def test_capture(self):
global_step = tf.contrib.framework.get_or_create_global_step()
# Some test computation
some_weights = tf.get_variable("weigths", [2, 128])
computation = tf.nn.softmax(some_weights)
hook = hooks.MetadataCaptureHook(params={"step": 5},
model_dir=self.model_dir)
hook.begin()
with self.test_session() as sess:
sess.run(tf.global_variables_initializer())
#pylint: disable=W0212
mon_sess = monitored_session._HookedSession(sess, [hook])
# Should not trigger for step 0
sess.run(tf.assign(global_step, 0))
mon_sess.run(computation)
self.assertEqual(gfile.ListDirectory(self.model_dir), [])
# Should trigger *after* step 5
sess.run(tf.assign(global_step, 5))
mon_sess.run(computation)
self.assertEqual(gfile.ListDirectory(self.model_dir), [])
mon_sess.run(computation)
self.assertEqual(set(gfile.ListDirectory(self.model_dir)),
set(["run_meta", "tfprof_log", "timeline.json"]))
def validate_hourly(working_dir, validate_name=None):
""" compiles a list of games based on the new hourly directory format. Then
calls validate on it """
holdout_dirs = gfile.ListDirectory(fsdb.holdout_dir())
holdout_files = (os.path.join(fsdb.holdout_dir(), d, f)
for d in reversed(gfile.ListDirectory(fsdb.holdout_dir()))
for f in gfile.ListDirectory(os.path.join(fsdb.holdout_dir(),d))
if gfile.IsDirectory(os.path.join(fsdb.holdout_dir(),d)))
holdout_files = list(itertools.islice(holdout_files, 20000))
random.shuffle(holdout_files)
dual_net.validate(holdout_files)
def execute_job():
config = JobExecutorConfig()
make_needed_dirs(config)
configure_logging(config)
if config.prediction == 'BERT':
match_predictor = BertMatchPredictor()
elif config.prediction == 'KEYED_VECTORS':
match_predictor = KeyedVectorsFormatPredictor()
else:
raise Exception("Wrong prediction mode")
while True:
logging.info("job iteration started")
dir_in = config.dir_in
files_names = [
f for f in gfile.ListDirectory(dir_in)
if not gfile.IsDirectory(join(dir_in, f))
]
for file_name in files_names:
logging.info(file_name)
file_path = join(dir_in, file_name)
try:
match_predictor.predict(dir_in, file_name, config.dir_result)
gfile.Rename(file_path, join(config.dir_success, file_name))
except Exception:
logging.error(traceback.format_exc())
gfile.Rename(file_path, join(config.dir_error, file_name))
logging.info("job iteration finished")
time.sleep(config.interval)
def __init__(self,
img_name_to_token_ids,
img_feature_dir,
num_timesteps,
vocab,
deterministic=False):
'''
:param img_name_to_token_ids: 图像到描述字典
:param img_feature_dir: 图像特征 保存文件目录
:param num_timesteps: 时间步的数量
:param vocab: 词表
:param deterministic: 是否打乱
'''
self._vocab = vocab
self._all_img_feature_filepaths = [] # 拼接出 图像特征文件的 路径
for filename in gfile.ListDirectory(img_feature_dir):
self._all_img_feature_filepaths.append(
os.path.join(img_feature_dir, filename))
self._img_name_to_token_ids = img_name_to_token_ids
self._num_timesteps = num_timesteps
self._indicator = 0 # batch_size 的 起始点
self._deterministic = deterministic
self._img_feature_filenames = [] # 保存所有图像特征的路径
self._img_feature_data = [] # 保存 所有 图像特征
self._load_img_feature_pickle()
if not self._deterministic:
self._random_shuffle()
def produce_timeline_profile(profile_dir, resources_dir, profile_cnt, options):
"""Produces a timeline profile."""
