def main(_):
tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.INFO)
processors = {
"cola": ColaProcessor,
"mnli": MnliProcessor,
"mrpc": MrpcProcessor,
"xnli": XnliProcessor,
}
tokenization.validate_case_matches_checkpoint(FLAGS.do_lower_case,
FLAGS.init_checkpoint)
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.compat.v1.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.compat.v1.estimator.tpu.InputPipelineConfig.PER_HOST_V2
session_config = tf.compat.v1.ConfigProto(
inter_op_parallelism_threads=FLAGS.num_inter_threads,
intra_op_parallelism_threads=FLAGS.num_intra_threads)
run_config = tf.compat.v1.estimator.tpu.RunConfig(
cluster=tpu_cluster_resolver,
master=FLAGS.master,
model_dir=FLAGS.output_dir,
save_checkpoints_steps=FLAGS.save_checkpoints_steps,
tpu_config=tf.compat.v1.estimator.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=session_config)
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.compat.v1.estimator.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.compat.v1.logging.info("***** Running training *****")
tf.compat.v1.logging.info(" Num examples = %d", len(train_examples))
tf.compat.v1.logging.info(" Batch size = %d", FLAGS.train_batch_size)
tf.compat.v1.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.compat.v1.logging.info("***** Running evaluation *****")
tf.compat.v1.logging.info(
" Num examples = %d (%d actual, %d padding)", len(eval_examples),
num_actual_eval_examples,
len(eval_examples) - num_actual_eval_examples)
tf.compat.v1.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) + 1
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)
start = time.time()
result = estimator.evaluate(input_fn=eval_input_fn,
steps=eval_steps,
hooks=[LoggerHook()])
end = time.time() - start
result['global_step'] = str(eval_steps)
result['latency_total'] = str(end)
result['latency_per_step'] = str(end / eval_steps)
if FLAGS.eval_batch_size != 1:
result['samples_per_sec'] = str(FLAGS.eval_batch_size /
(end / eval_steps))
output_eval_file = os.path.join(FLAGS.output_dir, "eval_results.txt")
with tf.compat.v1.gfile.GFile(output_eval_file, "w") as writer:
tf.compat.v1.logging.info("***** Eval results *****")
for key in sorted(result.keys()):
tf.compat.v1.logging.info(" %s = %s", key, str(result[key]))
writer.write("%s = %s\n" % (key, str(result[key])))
# BELOW IS LPOT TUNING AND BENCHMARK CODE
class Dataset(object):
def __init__(self, file_name, batch_size):
self.file_name = file_name
self.batch_size = batch_size
def __getitem__(self, idx):
return (self.file_name, self.batch_size), 0
def __len__(self):
return 1
def collate_fn(batch):
"""Puts each data field into a pd frame with outer dimension batch size"""
elem = batch[0]
return elem
from lpot.metric import METRICS
class Accuracy(object):
def __init__(self):
self.metric = METRICS('tensorflow')['Accuracy']()
# it's ugly that the label is in the iterator
def update(self, preds, label):
logits, labels = preds
self.metric.update(logits, labels)
def reset(self):
self.metric.reset()
def result(self):
return self.metric.result()
if FLAGS.tune:
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
convert_examples_to_features(examples=eval_examples,
label_list=label_list,
max_seq_length=FLAGS.max_seq_length,
tokenizer=tokenizer,
output_file=eval_file)
estimator_input_fn = input_fn_builder(input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=False)
from lpot.experimental import Quantization, common
quantizer = Quantization(FLAGS.config)
dataset = Dataset(eval_file, FLAGS.eval_batch_size)
quantizer.model = common.Model(estimator, input_fn=estimator_input_fn)
quantizer.calib_dataloader = common.DataLoader(dataset,
collate_fn=collate_fn)
quantizer.eval_dataloader = common.DataLoader(dataset,
collate_fn=collate_fn)
quantizer.metric = common.Metric(metric_cls=Accuracy)
q_model = quantizer()
q_model.save(FLAGS.output_model)
if FLAGS.benchmark:
eval_examples = processor.get_dev_examples(FLAGS.data_dir)
eval_file = os.path.join(FLAGS.output_dir, "eval.tf_record")
from lpot.experimental import Benchmark, common
evaluator = Benchmark(FLAGS.config)
dataset = Dataset(eval_file, FLAGS.eval_batch_size)
evaluator.b_dataloader = common.DataLoader(\
dataset, batch_size=FLAGS.eval_batch_size, collate_fn=collate_fn)
evaluator.metric = common.Metric(metric_cls=Accuracy)
from lpot.model.model import get_model_type
model_type = get_model_type(FLAGS.input_model)
if model_type == 'frozen_pb':
evaluator.model = FLAGS.input_model
else:
estimator_input_fn = input_fn_builder(
input_file=eval_file,
seq_length=FLAGS.max_seq_length,
is_training=False,
drop_remainder=False)
evaluator.model = common.Model(estimator,
input_fn=estimator_input_fn)
evaluator(FLAGS.mode)
pass
def reset(self):
self.pred_list = []
self.label_list = []
self.samples = 0
pass
def result(self):
correct_num = np.sum(
np.array(self.pred_list) == np.array(self.label_list))
return correct_num / self.samples
# Quantize with customized dataloader and metric
quantizer = Quantization('./conf.yaml')
dataset = Dataset()
quantizer.metric = common.Metric(MyMetric, 'hello_metric')
quantizer.calib_dataloader = common.DataLoader(dataset, batch_size=1)
quantizer.eval_dataloader = common.DataLoader(dataset, batch_size=1)
quantizer.model = common.Model('../models/saved_model')
q_model = quantizer()
# Optional, run quantized model
import tensorflow as tf
with tf.compat.v1.Graph().as_default(), tf.compat.v1.Session() as sess:
tf.compat.v1.import_graph_def(q_model.graph_def, name='')
styled_image = sess.run(['output:0'],
feed_dict={'input:0': dataset.test_images})
print("Inference is done.")
