Files
ray/python/ray/tune/stopper.py
T
Kai Fricke 5f04ade6ef [tune] add more stoppers and stopper documentation (#12750)
* Add new stoppers & docs

* Add tests for maximum iteration stopper and trial plateau stopper

* Update python/ray/tune/stopper.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Update doc/source/tune/api_docs/stoppers.rst

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Update doc/source/tune/api_docs/stoppers.rst

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>

* Apply suggestions from code review

* Apply suggestions from code review

* Update python/ray/tune/stopper.py

Co-authored-by: Richard Liaw <rliaw@berkeley.edu>
2020-12-12 01:47:19 -08:00

351 lines
12 KiB
Python

import warnings
from typing import Dict, Optional
import time
from collections import defaultdict, deque
import numpy as np
from ray import logger
class Stopper:
"""Base class for implementing a Tune experiment stopper.
Allows users to implement experiment-level stopping via ``stop_all``. By
default, this class does not stop any trials. Subclasses need to
implement ``__call__`` and ``stop_all``.
.. code-block:: python
import time
from ray import tune
from ray.tune import Stopper
class TimeStopper(Stopper):
def __init__(self):
self._start = time.time()
self._deadline = 300
def __call__(self, trial_id, result):
return False
def stop_all(self):
return time.time() - self._start > self.deadline
tune.run(Trainable, num_samples=200, stop=TimeStopper())
"""
def __call__(self, trial_id, result):
"""Returns true if the trial should be terminated given the result."""
raise NotImplementedError
def stop_all(self):
"""Returns true if the experiment should be terminated."""
raise NotImplementedError
class CombinedStopper(Stopper):
"""Combine several stoppers via 'OR'.
Args:
*stoppers (Stopper): Stoppers to be combined.
Example:
.. code-block:: python
from ray.tune.stopper import CombinedStopper, \
MaximumIterationStopper, TrialPlateauStopper
stopper = CombinedStopper(
MaximumIterationStopper(max_iter=20),
TrialPlateauStopper(metric="my_metric")
)
tune.run(train, stop=stopper)
"""
def __init__(self, *stoppers: Stopper):
self._stoppers = stoppers
def __call__(self, trial_id, result):
return any(s(trial_id, result) for s in self._stoppers)
def stop_all(self):
return any(s.stop_all() for s in self._stoppers)
class NoopStopper(Stopper):
def __call__(self, trial_id, result):
return False
def stop_all(self):
return False
class FunctionStopper(Stopper):
"""Provide a custom function to check if trial should be stopped.
The passed function will be called after each iteration. If it returns
True, the trial will be stopped.
Args:
function (Callable[[str, Dict], bool): Function that checks if a trial
should be stopped. Must accept the `trial_id` string and `result`
dictionary as arguments. Must return a boolean.
"""
def __init__(self, function):
self._fn = function
def __call__(self, trial_id, result):
return self._fn(trial_id, result)
def stop_all(self):
return False
@classmethod
def is_valid_function(cls, fn):
is_function = callable(fn) and not issubclass(type(fn), Stopper)
if is_function and hasattr(fn, "stop_all"):
raise ValueError(
"Stop object must be ray.tune.Stopper subclass to be detected "
"correctly.")
return is_function
class MaximumIterationStopper(Stopper):
"""Stop trials after reaching a maximum number of iterations
Args:
max_iter (int): Number of iterations before stopping a trial.
"""
def __init__(self, max_iter: int):
self._max_iter = max_iter
self._iter = defaultdict(lambda: 0)
def __call__(self, trial_id: str, result: Dict):
self._iter[trial_id] += 1
return self._iter[trial_id] >= self._max_iter
def stop_all(self):
return False
class ExperimentPlateauStopper(Stopper):
"""Early stop the experiment when a metric plateaued across trials.
Stops the entire experiment when the metric has plateaued
for more than the given amount of iterations specified in
the patience parameter.
Args:
metric (str): The metric to be monitored.
std (float): The minimal standard deviation after which
the tuning process has to stop.
top (int): The number of best models to consider.
mode (str): The mode to select the top results.
Can either be "min" or "max".
patience (int): Number of epochs to wait for
a change in the top models.
Raises:
ValueError: If the mode parameter is not "min" nor "max".
ValueError: If the top parameter is not an integer
greater than 1.
ValueError: If the standard deviation parameter is not
a strictly positive float.
ValueError: If the patience parameter is not
a strictly positive integer.
"""
def __init__(self, metric, std=0.001, top=10, mode="min", patience=0):
if mode not in ("min", "max"):
raise ValueError("The mode parameter can only be"
" either min or max.")
if not isinstance(top, int) or top <= 1:
raise ValueError("Top results to consider must be"
" a positive integer greater than one.")
if not isinstance(patience, int) or patience < 0:
raise ValueError("Patience must be"
" a strictly positive integer.")
if not isinstance(std, float) or std <= 0:
