import copy
from datetime import datetime
import logging
import os
from pathlib import Path
import platform
import sys
import tempfile
import time
from contextlib import redirect_stderr, redirect_stdout
from typing import TYPE_CHECKING, Any, Callable, Dict, List, Optional, Union
import ray
import ray.cloudpickle as ray_pickle
from ray.air._internal.util import skip_exceptions, exception_cause
from ray.air.constants import (
TIMESTAMP,
TIME_THIS_ITER_S,
TRAINING_ITERATION,
)
from ray.train._internal.checkpoint_manager import _TrainingResult
from ray.train._internal.storage import StorageContext, _exists_at_fs_path
from ray.train import Checkpoint
from ray.train.constants import DEFAULT_STORAGE_PATH
from ray.tune.result import (
DEBUG_METRICS,
DONE,
EPISODES_THIS_ITER,
EPISODES_TOTAL,
HOSTNAME,
NODE_IP,
PID,
RESULT_DUPLICATE,
SHOULD_CHECKPOINT,
STDERR_FILE,
STDOUT_FILE,
TIME_TOTAL_S,
TIMESTEPS_THIS_ITER,
TIMESTEPS_TOTAL,
TRIAL_ID,
TRIAL_INFO,
)
from ray.tune.utils import UtilMonitor
from ray.tune.utils.log import disable_ipython
from ray.tune.utils.util import Tee
from ray.tune.execution.placement_groups import PlacementGroupFactory
from ray.util.annotations import DeveloperAPI, PublicAPI
if TYPE_CHECKING:
from ray.tune.logger import Logger
logger = logging.getLogger(__name__)
SETUP_TIME_THRESHOLD = 10
# File containing dict data returned by user from `Trainable.save_checkpoint`
_DICT_CHECKPOINT_FILE_NAME = "_dict_checkpoint.pkl"
[docs]@PublicAPI
class Trainable:
"""Abstract class for trainable models, functions, etc.
A call to ``train()`` on a trainable will execute one logical iteration of
training. As a rule of thumb, the execution time of one train call should
be large enough to avoid overheads (i.e. more than a few seconds), but
short enough to report progress periodically (i.e. at most a few minutes).
Calling ``save()`` should save the training state of a trainable to disk,
and ``restore(path)`` should restore a trainable to the given state.
Generally you only need to implement ``setup``, ``step``,
``save_checkpoint``, and ``load_checkpoint`` when subclassing Trainable.
Other implementation methods that may be helpful to override are
``log_result``, ``reset_config``, ``cleanup``, and ``_export_model``.
Tune will convert this class into a Ray actor, which runs on a separate process.
By default, Tune will also change the current working directory of this process to
its corresponding trial-level log directory ``self.logdir``.
This is designed so that different trials that run on the same physical node won't
accidentally write to the same location and overstep each other.
The behavior of changing the working directory can be disabled by setting the
`RAY_CHDIR_TO_TRIAL_DIR=0` environment variable. This allows access to files
in the original working directory, but relative paths should be used for read only
purposes, and you must make sure that the directory is synced on all nodes if
running on multiple machines.
The `TUNE_ORIG_WORKING_DIR` environment variable was the original workaround for
accessing paths relative to the original working directory. This environment
variable is deprecated, and the `RAY_CHDIR_TO_TRIAL_DIR` environment variable
described above should be used instead.
This class supports checkpointing to and restoring from remote storage.
"""
[docs] def __init__(
self,
config: Dict[str, Any] = None,
logger_creator: Callable[[Dict[str, Any]], "Logger"] = None, # Deprecated (2.7)
storage: Optional[StorageContext] = None,
):
"""Initialize a Trainable.
Sets up logging and points ``self.logdir`` to a directory in which
training outputs should be placed.
Subclasses should prefer defining ``setup()`` instead of overriding
``__init__()`` directly.