timeline_path = os.path.join(resources_dir, PROFILER_COMMON_PREFIX + 'timeline')
if not os.path.isfile(timeline_path):
profiles = {}
log_path = os.path.join(PROFILER_TMP_DIR, PROFILER_TMP_NAME)
options['output'] = 'timeline:outfile=' + log_path
opts = model_analyzer._build_options(options) # pylint: disable=protected-access
for idx, prof in enumerate(gfile.ListDirectory(profile_dir)):
prof_file = os.path.join(profile_dir, prof)
if not os.path.isfile(prof_file):
continue
chosen_profile = os.path.join(resources_dir, PROFILER_COMMON_PREFIX + 'timeline_' + prof)
profiles[prof] = chosen_profile
if os.path.isfile(chosen_profile):
if idx == 0:
target_ts = get_timestamp(chosen_profile)
continue
tf.logging.info("Parse profile context %r" % prof_file)
remove_tmp_files()
# Parse profile context
ProfilerFromFile(prof_file.encode('utf-8'))
pwtf.Profile(options['view'].encode('utf-8'), opts.SerializeToString())
DeleteProfiler()
if idx == 0:
prof_names = get_informative_profiles(PROFILER_TMP_DIR, profile_cnt)
target_ts = get_timestamp(os.path.join(PROFILER_TMP_DIR, prof_names[0]))
else:
prof_names = get_profiles_by_timestamp(PROFILER_TMP_DIR, target_ts, profile_cnt)
tf.logging.info("Choose %r as the most informative profile context for %r" % (prof_names, prof))
gen_profile([os.path.join(PROFILER_TMP_DIR, name) for name in prof_names], chosen_profile)
merge_profiles(profiles, timeline_path)
return load_profile(timeline_path)
def load_config(config_path, config=None):
"""Loads configs from (possibly multiple) file(s).
Args:
config_path: Paths to configuration files. This can be a `list` of
config file names, or a path to a directory in which all files
are loaded, or a string of multiple file names separated by commas.
config (dict, optional): A config dict to which new configurations are
added. If `None`, a new config dict is created.
Returns:
A `dict` of configurations.
"""
fnames = []
if isinstance(config_path, (list, tuple)):
fnames = list(config_path)
elif gfile.IsDirectory(config_path):
for fname in gfile.ListDirectory(config_path):
fname = os.path.join(config_path, fname)
if not gfile.IsDirectory(fname):
fnames.append(fname)
else:
for fname in config_path.split(","):
fname = fname.strip()
if not fname:
continue
fnames.append(fname)
if config is None:
config = {}
for fname in fnames:
config = load_config_single(fname, config)
return config
def _download(self, trial):
"""Downloads a single url given by the trial (thread safe).
Args:
trial (UriTrial): Object containing info about download.
Raises:
ValueError: If the destination dir is not empty
"""
log = util.build_log(prefix=trial.id)
# Check the download dir is empty
if (gfile.Exists(trial.output_path)
and gfile.ListDirectory(trial.output_path)):
raise ValueError('Download dir {} should be empty'.format(
trial.output_path))
gfile.MakeDirs(trial.output_path)
log('Start downloading...')
self._backend.download(trial)
# TODO(epot): Compute the checksum
# Update the output path
trial.output_path = get_download_filepath(trial)
log('Download complete at {}', trial.output_path)
def validate_holdout_selfplay():
"""Validate on held-out selfplay data."""
holdout_dirs = (
os.path.join(fsdb.holdout_dir(), d)
for d in reversed(gfile.ListDirectory(fsdb.holdout_dir()))
if gfile.IsDirectory(os.path.join(fsdb.holdout_dir(), d))
for f in gfile.ListDirectory(os.path.join(fsdb.holdout_dir(), d)))
# This is a roundabout way of computing how many hourly directories we need
# to read in order to encompass 20,000 holdout games.
holdout_dirs = set(itertools.islice(holdout_dirs), 20000)
cmd = ['python3', 'validate.py'] + list(holdout_dirs) + [
'--use_tpu', '--tpu_name={}'.format(TPU_NAME),
'--flagfile=rl_loop/distributed_flags', '--expand_validation_dirs'
]
mask_flags.run(cmd)
def CopyRecursively(src, dst, overwrite=False):
entries = gfile.ListDirectory(src)
for entry in entries:
src_path = os.path.join(src, entry)
dst_path = os.path.join(dst, entry)
if gfile.IsDirectory(src_path):
gfile.MkDir(dst_path)
CopyRecursively(src_path, dst_path, overwrite)
else:
gfile.Copy(src_path, dst_path, overwrite)
def get_download_filepath(trial):
"""Extract the downloaded file from the completed trial."""
