def test_is_summary_directory(self):
"""Test is_summary_directory method success."""
summary_base_dir = tempfile.mkdtemp(dir=self.base_dir)
file_count = 1
directory_count = 1
gen_directories_and_files(summary_base_dir, file_count, directory_count)
summary_watcher = SummaryWatcher()
flag = summary_watcher.is_summary_directory(summary_base_dir, './')
assert flag
flag = summary_watcher.is_summary_directory(summary_base_dir, './\x00')
assert not flag
shutil.rmtree(summary_base_dir)
class ExplainManager:
"""ExplainManager."""
def __init__(self, summary_base_dir: str):
self._summary_base_dir = summary_base_dir
self._loader_pool = OrderedDict()
self._loading_status = _ExplainManagerStatus.INIT.value
self._status_mutex = threading.Lock()
self._loader_pool_mutex = threading.Lock()
self._max_loaders_num = _MAX_LOADERS_NUM
self._summary_watcher = SummaryWatcher()
@property
def summary_base_dir(self):
"""Return the base directory for summary records."""
return self._summary_base_dir
def start_load_data(self, reload_interval: int = 0):
"""
Start individual thread to cache explain_jobs and loading summary data periodically.
Args:
reload_interval (int): Specify the loading period in seconds. If interval == 0, data will only be loaded
once. Default: 0.
"""
thread = threading.Thread(target=self._repeat_loading,
name='start_load_thread',
args=(reload_interval, ),
daemon=True)
time.sleep(1)
thread.start()
def get_job(self, loader_id: str) -> Optional[ExplainLoader]:
"""
Return ExplainLoader given loader_id.
If explain job w.r.t given loader_id is not found, None will be returned.
Args:
loader_id (str): The id of expected ExplainLoader
Return:
explain_job
"""
self._check_status_valid()
with self._loader_pool_mutex:
if loader_id in self._loader_pool:
self._loader_pool[loader_id].query_time = datetime.now(
).timestamp()
self._loader_pool.move_to_end(loader_id, last=False)
return self._loader_pool[loader_id]
try:
loader = self._generate_loader_from_relative_path(loader_id)
loader.query_time = datetime.now().timestamp()
self._add_loader(loader)
self._reload_data_again()
except ParamValueError:
logger.warning(
'Cannot find summary in path: %s. No explain_job will be returned.',
loader_id)
return None
return loader
def get_job_list(self, offset=0, limit=None):
"""
Return List of explain jobs. includes job ID, create and update time.
Args:
offset (int): An offset for page. Ex, offset is 0, mean current page is 1. Default value is 0.
limit (int): The max data items for per page. Default value is 10.
Returns:
tuple[total, directories], total indicates the overall number of explain directories and directories
indicate list of summary directory info including the following attributes.
- relative_path (str): Relative path of summary directory, referring to settings.SUMMARY_BASE_DIR,
starting with "./".
- create_time (datetime): Creation time of summary file.
- update_time (datetime): Modification time of summary file.
"""
total, dir_infos = \
self._summary_watcher.list_explain_directories(self._summary_base_dir, offset=offset, limit=limit)
return total, dir_infos
def _repeat_loading(self, repeat_interval):
"""Periodically loading summary."""
while True:
try:
logger.info('Start to load data, repeat interval: %r.',
repeat_interval)
self._load_data()
if not repeat_interval:
return
time.sleep(repeat_interval)
except UnknownError as ex:
logger.error(
'Unexpected error happens when loading data. Loading status: %s, loading pool size: %d'
'Detail: %s', self._loading_status, len(self._loader_pool),
str(ex))
def _load_data(self):
"""
Prepare loaders in cache and start loading the data from summaries.
Only a limited number of loaders will be cached in terms of updated_time or query_time. The size of cache
pool is determined by _MAX_LOADERS_NUM. When the manager start loading data, only the lastest _MAX_LOADER_NUM
summaries will be loaded in cache. If a cached loader if queries by 'get_job', the query_time of the loader
will be updated as well as the the loader moved to the end of cache. If an uncached summary is queried,
a new loader instance will be generated and put to the end cache.