Args:
config: Trainable-specific configuration data. By default
will be saved as ``self.config``.
logger_creator: (Deprecated) Function that creates a ray.tune.Logger
object. If unspecified, a default logger is created.
storage: StorageContext object that contains persistent storage paths
"""
self.config = config or {}
trial_info = self.config.pop(TRIAL_INFO, None)
if self.is_actor():
disable_ipython()
# TODO(ml-team): Remove `logger_creator` in 2.7.
# TODO(justinvyu): Rename/remove logdir.
self._result_logger = self._logdir = None
self._create_logger(self.config, logger_creator)
self._stdout_context = self._stdout_fp = self._stdout_stream = None
self._stderr_context = self._stderr_fp = self._stderr_stream = None
self._stderr_logging_handler = None
stdout_file = self.config.pop(STDOUT_FILE, None)
stderr_file = self.config.pop(STDERR_FILE, None)
self._iteration = 0
self._time_total = 0.0
self._timesteps_total = None
self._episodes_total = None
self._time_since_restore = 0.0
self._timesteps_since_restore = 0
self._iterations_since_restore = 0
self._last_result = None
self._restored = False
self._trial_info = trial_info
self._stdout_file = stdout_file
self._stderr_file = stderr_file
self._start_time = time.time()
self._local_ip = ray.util.get_node_ip_address()
self._storage = storage
if storage:
assert storage.trial_fs_path
logger.debug(f"StorageContext on the TRAINABLE:\n{storage}")
self._open_logfiles(stdout_file, stderr_file)
self.setup(copy.deepcopy(self.config))
setup_time = time.time() - self._start_time
if setup_time > SETUP_TIME_THRESHOLD:
logger.info(
"Trainable.setup took {:.3f} seconds. If your "
"trainable is slow to initialize, consider setting "
"reuse_actors=True to reduce actor creation "
"overheads.".format(setup_time)
)
log_sys_usage = self.config.get("log_sys_usage", False)
self._monitor = UtilMonitor(start=log_sys_usage)
[docs] @classmethod
def default_resource_request(
cls, config: Dict[str, Any]
) -> Optional[PlacementGroupFactory]:
"""Provides a static resource requirement for the given configuration.
This can be overridden by sub-classes to set the correct trial resource
allocation, so the user does not need to.
.. testcode::
@classmethod
def default_resource_request(cls, config):
return PlacementGroupFactory([{"CPU": 1}, {"CPU": 1}])
Args:
config[Dict[str, Any]]: The Trainable's config dict.
Returns:
PlacementGroupFactory: A PlacementGroupFactory consumed by Tune
for queueing.
"""
return None
[docs] @classmethod
def resource_help(cls, config: Dict):
"""Returns a help string for configuring this trainable's resources.
Args:
config: The Trainer's config dict.
"""
return ""
def get_current_ip_pid(self):
return self._local_ip, os.getpid()
[docs] def get_auto_filled_metrics(
self,
now: Optional[datetime] = None,
time_this_iter: Optional[float] = None,
timestamp: Optional[int] = None,
debug_metrics_only: bool = False,
) -> dict:
"""Return a dict with metrics auto-filled by the trainable.
If ``debug_metrics_only`` is True, only metrics that don't
require at least one iteration will be returned
(``ray.tune.result.DEBUG_METRICS``).
"""
if now is None:
now = datetime.today()
autofilled = {
TRIAL_ID: self.trial_id,
"date": now.strftime("%Y-%m-%d_%H-%M-%S"),
"timestamp": timestamp if timestamp else int(time.mktime(now.timetuple())),
TIME_THIS_ITER_S: time_this_iter,
TIME_TOTAL_S: self._time_total,
PID: os.getpid(),
HOSTNAME: platform.node(),
NODE_IP: self._local_ip,
"config": self.config,
"time_since_restore": self._time_since_restore,
"iterations_since_restore": self._iterations_since_restore,
}
if self._timesteps_since_restore:
autofilled["timesteps_since_restore"] = self._timesteps_since_restore
if debug_metrics_only:
autofilled = {k: v for k, v in autofilled.items() if k in DEBUG_METRICS}
return autofilled
def is_actor(self):
try:
actor_id = ray._private.worker.global_worker.actor_id
return actor_id != actor_id.nil()
except Exception:
# If global_worker is not instantiated, we're not in an actor
return False
[docs] def train_buffered(self, buffer_time_s: float, max_buffer_length: int = 1000):
"""Runs multiple iterations of training.
Calls ``train()`` internally. Collects and combines multiple results.
This function will run ``self.train()`` repeatedly until one of
the following conditions is met: 1) the maximum buffer length is
reached, 2) the maximum buffer time is reached, or 3) a checkpoint
was created. Even if the maximum time is reached, it will always
block until at least one result is received.
Args:
buffer_time_s: Maximum time to buffer. The next result
received after this amount of time has passed will return
the whole buffer.
max_buffer_length: Maximum number of results to buffer.
