# -*- coding: utf-8 -*-
# pylint:disable=protected-access,too-many-public-methods,too-many-lines
"""
:mod:`orion.core.worker.experiment` -- Description of an optimization attempt
=============================================================================
.. module:: experiment
:platform: Unix
:synopsis: Manage history of trials corresponding to a black box process
"""
import copy
import datetime
import getpass
import logging
import sys
import orion.core
from orion.core.cli.evc import fetch_branching_configuration
from orion.core.evc.adapters import Adapter, BaseAdapter
from orion.core.evc.conflicts import detect_conflicts, ExperimentNameConflict
from orion.core.io.database import DuplicateKeyError
from orion.core.io.experiment_branch_builder import ExperimentBranchBuilder
from orion.core.io.interactive_commands.branching_prompt import BranchingPrompt
from orion.core.io.space_builder import SpaceBuilder
import orion.core.utils.backward as backward
from orion.core.utils.exceptions import RaceCondition
from orion.core.worker.primary_algo import PrimaryAlgo
from orion.core.worker.strategy import (BaseParallelStrategy,
Strategy)
from orion.storage.base import get_storage, ReadOnlyStorageProtocol
log = logging.getLogger(__name__)
# pylint: disable=too-many-public-methods
[docs]class Experiment:
"""Represents an entry in database/experiments collection.
Attributes
----------
name : str
Unique identifier for this experiment per `user`.
id: object
id of the experiment in the database if experiment is configured. Value is `None`
if the experiment is not configured.
refers : dict or list of `Experiment` objects, after initialization is done.
A dictionary pointing to a past `Experiment` id, ``refers[parent_id]``, whose
trials we want to add in the history of completed trials we want to re-use.
For convenience and database effiency purpose, all experiments of a common tree shares
`refers[root_id]`, with the root experiment refering to itself.
version: int
Current version of this experiment.
metadata : dict
Contains managerial information about this `Experiment`.
pool_size : int
How many workers can participate asynchronously in this `Experiment`.
max_trials : int
How many trials must be evaluated, before considering this `Experiment` done.
This attribute can be updated if the rest of the experiment configuration
is the same. In that case, if trying to set to an already set experiment,
it will overwrite the previous one.
algorithms : dict of dicts or an `PrimaryAlgo` object, after initialization is done.
Complete specification of the optimization and dynamical procedures taking
place in this `Experiment`.
Metadata
--------
user : str
System user currently owning this running process, the one who invoked **Oríon**.
datetime : `datetime.datetime`
When was this particular configuration submitted to the database.
orion_version : str
Version of **Oríon** which suggested this experiment. `user`'s current
**Oríon** version.
user_script : str
Full absolute path to `user`'s executable.
user_args : list of str
Contains separate arguments to be passed when invoking `user_script`,
possibly templated for **Oríon**.
user_vcs : str, optional
User's version control system for this executable's code repository.
user_version : str, optional
Current user's repository version.
user_commit_hash : str, optional
Current `Experiment`'s commit hash for **Oríon**'s invocation.
"""
__slots__ = ('name', 'refers', 'metadata', 'pool_size', 'max_trials', 'version',
'algorithms', 'producer', 'working_dir', '_init_done', '_id',
'_node', '_storage')
non_branching_attrs = ('pool_size', 'max_trials')
def __init__(self, name, user=None, version=None):
"""Initialize an Experiment object with primary key (:attr:`name`, :attr:`user`).
Try to find an entry in `Database` with such a key and config this object
from it import, if successful. Else, init with default/empty values and
insert new entry with this object's attributes in database.
.. note::
Practically initialization has not finished until `config`'s setter
is called.
:param name: Describe a configuration with a unique identifier per :attr:`user`.
