def testMaxConcurrentSuggestions(self):
"""Checks that next_trials() supports throttling."""
experiment_spec = {
"run": "PPO",
"num_samples": 6,
}
experiments = [Experiment.from_json("test", experiment_spec)]
searcher = _MockSuggestionAlgorithm(max_concurrent=4)
searcher.add_configurations(experiments)
trials = searcher.next_trials()
self.assertEqual(len(trials), 4)
self.assertEqual(searcher.next_trials(), [])
finished_trial = trials.pop()
searcher.on_trial_complete(finished_trial.trial_id)
self.assertEqual(len(searcher.next_trials()), 1)
finished_trial = trials.pop()
searcher.on_trial_complete(finished_trial.trial_id)
finished_trial = trials.pop()
searcher.on_trial_complete(finished_trial.trial_id)
finished_trial = trials.pop()
searcher.on_trial_complete(finished_trial.trial_id)
self.assertEqual(len(searcher.next_trials()), 1)
self.assertEqual(len(searcher.next_trials()), 0)
def testSearchAlgSchedulerEarlyStop(self):
"""Early termination notif to Searcher can be turned off."""
class _MockScheduler(FIFOScheduler):
def on_trial_result(self, *args, **kwargs):
return TrialScheduler.STOP
ray.init(num_cpus=4, num_gpus=2)
experiment_spec = {"run": "__fake", "stop": {"training_iteration": 2}}
experiments = [Experiment.from_json("test", experiment_spec)]
searcher = _MockSuggestionAlgorithm(use_early_stopped_trials=True)
searcher.add_configurations(experiments)
runner = TrialRunner(search_alg=searcher, scheduler=_MockScheduler())
runner.step()
runner.step()
self.assertEqual(len(searcher.final_results), 1)
searcher = _MockSuggestionAlgorithm(use_early_stopped_trials=False)
searcher.add_configurations(experiments)
runner = TrialRunner(search_alg=searcher, scheduler=_MockScheduler())
runner.step()
runner.step()
self.assertEqual(len(searcher.final_results), 0)
def testSearchAlgStalled(self):
"""Checks that runner and searcher state is maintained when stalled."""
ray.init(num_cpus=4, num_gpus=2)
experiment_spec = {
"run": "__fake",
"num_samples": 3,
"stop": {
"training_iteration": 1
}
}
experiments = [Experiment.from_json("test", experiment_spec)]
search_alg = _MockSuggestionAlgorithm(max_concurrent=1)
search_alg.add_configurations(experiments)
searcher = search_alg.searcher
runner = TrialRunner(search_alg=search_alg)
runner.step()
trials = runner.get_trials()
self.assertEqual(trials[0].status, Trial.RUNNING)
runner.step()
self.assertEqual(trials[0].status, Trial.TERMINATED)
trials = runner.get_trials()
runner.step()
self.assertEqual(trials[1].status, Trial.RUNNING)
self.assertEqual(len(searcher.live_trials), 1)
searcher.stall = True
runner.step()
self.assertEqual(trials[1].status, Trial.TERMINATED)
self.assertEqual(len(searcher.live_trials), 0)
self.assertTrue(all(trial.is_finished() for trial in trials))
self.assertFalse(search_alg.is_finished())
self.assertFalse(runner.is_finished())
searcher.stall = False
runner.step()
trials = runner.get_trials()
self.assertEqual(trials[2].status, Trial.RUNNING)
self.assertEqual(len(searcher.live_trials), 1)
runner.step()
self.assertEqual(trials[2].status, Trial.TERMINATED)
self.assertEqual(len(searcher.live_trials), 0)
self.assertTrue(search_alg.is_finished())
self.assertTrue(runner.is_finished())
def testNestedResults(self):
def create_result(i):
return {"test": {"1": {"2": {"3": i, "4": False}}}}
flattened_keys = list(flatten_dict(create_result(0)))
class _MockScheduler(FIFOScheduler):
results = []
def on_trial_result(self, trial_runner, trial, result):