"""
results = []
now = time.time()
send_buffer_at = now + buffer_time_s
while now < send_buffer_at or not results: # At least one result
result = self.train()
results.append(result)
if result.get(DONE, False):
# If the trial is done, return
break
elif result.get(SHOULD_CHECKPOINT, False):
# If a checkpoint was created, return
break
elif result.get(RESULT_DUPLICATE):
# If the function API trainable completed, return
break
elif len(results) >= max_buffer_length:
# If the buffer is full, return
break
now = time.time()
return results
[docs] def train(self):
"""Runs one logical iteration of training.
Calls ``step()`` internally. Subclasses should override ``step()``
instead to return results.
This method automatically fills the following fields in the result:
`done` (bool): training is terminated. Filled only if not provided.
`time_this_iter_s` (float): Time in seconds this iteration
took to run. This may be overridden in order to override the
system-computed time difference.
`time_total_s` (float): Accumulated time in seconds for this
entire experiment.
`training_iteration` (int): The index of this
training iteration, e.g. call to train(). This is incremented
after `step()` is called.
`pid` (str): The pid of the training process.
`date` (str): A formatted date of when the result was processed.
`timestamp` (str): A UNIX timestamp of when the result
was processed. This may be overridden.
`hostname` (str): Hostname of the machine hosting the training
process.
`node_ip` (str): Node ip of the machine hosting the training
process.
Returns:
A dict that describes training progress.
"""
start = time.time()
try:
result = self.step()
except Exception as e:
skipped = skip_exceptions(e)
raise skipped from exception_cause(skipped)
assert isinstance(result, dict), "step() needs to return a dict."
# We do not modify internal state nor update this result if duplicate.
if RESULT_DUPLICATE in result:
return result
result = result.copy()
self._iteration += 1
self._iterations_since_restore += 1
if result.get(TIME_THIS_ITER_S) is not None:
time_this_iter = result[TIME_THIS_ITER_S]
else:
time_this_iter = time.time() - start
self._time_total += time_this_iter
self._time_since_restore += time_this_iter
result_timestamp = result.get(TIMESTAMP, None)
result.setdefault(DONE, False)
# self._timesteps_total should only be tracked if increments are provided
if result.get(TIMESTEPS_THIS_ITER) is not None:
if self._timesteps_total is None:
self._timesteps_total = 0
self._timesteps_total += result[TIMESTEPS_THIS_ITER]
self._timesteps_since_restore += result[TIMESTEPS_THIS_ITER]
# self._episodes_total should only be tracked if increments provided
if result.get(EPISODES_THIS_ITER) is not None:
if self._episodes_total is None:
self._episodes_total = 0
self._episodes_total += result[EPISODES_THIS_ITER]
# self._timesteps_total should not override user-provided total
if self._timesteps_total is not None:
result.setdefault(TIMESTEPS_TOTAL, self._timesteps_total)
if self._episodes_total is not None:
result.setdefault(EPISODES_TOTAL, self._episodes_total)
result.setdefault(TRAINING_ITERATION, self._iteration)
now = datetime.today()
result.update(
self.get_auto_filled_metrics(
now=now, time_this_iter=time_this_iter, timestamp=result_timestamp
)
)
monitor_data = self._monitor.get_data()
if monitor_data:
result.update(monitor_data)
self.log_result(result)
if self._stdout_context:
self._stdout_stream.flush()
if self._stderr_context:
self._stderr_stream.flush()
self._last_result = result
if self._storage:
# Launch background tasks to sync artifacts at some specified frequency.
self._storage.persist_artifacts()
return result
def get_state(self):
return {
"iteration": self._iteration,
"timesteps_total": self._timesteps_total,
"time_total": self._time_total,
"episodes_total": self._episodes_total,
"last_result": self._last_result,
"ray_version": ray.__version__,
}
def _create_checkpoint_dir(
self, checkpoint_dir: Optional[str] = None
) -> Optional[str]:
# NOTE: There's no need to supply the checkpoint directory inside
# the local trial dir, since it'll get persisted to the right location.
if checkpoint_dir:
os.makedirs(checkpoint_dir, exist_ok=True)
return checkpoint_dir
else:
return tempfile.mkdtemp()
[docs] @DeveloperAPI
def save(self, checkpoint_dir: Optional[str] = None) -> _TrainingResult:
"""Saves the current model state to a checkpoint.