:type name: str
"""
log.debug("Creating Experiment object with name: %s", name)
self._init_done = False
self._id = None
self.name = name
self._node = None
self.refers = {}
if user is None:
user = getpass.getuser()
self.metadata = {'user': user}
self.pool_size = None
self.max_trials = None
self.algorithms = None
self.working_dir = None
self.producer = {'strategy': None}
self.version = 1
# this needs to be an attribute because we override it in ExperienceView
self._storage = get_storage()
config = self._storage.fetch_experiments({'name': name})
if config:
log.debug("Found existing experiment, %s, under user, %s, registered in database.",
name, user)
if len(config) > 1:
max_version = max(config, key=lambda exp: exp.get('version', 1)).get('version', 1)
if version is None:
self.version = max_version
else:
self.version = version
if self.version > max_version:
log.warning("Version %s was specified but most recent version is only %s. "
"Using %s.", self.version, max_version, max_version)
self.version = min(self.version, max_version)
log.info("Many versions for experiment %s have been found. Using latest "
"version %s.", name, self.version)
config = filter(lambda exp: exp.get('version', 1) == self.version, config)
config = sorted(config, key=lambda x: x['metadata']['datetime'],
reverse=True)[0]
backward.populate_priors(config['metadata'])
for attrname in self.__slots__:
if not attrname.startswith('_') and attrname in config:
setattr(self, attrname, config[attrname])
self._id = config['_id']
[docs] def fetch_trials(self, with_evc_tree=False):
"""Fetch all trials of the experiment"""
return self._select_evc_call(with_evc_tree, 'fetch_trials')
[docs] def get_trial(self, trial=None, uid=None):
"""Fetch a single Trial, see `orion.storage.base.BaseStorage.get_trial`"""
return self._storage.get_trial(trial, uid)
[docs] def connect_to_version_control_tree(self, node):
"""Connect the experiment to its node in a version control tree
.. seealso::
:class:`orion.core.evc.experiment.ExperimentNode`
:param node: Node giving access to the experiment version control tree.
:type name: None or `ExperimentNode`
"""
self._node = node
[docs] def retrieve_result(self, trial, *args, **kwargs):
"""See :func:`~orion.storage.BaseStorageProtocol.retrieve_result`"""
return self._storage.retrieve_result(trial, *args, **kwargs)
[docs] def set_trial_status(self, *args, **kwargs):
"""See :func:`~orion.storage.BaseStorageProtocol.set_trial_status`"""
return self._storage.set_trial_status(*args, **kwargs)
[docs] def reserve_trial(self, score_handle=None):
"""Find *new* trials that exist currently in database and select one of
them based on the highest score return from `score_handle` callable.
Parameters
----------
score_handle: callable object, optional
A way to decide which trial out of the *new* ones to
to pick as *reserved*, defaults to a random choice.
Deprecated
Returns
-------
Selected `Trial` object, None if could not find any.
"""
log.debug('reserving trial with (score: %s)', score_handle)
if score_handle is not None:
log.warning("Argument `score_handle` is deprecated")
self.fix_lost_trials()
selected_trial = self._storage.reserve_trial(self)
log.debug('reserved trial (trial: %s)', selected_trial)
return selected_trial
[docs] def fix_lost_trials(self):
"""Find lost trials and set them to interrupted.
A lost trial is defined as a trial whose heartbeat as not been updated since two times
the wait time for monitoring. This usually means that the trial is stalling or has been
interrupted in some way without its status being changed. This functions finds such
trials and set them as interrupted so they can be launched again.
"""
trials = self._storage.fetch_lost_trials(self)
for trial in trials:
log.debug('Setting lost trial %s status to interrupted...', trial.id)
updated = self._storage.set_trial_status(trial, status='interrupted')
log.debug('success' if updated else 'failed')
[docs] def update_completed_trial(self, trial, results_file):
"""Inform database about an evaluated `trial` with results.
:param trial: Corresponds to a successful evaluation of a particular run.
:type trial: `Trial`
.. note::
Change status from *reserved* to *completed*.
"""
trial.status = 'completed'
trial.end_time = datetime.datetime.utcnow()
self._storage.retrieve_result(trial, results_file)
# push trial results updates the entire trial status included
self._storage.push_trial_results(trial)
[docs] def register_lie(self, lying_trial):
"""Register a *fake* trial created by the strategist.
