def test_check_training_not_completed_training(sampler):
"""
Assert the flow is forced to train if training did not complete when
the sampler was checkpointed.
"""
sampler.completed_training = False
train, force = NestedSampler.check_training(sampler)
assert train is True
assert force is True
def test_check_training_train_on_empty(sampler):
"""
Assert the flow is forced to train if training the pool is empty and
`train_on_empty` is true but the proposal was not in the process of
popluating.
"""
sampler.completed_training = True
sampler.train_on_empty = True
sampler.proposal = MagicMock()
sampler.proposal.populated = False
sampler.proposal.populating = False
train, force = NestedSampler.check_training(sampler)
assert train is True
assert force is True
def test_check_training_acceptance(sampler):
"""
Assert that training will be true but not forced if the acceptance
threshold is met and retraining on acceptance is enabled.
"""
sampler.completed_training = True
sampler.train_on_empty = True
sampler.proposal = MagicMock()
sampler.proposal.populated = True
sampler.proposal.populating = False
sampler.acceptance_threshold = 0.1
sampler.mean_block_acceptance = 0.01
sampler.retrain_acceptance = True
train, force = NestedSampler.check_training(sampler)
assert train is True
assert force is False
def test_check_training_false(sampler, config):
"""
Test a range of different scenarios that should all not start training.
"""
sampler.completed_training = True
sampler.train_on_empty = config.get('train_on_empty', False)
sampler.proposal = MagicMock()
sampler.proposal.populated = config.get('populated', False)
sampler.proposal.populating = config.get('populating', False)
sampler.acceptance_threshold = config.get('acceptance_threshold', 0.1)
sampler.mean_block_acceptance = config.get('mean_acceptance', 0.2)
sampler.retrain_acceptance = config.get('retrain_acceptance', False)
sampler.iteration = config.get('iteration', 3000)
sampler.last_updated = config.get('last_updated', 2500)
sampler.training_frequency = config.get('training_frequency', 1000)
train, force = NestedSampler.check_training(sampler)
assert train is False
assert force is False
def test_check_training_iteration(sampler):
"""
Assert that training will be true but not forced if a training iteration
is reached (n iterations have passed since last updated).
"""
sampler.completed_training = True
sampler.train_on_empty = True
sampler.proposal = MagicMock()
sampler.proposal.populated = True
sampler.proposal.populating = False
sampler.acceptance_threshold = 0.1
sampler.mean_block_acceptance = 0.2
sampler.retrain_acceptance = False
sampler.iteration = 3521
sampler.last_updated = 2521
sampler.training_frequency = 1000
train, force = NestedSampler.check_training(sampler)
assert train is True
assert force is False