Subclasses should override ``save_checkpoint()`` instead to save state.
Args:
checkpoint_dir: Optional dir to place the checkpoint.
Returns:
The given or created checkpoint directory.
Note the return value matches up with what is expected of `restore()`.
"""
checkpoint_dir = self._create_checkpoint_dir(checkpoint_dir=checkpoint_dir)
# User saves checkpoint
checkpoint_dict_or_path = self.save_checkpoint(checkpoint_dir)
if not isinstance(self, ray.tune.trainable.FunctionTrainable):
# TODO(justinvyu): [cls_trainable_support]
# This is to get class Trainables to work in the new persistence mode.
# Need to handle checkpoint_dict_or_path == path, dict, or None
# Also need to upload to cloud, since `train.report` never gets called.
if isinstance(checkpoint_dict_or_path, dict):
with Path(checkpoint_dir, _DICT_CHECKPOINT_FILE_NAME).open("wb") as f:
ray_pickle.dump(checkpoint_dict_or_path, f)
elif isinstance(checkpoint_dict_or_path, str):
if checkpoint_dict_or_path != checkpoint_dir:
raise ValueError(
"The returned checkpoint path from `save_checkpoint` "
"must be None or the same as the provided path argument."
f"Got {checkpoint_dict_or_path} != {checkpoint_dir}"
)
local_checkpoint = Checkpoint.from_directory(checkpoint_dir)
metrics = self._last_result.copy() if self._last_result else {}
if self._storage:
# The checkpoint index is updated with the current result.
# NOTE: This is no longer using "iteration" as the folder indexing
# to be consistent with fn trainables.
self._storage._update_checkpoint_index(metrics)
persisted_checkpoint = self._storage.persist_current_checkpoint(
local_checkpoint
)
checkpoint_result = _TrainingResult(
checkpoint=persisted_checkpoint, metrics=metrics
)
# Persist trial artifacts to storage.
self._storage.persist_artifacts(
force=self._storage.sync_config.sync_artifacts_on_checkpoint
)
else:
# `storage=None` only happens when initializing the
# Trainable manually, outside of Tune/Train.
# In this case, no storage is set, so the default behavior
# is to just not upload anything and report a local checkpoint.
# This is fine for the main use case of local debugging.
checkpoint_result = _TrainingResult(
checkpoint=local_checkpoint, metrics=metrics
)
else:
checkpoint_result: _TrainingResult = checkpoint_dict_or_path
assert self._last_result
# Update the checkpoint result to include auto-filled metrics.
checkpoint_result.metrics.update(self._last_result)
assert isinstance(checkpoint_result, _TrainingResult)
return checkpoint_result
[docs] @DeveloperAPI
def restore(self, checkpoint_path: Union[str, Checkpoint, _TrainingResult]):
"""Restores training state from a given model checkpoint.
These checkpoints are returned from calls to save().
Subclasses should override ``load_checkpoint()`` instead to
restore state.
This method restores additional metadata saved with the checkpoint.
`checkpoint_path` should match with the return from ``save()``.
Args:
checkpoint_path: Path to restore checkpoint from. If this
path does not exist on the local node, it will be fetched
from external (cloud) storage if available, or restored
from a remote node.
checkpoint_node_ip: If given, try to restore
checkpoint from this node if it doesn't exist locally or
on cloud storage.
fallback_to_latest: If True, will try to recover the
latest available checkpoint if the given ``checkpoint_path``
could not be found.
"""
# TODO(justinvyu): Clean up this interface
if isinstance(checkpoint_path, str):
checkpoint_path = Checkpoint.from_directory(checkpoint_path)
if isinstance(checkpoint_path, Checkpoint):
checkpoint_result = _TrainingResult(checkpoint=checkpoint_path, metrics={})
else:
checkpoint_result: _TrainingResult = checkpoint_path
assert isinstance(checkpoint_result, _TrainingResult), type(checkpoint_result)
checkpoint = checkpoint_result.checkpoint
checkpoint_metrics = checkpoint_result.metrics
self._iteration = checkpoint_metrics.get(TRAINING_ITERATION, 0)
self._time_total = checkpoint_metrics.get(TIME_TOTAL_S, 0)