The main difference between fake trial and orignal ones is the addition of a fake objective
result, and status being set to completed. The id of the fake trial is different than the id
of the original trial, but the original id can be computed using the hashcode on parameters
of the fake trial. See mod:`orion.core.worker.strategy` for more information and the
Strategist object and generation of fake trials.
Parameters
----------
trials: `Trial` object
Fake trial to register in the database
Raises
------
orion.core.io.database.DuplicateKeyError
If a trial with the same id already exist in the database. Since the id is computed
based on a hashing of the trial, this should mean that an identical trial already exist
in the database.
"""
lying_trial.status = 'completed'
lying_trial.end_time = datetime.datetime.utcnow()
self._storage.register_lie(lying_trial)
[docs] def register_trial(self, trial):
"""Register new trial in the database.
Inform database about *new* suggested trial with specific parameter values. Trials may only
be registered one at a time to avoid registration of duplicates.
Parameters
----------
trials: `Trial` object
Trial to register in the database
Raises
------
orion.core.io.database.DuplicateKeyError
If a trial with the same id already exist in the database. Since the id is computed
based on a hashing of the trial, this should mean that an identical trial already exist
in the database.
"""
stamp = datetime.datetime.utcnow()
trial.experiment = self._id
trial.status = 'new'
trial.submit_time = stamp
self._storage.register_trial(trial)
def _select_evc_call(self, with_evc_tree, function, *args, **kwargs):
if self._node is not None and with_evc_tree:
return getattr(self._node, function)(*args, **kwargs)
return getattr(self._storage, function)(self, *args, **kwargs)
[docs] def fetch_trials_by_status(self, status, with_evc_tree=False):
"""Fetch all trials with the given status
Trials are sorted based on `Trial.submit_time`
:return: list of `Trial` objects
"""
return self._select_evc_call(with_evc_tree, 'fetch_trial_by_status', status)
[docs] def fetch_noncompleted_trials(self, with_evc_tree=False):
"""Fetch non-completed trials of this `Experiment` instance.
Trials are sorted based on `Trial.submit_time`
.. note::
It will return all non-completed trials, including new, reserved, suspended,
interrupted and broken ones.
:return: list of non-completed `Trial` objects
"""
return self._select_evc_call(with_evc_tree, 'fetch_noncompleted_trials')
# pylint: disable=invalid-name
@property
def id(self):
"""Id of the experiment in the database if configured.
Value is `None` if the experiment is not configured.
"""
return self._id
@property
def node(self):
"""Node of the experiment in the version control tree.
Value is `None` if the experiment is not connected to the version control tree.
.. seealso::
:py:meth:`orion.core.worker.experiment.Experiment.connect_to_version_control_tree`
"""
return self._node
@property
def is_done(self):
"""Return True, if this experiment is considered to be finished.
1. Count how many trials have been completed and compare with `max_trials`.
2. Ask `algorithms` if they consider there is a chance for further improvement.
.. note::
To be used as a terminating condition in a ``Worker``.
"""
num_completed_trials = self._storage.count_completed_trials(self)
return ((num_completed_trials >= self.max_trials) or
(self._init_done and self.algorithms.is_done))
@property
def is_broken(self):
"""Return True, if this experiment is considered to be broken.
Count how many trials are broken and return True if that number has reached
as given threshold.
"""
num_broken_trials = self._storage.count_broken_trials(self)
return num_broken_trials >= orion.core.config.worker.max_broken
@property
def space(self):
"""Return problem's parameter `orion.algo.space.Space`.
.. note:: It will return None, if experiment init is not done.
