def _has_event_file_been_skipped(self, missing_event_file_name: str) -> bool:
"""
Checks if an event file will ever be downloaded.
if event_file present --> return False
if the worker has written the next event file --> return True
if none of the above --> return False
:param missing_event_file_name:
:return:
"""
self.logger.info(f" Index Reader: Event File {missing_event_file_name} not found.")
missing_worker = parse_worker_name_from_file(missing_event_file_name)
missing_step = IndexFileLocationUtils.parse_step_from_index_file_name(
missing_event_file_name
)
event_files = self.list_event_files(missing_event_file_name)
for event_file in event_files:
if missing_worker == parse_worker_name_from_file(event_file):
step = IndexFileLocationUtils.parse_step_from_index_file_name(event_file)
if missing_step == step:
"""
The missing step file may have been written to disk before
we perform the list operation.
"""
return False
self.logger.warn(
f" Index Reader: Event File {missing_event_file_name} was written but not found "
f"\nHowever Event File {event_file} found."
)
self.logger.warn(f"IndexReader: Skipping {missing_event_file_name} ")
return True
return False
def _update_last_index_token(self, new_index_token: str) -> None:
"""
This function updates the last_index_token in the following scenarios:
1. last_complete_step >= last_index_token_step :
this means that the token isn't pointing to the latest completed step
2. number of steps available ( complete or incomplete ) - (last_completed_step+1) > window_size_limit:
we maintain a window to stop querying for older steps that have not completed.
if the total number of steps, we are querying for completion is greater than our window_size_limit
we update the last_index_token and last_complete_step by (window_size_limit // 2)
:param new_index_token:
:return:None
"""
if self.last_index_token is None:
last_index_token_step = 0
else:
last_index_token_step = IndexFileLocationUtils.parse_step_from_index_file_name(
self.last_index_token)
# Case 1: This case is not satisfied when all workers in a
# distributed training job have not written a step
if self.last_complete_step >= last_index_token_step:
prefix = IndexFileLocationUtils.get_prefix_from_index_file(
new_index_token)
# sort lexicographically and select the last worker
last_worker = sorted(list(self.worker_set))[-1]
# below converts worker_name to serialized workerName
# if it's a tf device, else no effect
last_worker_serialized = serialize_tf_device(last_worker)
self.last_index_token = IndexFileLocationUtils.get_index_key_for_step(
prefix, self.last_complete_step, last_worker_serialized)
self.logger.debug(
f"Updated last index token to:{self.last_index_token}")
# Case 2: This case is satisfied if the number of incomplete steps
# is greater than the INCOMPLETE_STEP_WAIT_WINDOW
available_step = self._global_to_mode.keys()
if (len(available_step) - (self.last_complete_step + 1) >
self._incomplete_wait_for_step_window):
prefix = IndexFileLocationUtils.get_prefix_from_index_file(
new_index_token)
last_worker = sorted(list(self.worker_set))[-1]
# below converts worker_name to serialized workerName
# if it's a tf device, else no effect
last_worker_serialized = serialize_tf_device(last_worker)
self.last_index_token = IndexFileLocationUtils.get_index_key_for_step(
prefix,
self.last_complete_step +
(self._incomplete_wait_for_step_window // 2),
last_worker_serialized,
)
self.last_complete_step = IndexFileLocationUtils.parse_step_from_index_file_name(
self.last_index_token)
self.logger.info(
f"Waiting for: {len(available_step) - (self.last_complete_step + 1)} Steps. \n"
f"INCOMPLETE_STEP_WAIT_WINDOW: {self._incomplete_wait_for_step_window}. \n"
f"Marking the last {self._incomplete_wait_for_step_window // 2} incomplete steps as complete"
f"Updating last_index_token to: {self.last_index_token}. \n"
f"Updating last_complete_step to: {self.last_complete_step}. ")
def read_index_files(
self, start_after_key: str, range_steps=None
) -> Tuple[List[bytes], list, str, List[str]]:
"""
Read files like `trial_{datetime}/index/000/{step}_{worker}.json.
:param start_after_key: str
:param range_steps: str
:return: Tuple( responses, steps, start_after_key, workers)
"""
index_files = self.list_index_files()
steps = []
workers = []
responses = []
if start_after_key is not None:
start_after_index = bisect_left(index_files, start_after_key)
else:
start_after_index = 0
index_files = index_files[start_after_index:] # ignore files we have already read
for index_file in index_files:
if self.index_file_cache.has_not_read(index_file):
step = IndexFileLocationUtils.parse_step_from_index_file_name(index_file)
if (
range_steps is not None and step_in_range(range_steps, step)
) or range_steps is None:
steps.append(step)
workers.append(parse_worker_name_from_file(index_file))
self.logger.debug(
f"Sagemaker-Debugger: Read {os.path.getsize(index_file)} bytes from file {index_file}"
)
with open(index_file) as f:
responses.append(f.read().encode())
self.index_file_cache.add(index_file, start_after_key)
if len(index_files) > 0:
start_after_key = index_files[-1] # Last file that we have read
return responses, steps, start_after_key, workers
def read_index_files(
self, start_after_key: str, range_steps=None
) -> Tuple[List[bytes], list, str, List[str]]:
"""
Read files like `trial_{datetime}/index/000/{step}_{worker}.json.
:param start_after_key: str
:param range_steps:
:return: Tuple( responses, steps, start_after_key, workers)
"""
object_requests = []
steps = []
workers = []
index_files, start_after_key = self.list_index_files(start_after_key)
self.logger.debug(f'Loaded Index Files: {",".join(index_files)}')
for index_file in index_files:
if self.index_file_cache.has_not_read(index_file):
step = IndexFileLocationUtils.parse_step_from_index_file_name(index_file)
if (
range_steps is not None and step_in_range(range_steps, step)
) or range_steps is None:
steps.append(step)
workers.append(parse_worker_name_from_file(index_file))
object_requests.append(
ReadObjectRequest(format(f"s3://{self.bucket_name}/") + index_file)
)
self.index_file_cache.add(index_file, start_after_key)
responses = self.s3_handler.get_objects(object_requests)
return responses, steps, start_after_key, workers