self._time_since_restore = 0.0
self._iterations_since_restore = 0
# TODO(justinvyu): This stuff should be moved to rllib.
self._timesteps_total = checkpoint_metrics.get(TIMESTEPS_TOTAL)
self._timesteps_since_restore = 0
self._episodes_total = checkpoint_metrics.get(EPISODES_TOTAL)
if not _exists_at_fs_path(checkpoint.filesystem, checkpoint.path):
raise ValueError(
f"Could not recover from checkpoint as it does not exist on "
f"storage anymore. "
f"Got storage fs type `{checkpoint.filesystem.type_name}` and "
f"path: {checkpoint.path}"
)
# TODO(justinvyu): [cls_trainable_support]
# This is to conform to the public class Trainable `load_checkpoint` API.
if not isinstance(self, ray.tune.trainable.FunctionTrainable):
# Need to convert Checkpoint -> local path or dict
# (depending on what the output of save_checkpoint was)
with checkpoint.as_directory() as checkpoint_dir:
checkpoint_path = Path(checkpoint_dir)
dict_checkpoint_file = checkpoint_path / _DICT_CHECKPOINT_FILE_NAME
if dict_checkpoint_file.exists():
# If this was a dict checkpoint, load it as a dict
with open(dict_checkpoint_file, "rb") as f:
checkpoint_dict = ray_pickle.load(f)
self.load_checkpoint(checkpoint_dict)
else:
self.load_checkpoint(checkpoint_dir)
else:
# TODO(justinvyu): The Function Trainable case doesn't conform
# to the load_checkpoint API at the moment.
self.load_checkpoint(checkpoint_result)
self._restored = True
logger.info(f"Restored on {self._local_ip} from checkpoint: {checkpoint}")
[docs] def export_model(
self, export_formats: Union[List[str], str], export_dir: Optional[str] = None
):
"""Exports model based on export_formats.
Subclasses should override _export_model() to actually
export model to local directory.
Args:
export_formats: Format or list of (str) formats
that should be exported.
export_dir: Optional dir to place the exported model.
Defaults to self.logdir.
Returns:
A dict that maps ExportFormats to successfully exported models.
"""
if isinstance(export_formats, str):
export_formats = [export_formats]
export_dir = export_dir or self.logdir
return self._export_model(export_formats, export_dir)
[docs] def reset(self, new_config, logger_creator=None, storage=None):
"""Resets trial for use with new config.
Subclasses should override reset_config() to actually
reset actor behavior for the new config."""
self.config = new_config
self._storage = storage
trial_info = new_config.pop(TRIAL_INFO, None)
if trial_info:
self._trial_info = trial_info
self._result_logger.flush()
self._result_logger.close()
if logger_creator:
logger.debug("Logger reset.")
self._create_logger(new_config.copy(), logger_creator)
else:
logger.debug(
"Did not reset logger. Got: "
f"trainable.reset(logger_creator={logger_creator})."
)
stdout_file = new_config.pop(STDOUT_FILE, None)
stderr_file = new_config.pop(STDERR_FILE, None)
self._close_logfiles()
self._open_logfiles(stdout_file, stderr_file)
success = self.reset_config(new_config)
if not success:
return False
# Reset attributes. Will be overwritten by `restore` if a checkpoint
# is provided.
self._iteration = 0
self._time_total = 0.0
self._timesteps_total = None
self._episodes_total = None
self._time_since_restore = 0.0
self._timesteps_since_restore = 0
self._iterations_since_restore = 0
self._restored = False
return True
[docs] def reset_config(self, new_config: Dict):
"""Resets configuration without restarting the trial.
This method is optional, but can be implemented to speed up algorithms
such as PBT, and to allow performance optimizations such as running
experiments with reuse_actors=True.
Args:
new_config: Updated hyperparameter configuration
for the trainable.
Returns:
True if reset was successful else False.
"""
return False
def _create_logger(
self,
config: Dict[str, Any],
logger_creator: Callable[[Dict[str, Any]], "Logger"] = None,
):
"""Create logger from logger creator.
Sets _logdir and _result_logger.
`_logdir` is the **per trial** directory for the Trainable.