"""
try:
with self._status_mutex:
if self._loading_status == _ExplainManagerStatus.LOADING.value:
logger.info(
'Current status is %s, will ignore to load data.',
self._loading_status)
return
self._loading_status = _ExplainManagerStatus.LOADING.value
self._cache_loaders()
self._execute_loading()
if not self._loader_pool:
self._loading_status = _ExplainManagerStatus.INVALID.value
else:
self._loading_status = _ExplainManagerStatus.DONE.value
logger.info(
'Load event data end, status: %s, and loader pool size: %d',
self._loading_status, len(self._loader_pool))
except Exception as ex:
self._loading_status = _ExplainManagerStatus.INVALID.value
logger.exception(ex)
raise UnknownError(str(ex))
def _cache_loaders(self):
"""Cache explain loader in cache pool."""
dir_map_mtime_dict = []
_, summaries_info = self._summary_watcher.list_explain_directories(
self._summary_base_dir)
for summary_info in summaries_info:
summary_path = summary_info.get('relative_path')
summary_update_time = summary_info.get('update_time').timestamp()
if summary_path in self._loader_pool:
summary_update_time = max(
summary_update_time,
self._loader_pool[summary_path].query_time)
dir_map_mtime_dict.append((summary_info, summary_update_time))
sorted_summaries_info = sorted(dir_map_mtime_dict,
key=lambda x: x[1])[-_MAX_LOADERS_NUM:]
with self._loader_pool_mutex:
for summary_info, query_time in sorted_summaries_info:
summary_path = summary_info['relative_path']
if summary_path not in self._loader_pool:
loader = self._generate_loader_from_relative_path(
summary_path)
self._add_loader(loader)
else:
self._loader_pool[summary_path].query_time = query_time
self._loader_pool.move_to_end(summary_path, last=False)
def _generate_loader_from_relative_path(
self, relative_path: str) -> ExplainLoader:
"""Generate explain loader from the given relative path."""
self._check_summary_exist(relative_path)
current_dir = os.path.realpath(
FileHandler.join(self._summary_base_dir, relative_path))
loader_id = self._generate_loader_id(relative_path)
loader = ExplainLoader(loader_id=loader_id, summary_dir=current_dir)
return loader
def _add_loader(self, loader):
"""add loader to the loader_pool."""
if loader.train_id not in self._loader_pool:
self._loader_pool[loader.train_id] = loader
else:
self._loader_pool.move_to_end(loader.train_id)
while len(self._loader_pool) > self._max_loaders_num:
self._loader_pool.popitem(last=False)
def _execute_loading(self):
"""Execute the data loading."""
for loader_id in list(self._loader_pool.keys()):
try:
with self._loader_pool_mutex:
loader = self._loader_pool.get(loader_id, None)
if loader is None:
logger.debug(
'Loader %r has been deleted, will not load data',
loader_id)
return
loader.load()
except MindInsightException as ex:
logger.warning(
'Data loader %r load data failed. Delete data_loader. Detail: %s',
loader_id, ex)
with self._loader_pool_mutex:
self._delete_loader(loader_id)
def _delete_loader(self, loader_id):
"""delete loader given loader_id"""
if loader_id in self._loader_pool:
self._loader_pool.pop(loader_id)
logger.debug('delete loader %s', loader_id)
def _check_status_valid(self):
"""Check manager status."""
if self._loading_status == _ExplainManagerStatus.INIT.value:
raise exceptions.SummaryLogIsLoading(
'Data is loading, current status is %s' % self._loading_status)
def _check_summary_exist(self, loader_id):
"""Verify thee train_job is existed given loader_id."""
if not self._summary_watcher.is_summary_directory(
self._summary_base_dir, loader_id):
raise ParamValueError('Can not find the train job in the manager.')
def _reload_data_again(self):
"""Reload the data one more time."""
logger.debug('Start to reload data again.')
thread = threading.Thread(target=self._load_data,
name='reload_data_thread')
thread.daemon = False
thread.start()
@staticmethod
def _generate_loader_id(relative_path):
"""Generate loader id for given path"""
loader_id = relative_path
return loader_id
class DataLoaderGenerator(LoaderGenerator):
"""
DataLoaderGenerator generate a loader_dict of loader from summary logs.
Each loader helps deal the data of the events.
It helps DataManager to generate loaders.
"""
def __init__(self, summary_path):
"""
Init DataLoaderGenerator.
Args:
summary_path (str): A directory path, e.g. '/data/ImageNet/'.
"""
self._summary_path = self._check_and_normalize_summary_path(
summary_path)
self._summary_watcher = SummaryWatcher()
def _check_and_normalize_summary_path(self, summary_path):
"""
Check and normalize summary path.
Args:
summary_path (str): A directory path, e.g. '/data/ImageNet/'.
Returns:
str, normalized summary path.
"""
if summary_path is None:
logger.warning(
"Summary path is None. It will not init data loader generator."
)
raise ParamValueError("Summary path is None.")
summary_path = os.path.realpath(summary_path)
return summary_path
def generate_loaders(self, loader_pool):
"""
Generate loader from summary path, if summary path is empty, will return empty list.