raise ValueError("The standard deviation must be"
" a strictly positive float number.")
self._mode = mode
self._metric = metric
self._patience = patience
self._iterations = 0
self._std = std
self._top = top
self._top_values = []
def __call__(self, trial_id, result):
"""Return a boolean representing if the tuning has to stop."""
self._top_values.append(result[self._metric])
if self._mode == "min":
self._top_values = sorted(self._top_values)[:self._top]
else:
self._top_values = sorted(self._top_values)[-self._top:]
# If the current iteration has to stop
if self.has_plateaued():
# we increment the total counter of iterations
self._iterations += 1
else:
# otherwise we reset the counter
self._iterations = 0
# and then call the method that re-executes
# the checks, including the iterations.
return self.stop_all()
def has_plateaued(self):
return (len(self._top_values) == self._top
and np.std(self._top_values) <= self._std)
def stop_all(self):
"""Return whether to stop and prevent trials from starting."""
return self.has_plateaued() and self._iterations >= self._patience
class EarlyStopping(ExperimentPlateauStopper):
def __init__(self, *args, **kwargs):
warnings.warn(
"The `EarlyStopping` stopper has been renamed to "
"`ExperimentPlateauStopper`. The reference will be removed "
"in a future version of Ray. Please use ExperimentPlateauStopper"
"instead.", DeprecationWarning)
super(EarlyStopping, self).__init__(*args, **kwargs)
class TrialPlateauStopper(Stopper):
"""Early stop single trials when they reached a plateau.
When the standard deviation of the `metric` result of a trial is
below a threshold `std`, the trial plateaued and will be stopped
early.
Args:
metric (str): Metric to check for convergence.
std (float): Maximum metric standard deviation to decide if a
trial plateaued. Defaults to 0.01.
num_results (int): Number of results to consider for stdev
calculation.
grace_period (int): Minimum number of timesteps before a trial
can be early stopped
metric_threshold (Optional[float]):
Minimum or maximum value the result has to exceed before it can
be stopped early.
mode (Optional[str]): If a `metric_threshold` argument has been
passed, this must be one of [min, max]. Specifies if we optimize
for a large metric (max) or a small metric (min). If max, the
`metric_threshold` has to be exceeded, if min the value has to
be lower than `metric_threshold` in order to early stop.
"""
def __init__(self,
metric: str,
std: float = 0.01,
num_results: int = 4,
grace_period: int = 4,
metric_threshold: Optional[float] = None,
mode: Optional[str] = None):
self._metric = metric
self._mode = mode
self._std = std
self._num_results = num_results
self._grace_period = grace_period
self._metric_threshold = metric_threshold
if self._metric_threshold:
if mode not in ["min", "max"]:
raise ValueError(
f"When specifying a `metric_threshold`, the `mode` "
f"argument has to be one of [min, max]. "
f"Got: {mode}")
self._iter = defaultdict(lambda: 0)
self._trial_results = defaultdict(
lambda: deque(maxlen=self._num_results))
def __call__(self, trial_id: str, result: Dict):
metric_result = result.get(self._metric)
self._trial_results[trial_id].append(metric_result)
self._iter[trial_id] += 1
# If still in grace period, do not stop yet
if self._iter[trial_id] < self._grace_period:
return False
# If not enough results yet, do not stop yet
if len(self._trial_results[trial_id]) < self._num_results:
return False
# If metric threshold value not reached, do not stop yet
if self._metric_threshold is not None:
if self._mode == "min" and metric_result > self._metric_threshold:
return False
elif self._mode == "max" and \
metric_result < self._metric_threshold:
return False
# Calculate stdev of last `num_results` results
try:
current_std = np.std(self._trial_results[trial_id])
except Exception:
current_std = float("inf")
# If stdev is lower than threshold, stop early.
return current_std < self._std
def stop_all(self):
return False
class TimeoutStopper(Stopper):
"""Stops all trials after a certain timeout.
This stopper is automatically created when the `time_budget_s`
argument is passed to `tune.run()`.
Args:
timeout (int|float|datetime.timedelta): Either a number specifying
the timeout in seconds, or a `datetime.timedelta` object.
"""
def __init__(self, timeout):
from datetime import timedelta
if isinstance(timeout, timedelta):
self._timeout_seconds = timeout.total_seconds()
elif isinstance(timeout, (int, float)):
self._timeout_seconds = timeout
else:
raise ValueError(
"`timeout` parameter has to be either a number or a "
"`datetime.timedelta` object. Found: {}".format(type(timeout)))
# To account for setup overhead, set the start time only after
# the first call to `stop_all()`.
self._start = None
def __call__(self, trial_id, result):
return False
def stop_all(self):
if not self._start:
self._start = time.time()
return False
now = time.time()
if now - self._start >= self._timeout_seconds:
logger.info(f"Reached timeout of {self._timeout_seconds} seconds. "
f"Stopping all trials.")
return True
return False