Modify to meet precommit requirements.

dlrover/python/elastic_agent/torch/ckpt_saver.py

@@ -697,8 +697,8 @@ def _sync_node_checkpoint(
  self, master_client: MasterClient, step: int, timeout: int
  ):
  """
- Check whether all training nodes can save the checkpoint from the memory
- to the storage. If some nodes fail, other nodes need not save.
+ Check whether all training nodes can save the checkpoint from the
+ memory to the storage. If some nodes fail, other nodes need not save.
  """
  start = time.time()
  while True:

dlrover/trainer/torch/flash_checkpoint/engine.py

@@ -149,7 +149,7 @@ class CheckpointEngine(metaclass=ABCMeta):
  Args:
  checkpoint_dir (str): the directory to save checkpoint.
  storage: a CheckpointStorage instance to write/read the storage.
- comm_backend (str): the communication backend to create a process group,
+ comm_backend (str): communication backend to create a process group,
  The default is the backend of general main process group.
  """
 
@@ -260,7 +260,7 @@ def _notify_agent_to_create_saver(self):
  """Notify the agent in the main process to create a checkpoint saver"""
  if self._local_rank != 0:
  return
- # the agent side will release the lock if the training process restarts.
+ # agent side will release the lock if the training process restarts.
  queue = SharedQueue(name="factory")
 
  local_shard_num = self.get_local_shard_num()
@@ -398,7 +398,7 @@ def save_to_memory(self, step, state_dict, paths: Dict[str, str]):
 
  Args:
  step (int): the global iteration step.
- state_dict (dict): the state dict of the model and optimizer to save.
+ state_dict (dict): state dict of the model and optimizer to save.
  paths (dict): the key is a category in
  ["model_states", "optim_states"] of the state dict, and
  the value is the path of storage to save.
@@ -417,7 +417,7 @@ def save_to_storage(self, step, state_dict, paths: Dict[str, str]):
 
  Args:
  step (int): the iteration step.
- state_dict (dict): the state dict of the model and optimizer to save.
+ state_dict (dict): state dict of the model and optimizer to save.
  paths (dict): the key is a category in
  ["model_states", "optim_states"] of the state dict, and
  the value is the path of storage to save.

dlrover/trainer/torch/flash_checkpoint/full_ckpt_engine.py

@@ -107,7 +107,7 @@ def save_to_memory(self, step, state_dict, paths: Dict[str, str]):
 
  Args:
  step (int): the global iteration step.
- state_dict (dict): the state dict of the model and optimizer to save.
+ state_dict (dict): state dict of the model and optimizer to save.
  paths (dict): the key is a category in
  ["model_states", "optim_states"] of the state dict, and
  the value is the path of storage to save.
@@ -126,7 +126,7 @@ def save_to_storage(self, step, state_dict, paths):
 
  Args:
  step (int): the global iteration step.
- state_dict (dict): the state dict of the model and optimizer to save.
+ state_dict (dict): state dict of the model and optimizer to save.
  paths (dict): the key is a category in
  ["model_states", "optim_states"] of the state dict, and
  the value is the path of storage to save.
@@ -144,7 +144,7 @@ def save_to_storage(self, step, state_dict, paths):
 
  def load(self, resume_path=""):
  """
- The method will first try to load the state dict from the shared memory.
+ The method will first try to load state dict from the shared memory.
  If there is no state dict in the shared memory, the method will
  load the state dict from the storage.