Args:
loader_pool (dict[str, LoaderStruct]): Current loader pool in data_manager.
Returns:
dict[str, LoaderStruct], a dict of `Loader`.
"""
loader_dict = {}
if not FileHandler.exists(self._summary_path):
logger.warning(
"Summary path does not exist. It will not start loading events data. "
"Current path is %r.", self._summary_path)
return loader_dict
dir_map_mtime_dict = {}
min_modify_time = None
summaries_info = self._summary_watcher.list_summary_directories(
self._summary_path)
for item in summaries_info:
relative_path = item.get("relative_path")
current_dir = FileHandler.join(self._summary_path, relative_path)
dataloader = DataLoader(current_dir)
if not dataloader.has_valid_files():
logger.debug(
"Can not find valid train log file in folder %s , "
"will ignore.", relative_path)
continue
modify_time = item.get("update_time").timestamp()
# if loader exists in loader pool and newer time, update its time
loader_id = self._generate_loader_id(relative_path)
loader = loader_pool.get(loader_id)
if loader is not None and loader.latest_update_time > modify_time:
modify_time = loader.latest_update_time
if not min_modify_time:
# The first load, init min modify time
min_modify_time = modify_time
# We need to find `MAX_DATA_LOADER_SIZE` newly modified folders.
if len(dir_map_mtime_dict) < MAX_DATA_LOADER_SIZE:
if modify_time < min_modify_time:
min_modify_time = modify_time
dir_map_mtime_dict.update({relative_path: modify_time})
else:
if modify_time >= min_modify_time:
dir_map_mtime_dict.update({relative_path: modify_time})
sorted_dir_tuple = sorted(dir_map_mtime_dict.items(),
key=lambda d: d[1])[-MAX_DATA_LOADER_SIZE:]
for relative_path, modify_time in sorted_dir_tuple:
loader_id = self._generate_loader_id(relative_path)
loader = self._generate_loader_by_relative_path(relative_path)
loader_dict.update({loader_id: loader})
return loader_dict
def _generate_loader_by_relative_path(self, relative_path):
"""
Generate loader by relative path.
Args:
relative_path (str): Relative path of a summary directory, e.g. './log1'.
Returns:
dict[str, LoaderStruct], a dict of `Loader`.
"""
current_dir = os.path.realpath(
FileHandler.join(self._summary_path, relative_path))
data_loader = DataLoader(current_dir)
loader_id = self._generate_loader_id(relative_path)
loader = LoaderStruct(
loader_id=loader_id,
name=self._generate_loader_name(relative_path),
path=current_dir,
latest_update_time=FileHandler.file_stat(current_dir).mtime,
data_loader=data_loader)
return loader
def _generate_loader_id(self, relative_path):
"""
Generate loader id from relative path.
Args:
relative_path (str): Relative path of a summary directory, e.g. './log1'.
Returns:
str, loader_id for `Loader`.
"""
loader_id = relative_path
return loader_id
def _generate_loader_name(self, relative_path):
"""
Generate loader name from relative path.
Args:
relative_path (str): Relative path of a summary directory, e.g. './log1'.
Returns:
str, loader_name for `Loader`.
"""
loader_name = relative_path
return loader_name
def _get_relative_path_from_train_id(self, train_id):
"""
Get relative from train_id.
Args:
train_id (str): Train ID of a summary directory, e.g. './log1'.
Returns:
str, relative path of `Loader`.
"""
relative_path = train_id
return relative_path
def check_train_job_exist(self, train_id):
"""
Check if train job exists.
Args:
train_id (str): Train ID of a summary directory, e.g. './log1'.
Returns:
bool, if train job exists, return True.
"""
if not self._is_train_id_valid(train_id):
return False
relative_path = self._get_relative_path_from_train_id(train_id)
if self._summary_watcher.is_summary_directory(self._summary_path,
relative_path):
return True
return False
def _is_train_id_valid(self, train_id):
"""
Check if train_id is valid.
Args:
train_id (str): Train ID of a summary directory, e.g. './log1'.
Returns:
bool, if train id is valid, return True.
"""
if not train_id.startswith('./'):
logger.warning("The train_id does not start with './'.")
return False
if len(train_id.split("/")) > 2:
logger.warning("The train_id contains multiple '/'.")
return False
return True
def generate_loader_by_train_id(self, train_id):
"""
Generate loader by train_id.
Args:
train_id (str): Train ID of a summary directory, e.g. './log1'.
Returns:
dict[str, LoaderStruct], a dict of `Loader`.
"""
relative_path = self._get_relative_path_from_train_id(train_id)
loader = self._generate_loader_by_relative_path(relative_path)
return loader