"""
if logger_creator:
self._result_logger = logger_creator(config)
self._logdir = self._result_logger.logdir
else:
from ray.tune.logger import UnifiedLogger
logdir_prefix = datetime.today().strftime("%Y-%m-%d_%H-%M-%S")
ray._private.utils.try_to_create_directory(DEFAULT_STORAGE_PATH)
self._logdir = tempfile.mkdtemp(
prefix=logdir_prefix, dir=DEFAULT_STORAGE_PATH
)
self._result_logger = UnifiedLogger(config, self._logdir, loggers=None)
def _open_logfiles(self, stdout_file, stderr_file):
"""Create loggers. Open stdout and stderr logfiles."""
if stdout_file:
stdout_path = (Path(self._logdir) / stdout_file).expanduser().as_posix()
self._stdout_fp = open(stdout_path, "a+")
self._stdout_stream = Tee(sys.stdout, self._stdout_fp)
self._stdout_context = redirect_stdout(self._stdout_stream)
self._stdout_context.__enter__()
if stderr_file:
stderr_path = (Path(self._logdir) / stderr_file).expanduser().as_posix()
self._stderr_fp = open(stderr_path, "a+")
self._stderr_stream = Tee(sys.stderr, self._stderr_fp)
self._stderr_context = redirect_stderr(self._stderr_stream)
self._stderr_context.__enter__()
# Add logging handler to root ray logger
formatter = logging.Formatter(
"[%(levelname)s %(asctime)s] "
"%(filename)s: %(lineno)d "
"%(message)s"
)
self._stderr_logging_handler = logging.StreamHandler(self._stderr_fp)
self._stderr_logging_handler.setFormatter(formatter)
ray.logger.addHandler(self._stderr_logging_handler)
def _close_logfiles(self):
"""Close stdout and stderr logfiles."""
if self._stderr_logging_handler:
ray.logger.removeHandler(self._stderr_logging_handler)
if self._stdout_context:
self._stdout_stream.flush()
self._stdout_context.__exit__(None, None, None)
self._stdout_fp.close()
self._stdout_context = None
if self._stderr_context:
self._stderr_stream.flush()
self._stderr_context.__exit__(None, None, None)
self._stderr_fp.close()
self._stderr_context = None
[docs] def stop(self):
"""Releases all resources used by this trainable.
Calls ``Trainable.cleanup`` internally. Subclasses should override
``Trainable.cleanup`` for custom cleanup procedures.
"""
self._result_logger.flush()
self._result_logger.close()
if self._monitor.is_alive():
self._monitor.stop()
self._monitor.join()
self.cleanup()
self._close_logfiles()
@property
def logdir(self):
"""Directory of the results and checkpoints for this Trainable.
Note that the current working directory will also be changed to this.
"""
return self._logdir
@property
def trial_name(self):
"""Trial name for the corresponding trial of this Trainable.
This is not set if not using Tune.
.. testcode::
from ray.tune import Trainable
name = Trainable().trial_name
"""
if self._trial_info:
return self._trial_info.trial_name
else:
return "default"
@property
def trial_id(self):
"""Trial ID for the corresponding trial of this Trainable.
This is not set if not using Tune.
.. testcode::
from ray.tune import Trainable
trial_id = Trainable().trial_id
"""
if self._trial_info:
return self._trial_info.trial_id
else:
return "default"
@property
def trial_resources(self) -> Optional[PlacementGroupFactory]:
"""Resources currently assigned to the trial of this Trainable.
This is not set if not using Tune.
.. testcode::
from ray.tune import Trainable
trial_resources = Trainable().trial_resources
"""
if self._trial_info:
return self._trial_info.trial_resources
else:
return None
@property
def iteration(self):
"""Current training iteration.
This value is automatically incremented every time `train()` is called
and is automatically inserted into the training result dict.
"""
return self._iteration
@property
def training_iteration(self):
"""Current training iteration (same as `self.iteration`).
This value is automatically incremented every time `train()` is called
and is automatically inserted into the training result dict.
"""
return self._iteration
[docs] def get_config(self):
"""Returns configuration passed in by Tune."""
return self.config
[docs] def step(self):
"""Subclasses should override this to implement train().
The return value will be automatically passed to the loggers. Users
can also return `tune.result.DONE` or `tune.result.SHOULD_CHECKPOINT`
as a key to manually trigger termination or checkpointing of this
trial. Note that manual checkpointing only works when subclassing
Trainables.
.. versionadded:: 0.8.7
Returns:
A dict that describes training progress.
"""
raise NotImplementedError
[docs] def save_checkpoint(self, checkpoint_dir: str) -> Optional[Dict]:
"""Subclasses should override this to implement ``save()``.