"""
if self._init_done:
return self.algorithms.space
return None
@property
def configuration(self):
"""Return a copy of an `Experiment` configuration as a dictionary."""
config = dict()
for attrname in self.__slots__:
if attrname.startswith('_'):
continue
attribute = getattr(self, attrname)
if self._init_done and attrname == 'algorithms':
config[attrname] = attribute.configuration
else:
config[attrname] = attribute
if attrname == "refers" and isinstance(attribute.get("adapter"), BaseAdapter):
config[attrname] = copy.deepcopy(config[attrname])
config[attrname]['adapter'] = config[attrname]['adapter'].configuration
if self._init_done and attrname == "producer" and attribute.get("strategy"):
config[attrname] = copy.deepcopy(config[attrname])
config[attrname]['strategy'] = config[attrname]['strategy'].configuration
# Reason for deepcopy is that some attributes are dictionaries
# themselves, we don't want to accidentally change the state of this
# object from a getter.
return copy.deepcopy(config)
@property
def stats(self):
"""Calculate a stats dictionary for this particular experiment.
Returns
-------
stats : dict
Stats
-----
trials_completed : int
Number of completed trials
best_trials_id : int
Unique identifier of the `Trial` object in the database which achieved
the best known objective result.
best_evaluation : float
Evaluation score of the best trial
start_time : `datetime.datetime`
When Experiment was first dispatched and started running.
finish_time : `datetime.datetime`
When Experiment reached terminating condition and stopped running.
duration : `datetime.timedelta`
Elapsed time.
"""
completed_trials = self.fetch_trials_by_status('completed')
if not completed_trials:
return dict()
stats = dict()
stats['trials_completed'] = len(completed_trials)
stats['best_trials_id'] = None
trial = completed_trials[0]
stats['best_evaluation'] = trial.objective.value
stats['best_trials_id'] = trial.id
stats['start_time'] = self.metadata['datetime']
stats['finish_time'] = stats['start_time']
for trial in completed_trials:
# All trials are going to finish certainly after the start date
# of the experiment they belong to
if trial.end_time > stats['finish_time']: # pylint:disable=no-member
stats['finish_time'] = trial.end_time
objective = trial.objective.value
if objective < stats['best_evaluation']:
stats['best_evaluation'] = objective
stats['best_trials_id'] = trial.id
stats['duration'] = stats['finish_time'] - stats['start_time']
return stats
def _instantiate_config(self, config):
"""Check before dispatching experiment whether configuration corresponds
to a executable experiment environment.
1. Check `refers` and instantiate `Adapter` objects from it.
2. Try to build parameter space from user arguments.
3. Check whether configured algorithms correspond to [known]/valid
implementations of the ``Algorithm`` class. Instantiate these objects.
4. Check if experiment `is_done`, prompt for larger `max_trials` if it is. (TODO)
"""
# Just overwrite everything else given
for section, value in config.items():
if section not in self.__slots__:
log.info("Found section '%s' in configuration. Experiments "
"do not support this option. Ignoring.", section)
continue
if section.startswith('_'):
log.info("Found section '%s' in configuration. "
"Cannot set private attributes. Ignoring.", section)
continue
# Copy sub configuration to value confusing side-effects
# Only copy at this level, not `config` directly to avoid TypeErrors if config contains
# non-serializable objects (copy.deepcopy complains otherwise).
if isinstance(value, dict):
value = copy.deepcopy(value)
setattr(self, section, value)
# TODO: Can we get rid of this try-except clause?
try:
space_builder = SpaceBuilder()
space = space_builder.build(config['metadata']['priors'])
if not space:
raise ValueError("Parameter space is empty. There is nothing to optimize.")
# Instantiate algorithms
self.algorithms = PrimaryAlgo(space, self.algorithms)
except KeyError:
pass
self.refers.setdefault('parent_id', None)
self.refers.setdefault('root_id', self._id)
self.refers.setdefault('adapter', [])
if not isinstance(self.refers.get('adapter'), BaseAdapter):
self.refers['adapter'] = Adapter.build(self.refers['adapter'])
if not self.producer.get('strategy'):
self.producer = {'strategy': Strategy(of_type="MaxParallelStrategy")}
elif not isinstance(self.producer.get('strategy'), BaseParallelStrategy):
self.producer = {'strategy': Strategy(of_type=self.producer['strategy'])}
def _branch_config(self, conflicts, branching_configuration):
"""Ask for a different identifier for this experiment. Set :attr:`refers`
key to previous experiment's name, the one that we branched from.
:param config: Conflicting configuration that will change based on prompt.
