def test_checkpoint_max_to_keep(self):
"""Test that only the most recent checkpoints are kept."""
with mock.patch.object(model_builder, 'build',
autospec=True) as mock_builder:
mock_builder.return_value = SimpleModel()
hparams = model_hparams.create_hparams(
hparams_overrides='load_pretrained=false')
pipeline_config_path = get_pipeline_config_path(
MODEL_NAME_FOR_TEST)
config_kwarg_overrides = _get_config_kwarg_overrides()
model_dir = tempfile.mkdtemp(dir=self.get_temp_dir())
model_lib_v2.train_loop(hparams,
pipeline_config_path,
model_dir=model_dir,
train_steps=20,
checkpoint_every_n=2,
checkpoint_max_to_keep=3,
**config_kwarg_overrides)
ckpt_files = tf.io.gfile.glob(
os.path.join(model_dir, 'ckpt-*.index'))
self.assertEqual(len(ckpt_files), 3,
'{} not of length 3.'.format(ckpt_files))
def main(unused_argv):
flags.mark_flag_as_required('model_dir')
flags.mark_flag_as_required('pipeline_config_path')
tf.config.set_soft_device_placement(True)
if FLAGS.checkpoint_dir:
model_lib_v2.eval_continuously(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
sample_1_of_n_eval_examples=FLAGS.sample_1_of_n_eval_examples,
sample_1_of_n_eval_on_train_examples=(
FLAGS.sample_1_of_n_eval_on_train_examples),
checkpoint_dir=FLAGS.checkpoint_dir,
wait_interval=300, timeout=FLAGS.eval_timeout)
else:
if FLAGS.use_tpu:
resolver = tf.distribute.cluster_resolver.TPUClusterResolver()
tf.config.experimental_connect_to_cluster(resolver)
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)
elif FLAGS.num_workers > 1:
strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy()
else:
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
use_tpu=FLAGS.use_tpu)
def test_train_loop_then_eval_loop(self):
"""Tests that Estimator and input function are constructed correctly."""
hparams = model_hparams.create_hparams(
hparams_overrides='load_pretrained=false')
pipeline_config_path = get_pipeline_config_path(MODEL_NAME_FOR_TEST)
config_kwarg_overrides = _get_config_kwarg_overrides()
model_dir = tf.test.get_temp_dir()
train_steps = 2
model_lib_v2.train_loop(
hparams,
pipeline_config_path,
model_dir=model_dir,
train_steps=train_steps,
checkpoint_every_n=1,
**config_kwarg_overrides)
model_lib_v2.eval_continuously(
hparams,
pipeline_config_path,
model_dir=model_dir,
checkpoint_dir=model_dir,
train_steps=train_steps,
wait_interval=10,
**config_kwarg_overrides)
def test_checkpoint_max_to_keep(self):
"""Test that only the most recent checkpoints are kept."""
strategy = tf2.distribute.OneDeviceStrategy(device='/cpu:0')
with mock.patch.object(model_builder, 'build',
autospec=True) as mock_builder:
with strategy.scope():
mock_builder.return_value = SimpleModel()
model_dir = tempfile.mkdtemp(dir=self.get_temp_dir())
pipeline_config_path = get_pipeline_config_path(
MODEL_NAME_FOR_TEST)
new_pipeline_config_path = os.path.join(model_dir,
'new_pipeline.config')
config_util.clear_fine_tune_checkpoint(pipeline_config_path,
new_pipeline_config_path)
config_kwarg_overrides = _get_config_kwarg_overrides()
with strategy.scope():
model_lib_v2.train_loop(new_pipeline_config_path,
model_dir=model_dir,
train_steps=20,
checkpoint_every_n=2,
checkpoint_max_to_keep=3,
**config_kwarg_overrides)
ckpt_files = tf.io.gfile.glob(
os.path.join(model_dir, 'ckpt-*.index'))
self.assertEqual(len(ckpt_files), 3,
'{} not of length 3.'.format(ckpt_files))
def test_train_loop_then_eval_loop(self):
"""Tests that Estimator and input function are constructed correctly."""