# Get the downloaded file:
files = list(gfile.ListDirectory(trial.output_path))
if not files: # Should have been catched before, but just in case
raise ValueError('Download {} failed!'.format(trial.id))
elif len(files) > 1:
raise ValueError('Multiple files detected for {}.'.format(trial.id))
# Update the output path
filename = files[0]
return os.path.join(trial.output_path, filename)
def gather(
input_directory: 'where to look for games' = 'data/selfplay/',
output_directory: 'where to put collected games' = 'data/training_chunks/',
examples_per_record:
'how many tf.examples to gather in each chunk' = EXAMPLES_PER_RECORD):
qmeas.start_time('gather')
_ensure_dir_exists(output_directory)
models = [
model_dir.strip('/')
for model_dir in sorted(gfile.ListDirectory(input_directory))[-50:]
]
with timer("Finding existing tfrecords..."):
model_gamedata = {
model:
gfile.Glob(os.path.join(input_directory, model, '*.tfrecord.zz'))
for model in models
}
print("Found %d models" % len(models))
for model_name, record_files in sorted(model_gamedata.items()):
print(" %s: %s files" % (model_name, len(record_files)))
meta_file = os.path.join(output_directory, 'meta.txt')
try:
with gfile.GFile(meta_file, 'r') as f:
already_processed = set(f.read().split())
except tf.errors.NotFoundError:
already_processed = set()
num_already_processed = len(already_processed)
for model_name, record_files in sorted(model_gamedata.items()):
if set(record_files) <= already_processed:
continue
print("Gathering files for %s:" % model_name)
for i, example_batch in enumerate(
tqdm(
preprocessing.shuffle_tf_examples(examples_per_record,
record_files))):
output_record = os.path.join(
output_directory,
'{}-{}.tfrecord.zz'.format(model_name, str(i)))
preprocessing.write_tf_examples(output_record,
example_batch,
serialize=False)
already_processed.update(record_files)
print("Processed %s new files" %
(len(already_processed) - num_already_processed))
with gfile.GFile(meta_file, 'w') as f:
f.write('\n'.join(sorted(already_processed)))
qmeas.stop_time('gather')
def _load_data(self):
dataset = np.zeros((24 * 4 * 183, 64, 64, 3))
all_files = [x for x in gfile.ListDirectory('/home/adarsh/Desktop/Disentangled-Representation-Learning-Playground/data/cars/') if ".mat" in x]
for i, filename in enumerate(all_files):
data_mesh = _load_mesh(filename)
factor1 = np.array(list(range(4)))
factor2 = np.array(list(range(24)))
all_factors = np.transpose([
np.tile(factor1, len(factor2)),
np.repeat(factor2, len(factor1)),
np.tile(i,len(factor1) * len(factor2))])
indexes = self.index.features_to_index(all_factors)
dataset[indexes] = data_mesh
return dataset
def aggregate():
logger.info("Gathering game results")
os.makedirs(PATHS.TRAINING_CHUNK_DIR, exist_ok=True)
os.makedirs(PATHS.SELFPLAY_DIR, exist_ok=True)
models = [
model_dir.strip('/')
for model_dir in sorted(gfile.ListDirectory(PATHS.SELFPLAY_DIR))[-50:]
]
with timer("Finding existing tfrecords..."):
model_gamedata = {
model: gfile.Glob(os.path.join(PATHS.SELFPLAY_DIR, model, '*.zz'))
for model in models
}
logger.info("Found %d models" % len(models))
for model_name, record_files in sorted(model_gamedata.items()):
logger.info(" %s: %s files" % (model_name, len(record_files)))
meta_file = os.path.join(PATHS.TRAINING_CHUNK_DIR, 'meta.txt')
try:
with gfile.GFile(meta_file, 'r') as f:
already_processed = set(f.read().split())
except tf.errors.NotFoundError:
already_processed = set()
num_already_processed = len(already_processed)
for model_name, record_files in sorted(model_gamedata.items()):
if set(record_files) <= already_processed:
continue
logger.info("Gathering files for %s:" % model_name)
for i, example_batch in enumerate(
tqdm(
preprocessing.shuffle_tf_examples(
GLOBAL_PARAMETER_STORE.EXAMPLES_PER_RECORD,
record_files))):
output_record = os.path.join(
PATHS.TRAINING_CHUNK_DIR,
'{}-{}.tfrecord.zz'.format(model_name, str(i)))
preprocessing.write_tf_examples(output_record,
example_batch,
serialize=False)
already_processed.update(record_files)
logger.info("Processed %s new files" %
(len(already_processed) - num_already_processed))
with gfile.GFile(meta_file, 'w') as f:
f.write('\n'.join(sorted(already_processed)))
def produce_timeline_profile():
"""Produces a timeline profile."""