"""
experiment_brancher = ExperimentBranchBuilder(conflicts, branching_configuration)
needs_manual_resolution = (not experiment_brancher.is_resolved or
experiment_brancher.manual_resolution)
if needs_manual_resolution:
branching_prompt = BranchingPrompt(experiment_brancher)
if not sys.__stdin__.isatty():
raise ValueError(
"Configuration is different and generates a branching event:\n{}".format(
branching_prompt.get_status()))
branching_prompt.cmdloop()
if branching_prompt.abort or not experiment_brancher.is_resolved:
sys.exit()
adapter = experiment_brancher.create_adapters()
self._instantiate_config(experiment_brancher.conflicting_config)
self.refers['adapter'] = adapter
self.refers['parent_id'] = self._id
def _is_different_from(self, config):
"""Return True, if current `Experiment`'s configuration as described by
its attributes is different from the one suggested in `config`.
"""
is_diff = False
for section, value in config.items():
if section in self.non_branching_attrs or \
section not in self.__slots__ or \
section.startswith('_'):
continue
item = getattr(self, section)
if item != value:
log.warning("Config given is different from config found in db at section: %s",
section)
log.warning("Config+ :\n%s", value)
log.warning("Config- :\n%s", item)
is_diff = True
break
return is_diff
def __repr__(self):
"""Represent the object as a string."""
return "Experiment(name=%s, metadata.user=%s, version=%s)" % \
(self.name, self.metadata['user'], self.version)
# pylint: disable=too-few-public-methods
[docs]class ExperimentView(object):
"""Non-writable view of an experiment
.. seealso::
:py:class:`orion.core.worker.experiment.Experiment` for writable experiments.
"""
__slots__ = ('_experiment', )
# Attributes
valid_attributes = (["_id", "name", "refers", "metadata", "pool_size", "max_trials",
"version"] +
# Properties
["id", "node", "is_done", "space", "algorithms", "stats", "configuration"] +
# Methods
["fetch_trials", "fetch_trials_by_status",
"connect_to_version_control_tree", "get_trial"])
def __init__(self, name, user=None, version=None):
"""Initialize viewed experiment object with primary key (:attr:`name`, :attr:`user`).
Build an experiment from configuration found in `Database` with a key (name, user).
.. note::
A view is fully configured at initialiation. It cannot be reconfigured.
If no experiment is found for the key (name, user), a `ValueError` will be raised.
:param name: Describe a configuration with a unique identifier per :attr:`user`.
:type name: str
"""
self._experiment = Experiment(name, user, version)
if self._experiment.id is None:
raise ValueError("No experiment with given name '%s' for user '%s' inside database, "
"no view can be created." %
(self._experiment.name, self._experiment.metadata['user']))
# TODO: Views are not fully configured until configuration is refactored
# This snippet is to instantiate adapters anyhow, because it is required for
# experiment views in EVC.
self.refers.setdefault('parent_id', None)
self.refers.setdefault('root_id', self._id)
self.refers.setdefault('adapter', [])
if not isinstance(self.refers.get('adapter'), BaseAdapter):
self.refers['adapter'] = Adapter.build(self.refers['adapter'])
# try:
# self._experiment.configure(self._experiment.configuration, enable_branching=False,
# enable_update=False)
# except ValueError as e:
# if "Configuration is different and generates a branching event" in str(e):
# raise RuntimeError(
# "Configuration in the database does not correspond to the one generated by "
# "Experiment object. This is likely due to a backward incompatible update in "
# "Oríon. Please report to https://github.com/epistimio/orion/issues.") from e
# raise
self._experiment._storage = ReadOnlyStorageProtocol(get_storage())
def __getattr__(self, name):
"""Get attribute only if valid"""
if name not in self.valid_attributes:
raise AttributeError("Cannot access attribute %s on view-only experiments." % name)
return getattr(self._experiment, name)
def __repr__(self):
"""Represent the object as a string."""
return "ExperimentView(name=%s, metadata.user=%s, version=%s)" % \
(self.name, self.metadata['user'], self.version)