model_dir = tf.test.get_temp_dir()
pipeline_config_path = get_pipeline_config_path(MODEL_NAME_FOR_TEST)
new_pipeline_config_path = os.path.join(model_dir,
'new_pipeline.config')
config_util.clear_fine_tune_checkpoint(pipeline_config_path,
new_pipeline_config_path)
config_kwarg_overrides = _get_config_kwarg_overrides()
train_steps = 2
strategy = tf2.distribute.MirroredStrategy(['/cpu:0', '/cpu:1'])
with strategy.scope():
model_lib_v2.train_loop(new_pipeline_config_path,
model_dir=model_dir,
train_steps=train_steps,
checkpoint_every_n=1,
**config_kwarg_overrides)
model_lib_v2.eval_continuously(new_pipeline_config_path,
model_dir=model_dir,
checkpoint_dir=model_dir,
train_steps=train_steps,
wait_interval=1,
timeout=10,
**config_kwarg_overrides)
def test_export_metrics_json_serializable(self):
"""Tests that Estimator and input function are constructed correctly."""
strategy = tf2.distribute.OneDeviceStrategy(device='/cpu:0')
def export(data, _):
json.dumps(data)
with mock.patch.dict(exporter_lib_v2.INPUT_BUILDER_UTIL_MAP,
FAKE_BUILDER_MAP):
with strategy.scope():
model_dir = tf.test.get_temp_dir()
new_pipeline_config_path = os.path.join(
model_dir, 'new_pipeline.config')
pipeline_config_path = get_pipeline_config_path(
MODEL_NAME_FOR_TEST)
config_util.clear_fine_tune_checkpoint(
pipeline_config_path, new_pipeline_config_path)
train_steps = 2
with strategy.scope():
model_lib_v2.train_loop(
new_pipeline_config_path,
model_dir=model_dir,
train_steps=train_steps,
checkpoint_every_n=100,
performance_summary_exporter=export,
**_get_config_kwarg_overrides())
def test_checkpoint_max_to_keep(self):
"""Test that only the most recent checkpoints are kept."""
strategy = tf2.distribute.OneDeviceStrategy(device='/cpu:0')
with mock.patch.dict(exporter_lib_v2.INPUT_BUILDER_UTIL_MAP,
FAKE_BUILDER_MAP):
model_dir = tempfile.mkdtemp(dir=self.get_temp_dir())
pipeline_config_path = get_pipeline_config_path(
MODEL_NAME_FOR_TEST)
new_pipeline_config_path = os.path.join(model_dir,
'new_pipeline.config')
config_util.clear_fine_tune_checkpoint(pipeline_config_path,
new_pipeline_config_path)
config_kwarg_overrides = _get_config_kwarg_overrides()
with strategy.scope():
model_lib_v2.train_loop(new_pipeline_config_path,
model_dir=model_dir,
train_steps=20,
checkpoint_every_n=2,
checkpoint_max_to_keep=3,
**config_kwarg_overrides)
ckpt_files = tf.io.gfile.glob(
os.path.join(model_dir, 'ckpt-*.index'))
self.assertEqual(len(ckpt_files), 3,
'{} not of length 3.'.format(ckpt_files))
def main(unused_argv):
# ste the gpu (device:GPU:0)
print("Num GPUs Available: ",
len(tf.config.experimental.list_physical_devices('GPU')))
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
# Restrict TensorFlow to only use the first GPU
try:
tf.config.experimental.set_visible_devices(gpus[0], 'GPU')
tf.config.experimental.set_memory_growth(gpus[0], True)
logical_gpus = tf.config.experimental.list_logical_devices('GPU')
print(len(gpus), "Physical GPUs,", len(logical_gpus),
"Logical GPU")
except RuntimeError as e:
# Visible devices must be set before GPUs have been initialized
print(e)
flags.mark_flag_as_required('model_dir')
flags.mark_flag_as_required('pipeline_config_path')
tf.config.set_soft_device_placement(True)
if FLAGS.checkpoint_dir:
model_lib_v2.eval_continuously(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
sample_1_of_n_eval_examples=FLAGS.sample_1_of_n_eval_examples,
sample_1_of_n_eval_on_train_examples=(
FLAGS.sample_1_of_n_eval_on_train_examples),
checkpoint_dir=FLAGS.checkpoint_dir,
wait_interval=300,
timeout=FLAGS.eval_timeout)
else:
if FLAGS.use_tpu:
# TPU is automatically inferred if tpu_name is None and
# we are running under cloud ai-platform.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name)
tf.config.experimental_connect_to_cluster(resolver)
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)
