def test_suggest_promote_identic_objectives(
self,
asha: ASHA,
bracket: ASHABracket,
big_rung_0: RungDict,
big_rung_1: RungDict,
):
"""Test that identic objectives are handled properly"""
asha.brackets = [bracket]
bracket.owner = asha
n_trials = 9
resources = 1
results = {}
for param in np.linspace(0, 8, 9):
trial = create_trial_for_hb((resources, param), objective=0)
trial_hash = trial.compute_trial_hash(
trial,
ignore_fidelity=True,
ignore_experiment=True,
)
assert trial.objective is not None
results[trial_hash] = (trial.objective.value, trial)
bracket.rungs[0] = RungDict(n_trials=n_trials,
resources=resources,
results=results)
candidates = asha.suggest(2)
assert len(candidates) == 2
assert (sum(1 for trial in candidates
if trial.params[asha.fidelity_index] == 3) == 2)
def test_suggest_promote(self, asha: ASHA, bracket: ASHABracket,
rung_0: RungDict):
"""Test that correct point is promoted and returned."""
asha.brackets = [bracket]
assert bracket.owner is asha
bracket.rungs[0] = rung_0
trials = asha.suggest(1)
assert trials[0].params == {"epoch": 3, "lr": 0.0}
def test_register(self, asha: ASHA, bracket: ASHABracket, rung_0: RungDict,
rung_1: RungDict):
"""Check that a point is registered inside the bracket."""
asha.brackets = [bracket]
assert bracket.owner is asha
bracket.rungs = [rung_0, rung_1]
trial = create_trial_for_hb((1, 0.0), 0.0)
trial_id = asha.get_id(trial, ignore_fidelity=True)
asha.observe([trial])
assert len(bracket.rungs[0])
assert trial_id in bracket.rungs[0]["results"]
assert bracket.rungs[0]["results"][trial_id][0] == 0.0
assert bracket.rungs[0]["results"][trial_id][1].params == trial.params
def test_suggest_promote_many(
self,
asha: ASHA,
bracket: ASHABracket,
big_rung_0: RungDict,
big_rung_1: RungDict,
):
"""Test that correct points are promoted and returned."""
asha.brackets = [bracket]
assert bracket.owner is asha
bracket.rungs[0] = big_rung_0
bracket.rungs[1] = big_rung_1
candidates = asha.suggest(3)
assert len(candidates) == 2 + 1
assert (sum(1 for trial in candidates
if trial.params[asha.fidelity_index] == 9) == 2)
assert (sum(1 for trial in candidates
if trial.params[asha.fidelity_index] == 3) == 1)
def test_suggest_new(
self,
monkeypatch,
asha: ASHA,
bracket: ASHABracket,
rung_0: RungDict,
rung_1: RungDict,
rung_2: RungDict,
):
"""Test that a new point is sampled."""
asha.brackets = [bracket]
assert bracket.owner is asha
def sample(num=1, seed=None):
return [create_trial_for_hb(("fidelity", 0.5))]
monkeypatch.setattr(asha.space, "sample", sample)
trials = asha.suggest(1)
assert trials[0].params == {"epoch": 1, "lr": 0.5}
def test_suggest_inf_duplicates(
self,
monkeypatch,
asha: ASHA,
bracket: ASHABracket,
rung_0: RungDict,
rung_1: RungDict,
rung_2: RungDict,
):
"""Test that sampling inf collisions returns None."""
asha.brackets = [bracket]
assert bracket.owner is asha
fidelity = 1
zhe_trial = create_trial_for_hb((fidelity, 0.0))
asha.trial_to_brackets[asha.get_id(zhe_trial,
ignore_fidelity=True)] = 0
def sample(num=1, seed=None):
return [zhe_trial]
monkeypatch.setattr(asha.space, "sample", sample)
assert asha.suggest(1) == []
def test_suggest_duplicates(
self,
monkeypatch,
asha: ASHA,
bracket: ASHABracket,
rung_0: RungDict,
rung_1: RungDict,
rung_2: RungDict,
):
"""Test that sampling collisions are handled."""
asha.brackets = [bracket]
assert bracket.owner is asha
fidelity = 1
duplicate_trial = create_trial_for_hb((fidelity, 0.0))
new_trial = create_trial_for_hb((fidelity, 0.5))
duplicate_id_wo_fidelity = asha.get_id(duplicate_trial,
ignore_fidelity=True)
bracket.rungs[0] = dict(
n_trials=2,
resources=1,
results={duplicate_id_wo_fidelity: (0.0, duplicate_trial)},
)
asha.trial_to_brackets[duplicate_id_wo_fidelity] = 0
asha.register(duplicate_trial)
trials = [duplicate_trial, new_trial]
def sample(num=1, seed=None):
return trials
monkeypatch.setattr(asha.space, "sample", sample)
assert asha.suggest(1)[0].params == new_trial.params