self.results += [result]
return TrialScheduler.CONTINUE
def on_trial_complete(self, trial_runner, trial, result):
self.complete_result = result
def train(config, reporter):
for i in range(100):
reporter(**create_result(i))
algo = _MockSuggestionAlgorithm()
scheduler = _MockScheduler()
[trial] = tune.run(
train,
scheduler=scheduler,
search_alg=algo,
stop={
"test/1/2/3": 20
}).trials
self.assertEqual(trial.status, Trial.TERMINATED)
self.assertEqual(trial.last_result["test"]["1"]["2"]["3"], 20)
self.assertEqual(trial.last_result["test"]["1"]["2"]["4"], False)
self.assertEqual(trial.last_result[TRAINING_ITERATION], 21)
self.assertEqual(len(scheduler.results), 20)
self.assertTrue(
all(
set(result) >= set(flattened_keys)
for result in scheduler.results))
self.assertTrue(set(scheduler.complete_result) >= set(flattened_keys))
self.assertEqual(len(algo.results), 20)
self.assertTrue(
all(set(result) >= set(flattened_keys) for result in algo.results))
with self.assertRaises(TuneError):
[trial] = tune.run(train, stop={"1/2/3": 20})
with self.assertRaises(TuneError):
[trial] = tune.run(train, stop={"test": 1}).trials
def testSearchAlgFinished(self):
"""Checks that SearchAlg is Finished before all trials are done."""
ray.init(num_cpus=4, num_gpus=2)
experiment_spec = {"run": "__fake", "stop": {"training_iteration": 1}}
experiments = [Experiment.from_json("test", experiment_spec)]
searcher = _MockSuggestionAlgorithm(experiments, max_concurrent=10)
runner = TrialRunner(search_alg=searcher)
runner.step()
trials = runner.get_trials()
self.assertEqual(trials[0].status, Trial.RUNNING)
self.assertTrue(searcher.is_finished())
self.assertFalse(runner.is_finished())
runner.step()
self.assertEqual(trials[0].status, Trial.TERMINATED)
self.assertEqual(len(searcher.live_trials), 0)
self.assertTrue(searcher.is_finished())
self.assertTrue(runner.is_finished())
def testSearchAlgNotification(self):
"""Checks notification of trial to the Search Algorithm."""
ray.init(num_cpus=4, num_gpus=2)
experiment_spec = {"run": "__fake", "stop": {"training_iteration": 2}}
experiments = [Experiment.from_json("test", experiment_spec)]
searcher = _MockSuggestionAlgorithm(experiments, max_concurrent=10)
runner = TrialRunner(search_alg=searcher)
runner.step()
trials = runner.get_trials()
self.assertEqual(trials[0].status, Trial.RUNNING)
runner.step()
self.assertEqual(trials[0].status, Trial.RUNNING)
runner.step()
self.assertEqual(trials[0].status, Trial.TERMINATED)
self.assertEqual(searcher.counter["result"], 1)
self.assertEqual(searcher.counter["complete"], 1)
def testSearchAlgSchedulerInteraction(self):
"""Checks that TrialScheduler killing trial will notify SearchAlg."""
class _MockScheduler(FIFOScheduler):
def on_trial_result(self, *args, **kwargs):
return TrialScheduler.STOP
ray.init(num_cpus=4, num_gpus=2)
experiment_spec = {"run": "__fake", "stop": {"training_iteration": 2}}
experiments = [Experiment.from_json("test", experiment_spec)]
searcher = _MockSuggestionAlgorithm(experiments, max_concurrent=10)
runner = TrialRunner(search_alg=searcher, scheduler=_MockScheduler())
runner.step()
trials = runner.get_trials()
self.assertEqual(trials[0].status, Trial.RUNNING)
self.assertTrue(searcher.is_finished())
self.assertFalse(runner.is_finished())
runner.step()
self.assertEqual(trials[0].status, Trial.TERMINATED)
self.assertEqual(len(searcher.live_trials), 0)
self.assertTrue(searcher.is_finished())
self.assertTrue(runner.is_finished())