# elif FLAGS.num_workers > 1:
# strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy()
# else:
# strategy = tf.compat.v2.distribute.MirroredStrategy()
# with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
use_tpu=FLAGS.use_tpu,
checkpoint_every_n=FLAGS.checkpoint_every_n,
record_summaries=FLAGS.record_summaries)
def main(unused_argv):
flags.mark_flag_as_required('model_dir')
flags.mark_flag_as_required('pipeline_config_path')
tf.config.set_soft_device_placement(True)
if FLAGS.checkpoint_dir:
if FLAGS.eval_all_checkpoints:
model_lib_v2.eval_all_checkpoints(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
sample_1_of_n_eval_examples=FLAGS.sample_1_of_n_eval_examples,
sample_1_of_n_eval_on_train_examples=(
FLAGS.sample_1_of_n_eval_on_train_examples),
checkpoint_dir=FLAGS.checkpoint_dir)
else:
model_lib_v2.eval_continuously(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
sample_1_of_n_eval_examples=FLAGS.sample_1_of_n_eval_examples,
sample_1_of_n_eval_on_train_examples=(
FLAGS.sample_1_of_n_eval_on_train_examples),
checkpoint_dir=FLAGS.checkpoint_dir,
wait_interval=300,
timeout=FLAGS.eval_timeout)
else:
if FLAGS.use_tpu:
# TPU is automatically inferred if tpu_name is None and
# we are running under cloud ai-platform.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name)
tf.config.experimental_connect_to_cluster(resolver)
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)
elif FLAGS.num_workers > 1:
strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy()
else:
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
use_tpu=FLAGS.use_tpu,
checkpoint_every_n=FLAGS.checkpoint_every_n,
record_summaries=FLAGS.record_summaries)
def train(self, steps_per_epoch, checkpoint_every_n_epochs=10):
"""
Trains the model.
Args:
steps_per_epoch: Number of steps that are to be trained for one epoch
checkpoint_every_n_epochs: Epoch interval in which to save a checkpoint while training
"""
checkpoints_every_n_steps = steps_per_epoch * checkpoint_every_n_epochs
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=self.config_path,
model_dir=self.checkpoint_path,
checkpoint_every_n=checkpoints_every_n_steps,
checkpoint_max_to_keep=150,
record_summaries=True)
def main(_):
with open('system_dict.json') as json_file:
args = json.load(json_file)
tf.config.set_soft_device_placement(True)
if args["checkpoint_dir"]:
model_lib_v2.eval_continuously(
pipeline_config_path=args["pipeline_config_path"],
model_dir=args["model_dir"],
train_steps=args["num_train_steps"],
sample_1_of_n_eval_examples=args["sample_1_of_n_eval_examples"],
sample_1_of_n_eval_on_train_examples=(
args["sample_1_of_n_eval_on_train_examples"]),
checkpoint_dir=args["checkpoint_dir"],
wait_interval=300,
timeout=args["eval_timeout"])
else:
if args["use_tpu"]:
# TPU is automatically inferred if tpu_name is None and
# we are running under cloud ai-platform.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(
args["tpu_name"])
tf.config.experimental_connect_to_cluster(resolver)
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)
elif args["num_workers"] > 1:
strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy()
else:
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=args["pipeline_config_path"],
model_dir=args["model_dir"],
train_steps=args["num_train_steps"],
use_tpu=args["use_tpu"],
checkpoint_every_n=args["checkpoint_every_n"],
record_summaries=args["record_summaries"])
def run(self):
self._validate_pipeline_config()
if self._memory_growth:
gpus = tf.config.experimental.list_physical_devices("GPU")
if gpus:
try:
# Currently, memory growth needs to be the same across GPUs
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
except RuntimeError as e:
# Memory growth must be set before GPUs have been initialized
print(e)
print("Running train loop...")