Warning:
Do not rely on absolute paths in the implementation of
``Trainable.save_checkpoint`` and ``Trainable.load_checkpoint``.
Use ``validate_save_restore`` to catch ``Trainable.save_checkpoint``/
``Trainable.load_checkpoint`` errors before execution.
>>> from ray.tune.utils import validate_save_restore
>>> MyTrainableClass = ... # doctest: +SKIP
>>> validate_save_restore(MyTrainableClass) # doctest: +SKIP
.. versionadded:: 0.8.7
Args:
checkpoint_dir: The directory where the checkpoint
file must be stored. In a Tune run, if the trial is paused,
the provided path may be temporary and moved.
Returns:
A dict or None. If dict, the return value will
be automatically serialized by Tune. In that case,
``Trainable.load_checkpoint()`` will receive the dict upon restore.
Example:
>>> trainable, trainable1, trainable2 = ... # doctest: +SKIP
>>> print(trainable1.save_checkpoint("/tmp/checkpoint_1")) # doctest: +SKIP
"/tmp/checkpoint_1"
>>> print(trainable2.save_checkpoint("/tmp/checkpoint_2")) # doctest: +SKIP
{"some": "data"}
>>> trainable.save_checkpoint("/tmp/bad_example") # doctest: +SKIP
"/tmp/NEW_CHECKPOINT_PATH/my_checkpoint_file" # This will error.
"""
raise NotImplementedError
[docs] def load_checkpoint(self, checkpoint: Optional[Dict]):
"""Subclasses should override this to implement restore().
Warning:
In this method, do not rely on absolute paths. The absolute
path of the checkpoint_dir used in ``Trainable.save_checkpoint``
may be changed.
If ``Trainable.save_checkpoint`` returned a prefixed string, the
prefix of the checkpoint string returned by
``Trainable.save_checkpoint`` may be changed.
This is because trial pausing depends on temporary directories.
The directory structure under the checkpoint_dir provided to
``Trainable.save_checkpoint`` is preserved.
See the examples below.
Example:
>>> import os
>>> from ray.tune.trainable import Trainable
>>> class Example(Trainable):
... def save_checkpoint(self, checkpoint_path):
... my_checkpoint_path = os.path.join(checkpoint_path, "my/path")
... return my_checkpoint_path
... def load_checkpoint(self, my_checkpoint_path):
... print(my_checkpoint_path)
>>> trainer = Example()
>>> # This is used when PAUSED.
>>> checkpoint_result = trainer.save() # doctest: +SKIP
>>> trainer.restore(checkpoint_result) # doctest: +SKIP
If `Trainable.save_checkpoint` returned a dict, then Tune will directly pass
the dict data as the argument to this method.
Example:
>>> from ray.tune.trainable import Trainable
>>> class Example(Trainable):
... def save_checkpoint(self, checkpoint_path):
... return {"my_data": 1}
... def load_checkpoint(self, checkpoint_dict):
... print(checkpoint_dict["my_data"])
.. versionadded:: 0.8.7
Args:
checkpoint: If dict, the return value is as
returned by ``save_checkpoint``. Otherwise, the directory
the checkpoint was stored in.
"""
raise NotImplementedError
[docs] def setup(self, config: Dict):
"""Subclasses should override this for custom initialization.
.. versionadded:: 0.8.7
Args:
config: Hyperparameters and other configs given.
Copy of `self.config`.
"""
pass
[docs] def log_result(self, result: Dict):
"""Subclasses can optionally override this to customize logging.
The logging here is done on the worker process rather than
the driver.
.. versionadded:: 0.8.7
Args:
result: Training result returned by step().
"""
self._result_logger.on_result(result)
[docs] def cleanup(self):
"""Subclasses should override this for any cleanup on stop.
If any Ray actors are launched in the Trainable (i.e., with a RLlib
trainer), be sure to kill the Ray actor process here.
This process should be lightweight. Per default,
You can kill a Ray actor by calling `ray.kill(actor)`
on the actor or removing all references to it and waiting for garbage
collection
.. versionadded:: 0.8.7
"""
pass
def _export_model(self, export_formats: List[str], export_dir: str):
"""Subclasses should override this to export model.
Args:
export_formats: List of formats that should be exported.
export_dir: Directory to place exported models.
Return:
A dict that maps ExportFormats to successfully exported models.
"""
return {}
def _implements_method(self, key):
return hasattr(self, key) and callable(getattr(self, key))