# Find the largest profile, since every step is profiled for the "graph"
# view, and the largest step tends to be the most informative.
# TODO: Optimize backend to only process largest step in this scenario.
largest_profile_size = 0
for file_name in gfile.ListDirectory(PROFILER_LOG_DIR):
if 'profiler-ui.log_' in file_name:
file_path = os.path.join(PROFILER_LOG_DIR, file_name)
with gfile.GFile(file_path, 'rb') as profile:
file_size = profile.size()
if largest_profile_size < file_size:
largest_profile_size = file_size
path = os.path.join(PROFILER_LOG_DIR, file_name)
return load_profile(path)
def _load_data(self):
dataset = np.zeros((24 * 4 * 183, 64, 64, 3))
all_files = [x for x in gfile.ListDirectory(CARS3D_PATH) if ".mat" in x]
for i, filename in enumerate(all_files):
data_mesh = _load_mesh(filename)
factor1 = np.array(list(range(4)))
factor2 = np.array(list(range(24)))
all_factors = np.transpose([
np.tile(factor1, len(factor2)),
np.repeat(factor2, len(factor1)),
np.tile(i,
len(factor1) * len(factor2))
])
indexes = self.index.features_to_index(all_factors)
dataset[indexes] = data_mesh
return dataset
def format_input(input_path, size):
"""Reads input path, randomly selects a sub-sample and concatenates them.
Args:
input_path: `str`, directory to read files from.
size: `int`, number of files to read.
Returns:
List of `str` containing independent text reviews.
"""
files = [
path for path in gfile.ListDirectory(input_path)
if path.endswith(constants.FILE_EXTENSION)
]
files = np.random.choice(files, size, replace=False)
files = [os.path.join(input_path, filename) for filename in files]
return get_prediction_input(files)
def aggregate_json_results(base_path):
"""Aggregates all the result files in a directory into a namespaced dict.
Args:
base_path: String with the directory containing JSON files that only contain
dictionaries.
Returns:
Namespaced dictionary with the results.
"""
result = {}
compiled_pattern = re.compile(r"(.*)\.json")
for filename in gfile.ListDirectory(base_path):
match = compiled_pattern.match(filename)
if match:
path = os.path.join(base_path, filename)
with tf.gfile.GFile(path, "r") as f:
result[match.group(1)] = json.load(f)
return namespaced_dict(**result)
def __init__(self,
img_name_to_token_ids,
img_feature_dir,
num_timesteps,
vocab,
deterministic=False):
self._vocab = vocab
self._all_img_feature_filepaths = []
for filename in gfile.ListDirectory(img_feature_dir):
self._all_img_feature_filepaths.append(
os.path.join(img_feature_dir, filename))
# pprint.pprint(self._all_img_feature_filepaths)
self._img_name_to_token_ids = img_name_to_token_ids
self._num_timesteps = num_timesteps
self._indicator = 0
self._deterministic = deterministic
self._img_feature_filenames = []
self._img_feature_data = []
self._load_img_feature_pickle()
if not self._deterministic:
self._random_shuffle()
def __init__(self, img_feature_dir, valid_percent=0.1, test_percent=0.1, is_shuffle=True):
self._all_img_feature_filepaths = []
for filename in gfile.ListDirectory(img_feature_dir):
self._all_img_feature_filepaths.append(os.path.join(img_feature_dir, filename))
self.is_shuffle = is_shuffle
self._img_feature_data = []
self._img_feature_labels = []
self._load_img_feature_pickle()
if self.is_shuffle:
self._random_shuffle()
# 根据数据集加载出训练，验证，测试集
valid_len = int(self.size() * valid_percent)
test_len = int(self.size() * test_percent)
self._valid_data = self._img_feature_data[:valid_len, :]
self._valid_labels = self._img_feature_labels[:valid_len]
self._test_data = self._img_feature_data[valid_len:valid_len + test_len, :]
self._test_labels = self._img_feature_labels[valid_len:valid_len + test_len]
self._train_data = self._img_feature_data[valid_len + test_len:, :]
self._train_labels = self._img_feature_labels[valid_len + test_len:]
self._train_indicator = 0
print(self._valid_data.shape, self._valid_labels.shape)
print(self._test_data.shape, self._test_labels.shape)
print(self._train_data.shape, self._train_labels.shape)
def get_sgf_names(model):
game_dir = HOLDOUT_PATH.format(FLAGS.base_dir, model)
tf_records = map(os.path.basename, gfile.ListDirectory(game_dir))
sgfs = [record.replace('.tfrecord.zz', '.sgf') for record in tf_records]
return [PATH_TEMPLATE.format(FLAGS.base_dir, model, sgf) for sgf in sgfs]
def remove_tmp_files():
"""Removes temporary files created by the profiler."""
for file_name in gfile.ListDirectory(PROFILER_LOG_DIR):
if 'profiler-ui.' in file_name:
gfile.Remove(os.path.join(PROFILER_LOG_DIR, file_name))
def get_hour_dirs(root=None):
"""Gets the directories under selfplay_dir that match YYYY-MM-DD-HH."""
root = root or selfplay_dir()
return list(
filter(lambda s: re.match(r"\d{4}-\d{2}-\d{2}-\d{2}", s),
gfile.ListDirectory(root)))