strategy = tf.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=self._pipeline_config_path,
model_dir=self._training_loop_path,
checkpoint_max_to_keep=None # keep all checkpoints
)
testboundingbox = decoded_tensors['groundtruth_boxes'].numpy()
# show_oneimage_category(testimage, testlabel, testboundingbox, IMAGE_SIZE)
# cv2.imwrite('result.jpg', resultimage)
cwd = os.getcwd()
# Print the current working directory
print("Current working directory: {0}".format(cwd))
# Start the training, ref: https://github.com/tensorflow/models/blob/master/research/object_detection/model_main_tf2.py
pipeline_config_path = '/Developer/MyRepo/WaymoObjectDetection/2DObject/tfobjectdetection/tf_ssdresnet50_1024_pipeline_P100.config'
model_dir = '/Developer/MyRepo/mymodels/tf_ssdresnet50_output'
num_train_steps = 150000
steps_per_sec_list = []
checkpoint_every_n=1000
tf.config.set_soft_device_placement(True)
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
#in: https://github.com/tensorflow/models/blob/master/research/object_detection/model_lib_v2.py
model_lib_v2.train_loop(
pipeline_config_path=pipeline_config_path,
model_dir=model_dir,
train_steps=num_train_steps,
use_tpu=False,
checkpoint_every_n=1000,
record_summaries=True)
def main(unused_argv):
if FLAGS.checkpoint_dir:
print("\n-------Running evaluation")
else:
print("\n-------Running traingin!")
flags.mark_flag_as_required('model_dir')
flags.mark_flag_as_required('pipeline_config_path')
tf.config.set_soft_device_placement(True)
print(
"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++"
)
gpus = tf.config.experimental.list_physical_devices('GPU')
if gpus:
try:
# Currently, memory growth needs to be the same across GPUs
for gpu in gpus:
tf.config.experimental.set_memory_growth(gpu, True)
logical_gpus = tf.config.experimental.list_logical_devices('GPU')
print(len(gpus), "Physical GPUs,", len(logical_gpus),
"Logical GPUs")
except RuntimeError as e:
# Memory growth must be set before GPUs have been initialized
print(e)
print(
"++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++"
)
if FLAGS.checkpoint_dir:
model_lib_v2.eval_continuously(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
train_steps=FLAGS.num_train_steps,
sample_1_of_n_eval_examples=FLAGS.sample_1_of_n_eval_examples,
sample_1_of_n_eval_on_train_examples=(
FLAGS.sample_1_of_n_eval_on_train_examples),
checkpoint_dir=FLAGS.checkpoint_dir,
wait_interval=300,
timeout=FLAGS.eval_timeout)
else:
if FLAGS.use_tpu:
# TPU is automatically inferred if tpu_name is None and
# we are running under cloud ai-platform.
resolver = tf.distribute.cluster_resolver.TPUClusterResolver(
FLAGS.tpu_name)
tf.config.experimental_connect_to_cluster(resolver)
tf.tpu.experimental.initialize_tpu_system(resolver)
strategy = tf.distribute.experimental.TPUStrategy(resolver)
elif FLAGS.num_workers > 1:
strategy = tf.distribute.experimental.MultiWorkerMirroredStrategy()
else:
strategy = tf.compat.v2.distribute.MirroredStrategy()
with strategy.scope():
model_lib_v2.train_loop(
pipeline_config_path=FLAGS.pipeline_config_path,
model_dir=FLAGS.model_dir,
save_final_config=True,
train_steps=FLAGS.num_train_steps,
use_tpu=FLAGS.use_tpu,
checkpoint_every_n=FLAGS.checkpoint_every_n,
record_summaries=FLAGS.record_summaries)
