_Label_Classification_Processor,
}
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()
label_length = len(label_list)
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,
label_length=label_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_test_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,
label_length=label_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:
# 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,
label_length=label_length,
is_training=False,
drop_remainder=predict_drop_remainder)
result = estimator.predict(input_fn=predict_input_fn)
output_score_value_file = os.path.join(FLAGS.output_dir, "predicate_score_value.txt")
output_predicate_predict_file = os.path.join(FLAGS.output_dir, "predicate_predict.txt")
with tf.gfile.GFile(output_score_value_file, "w") as score_value_writer:
with tf.gfile.GFile(output_predicate_predict_file, "w") as predicate_predict_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_score_value = " ".join(
str(class_probability)
for class_probability in probabilities) + "\n"
predicate_predict = []
for idx, class_probability in enumerate(probabilities):
if class_probability > 0.5:
predicate_predict.append(label_list[idx])
output_line_predicate_predict = " ".join(predicate_predict) + "\n"
predicate_predict_writer.write(output_line_predicate_predict)
score_value_writer.write(output_line_score_value)
num_written_lines += 1
assert num_written_lines == num_actual_predict_examples
def train_eval_bert(m_type, data_path, train_data_file, dev_data_file, seq_len,
epoch):
""" parameter setting """
parameter_setting(m_type=m_type,
data_path=data_path,
train_file=train_data_file,
dev_file=dev_data_file,
max_seq_length=seq_len,
EPOCH=epoch)
""" logging setting """
tf.logging.set_verbosity(tf.logging.INFO) # DEBUG，INFO，WARN，ERROR, FATAL
""" processor """
processor = DomainClf_Processor()
""" tokenizer """
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
""" bert config """
bert_config = modeling.BertConfig.from_json_file(FLAGS.bert_config_file)
""" build output folder """
tf.gfile.MakeDirs(FLAGS.model_output_dir)
""" get label """
label_list = processor.get_labels()
""" tokenizer """
tokenizer = tokenization.FullTokenizer(vocab_file=FLAGS.vocab_file,
do_lower_case=FLAGS.do_lower_case)
""" train examples"""
train_examples = processor.get_train_examples(FLAGS.data_dir)
num_batch_single_epoch = int(len(train_examples) / FLAGS.train_batch_size)
num_warmup_steps = int(num_batch_single_epoch * FLAGS.warmup_proportion)
print("num_warmup_steps", num_warmup_steps)
"""
Train Setting
"""
train_file = os.path.join(FLAGS.model_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_batch_single_epoch)
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_batch_single_epoch)
"""
run setting
"""
is_per_host = tf.contrib.tpu.InputPipelineConfig.PER_HOST_V2
run_config = tf.contrib.tpu.RunConfig(
cluster=None,
master=FLAGS.master,
model_dir=FLAGS.model_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),
session_config=config)
""" model fn for estimator"""
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_batch_single_epoch,
num_warmup_steps=num_warmup_steps,
use_tpu=FLAGS.use_tpu,
use_one_hot_embeddings=FLAGS.use_tpu)
"""estimator setting """
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)
"""
Eval Setting
"""
""" eval examples """
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
num_actual_eval_examples = len(eval_examples)
eval_file = os.path.join(FLAGS.model_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)
eval_input_fn = file_based_input_fn_builder(
input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=False)
"""
early stopping hook
"""
early_stopping = tf.contrib.estimator.stop_if_no_decrease_hook(
estimator,
metric_name='eval_accuracy',
max_steps_without_decrease=FLAGS.iterations_per_loop * 3,
min_steps=num_batch_single_epoch *
1) # min_step=num_batch_single_epoch*5) # min_step
"""
train and eval
"""
train_spec = tf.estimator.TrainSpec(train_input_fn, hooks=[early_stopping])
eval_spec = tf.estimator.EvalSpec(
eval_input_fn, exporters=None,
throttle_secs=1) # throttle_secs 多少秒之内不进行 eval
result_eval = tf.estimator.train_and_evaluate(estimator, train_spec,
eval_spec)
print('\n\n result eval', result_eval)
