def test_sampling_random_state():
f = hp_module.Float('f', 1e-3, 1e3, sampling='log')
rand_sample = f.random_sample()
assert rand_sample >= f.min_value
assert rand_sample <= f.max_value
def log_scale(x, min_value, max_value):
return math.log(x / min_value) / math.log(max_value / min_value)
x = 1e-1
min_value, max_value = 1e-10, 1e10
# Scale x to [0, 1].
x_scaled = log_scale(x, min_value, max_value)
# Scale back.
x_rescaled = hp_module._log_sample(x_scaled, min_value, max_value)
assert np.allclose(x, x_rescaled)
f = hp_module.Float('f', 1e-3, 1e3, sampling='reverse_log')
rand_sample = f.random_sample()
assert rand_sample >= f.min_value
assert rand_sample <= f.max_value
def reverse_log_scale(x, a, b):
return 1 - math.log((b + a - x) / a) / math.log(b / a)
x = 1e5
min_value, max_value = 1e-10, 1e10
x_scaled = reverse_log_scale(x, min_value, max_value)
x_rescaled = hp_module._reverse_log_sample(x_scaled, min_value, max_value)
assert np.allclose(x, x_rescaled)
def test_Float():
# Test with step arg
linear = hp_module.Float('linear',
min_value=0.5,
max_value=9.5,
step=0.1,
default=9.)
linear = hp_module.Float.from_config(linear.get_config())
assert linear.default == 9.
assert 0.5 <= linear.random_sample() <= 9.5
assert isinstance(linear.random_sample(), float)
assert linear.random_sample(123) == linear.random_sample(123)
# Test without step arg
linear = hp_module.Float('linear',
min_value=0.5,
max_value=6.5,
default=2.)
linear = hp_module.Float.from_config(linear.get_config())
assert linear.default == 2.
assert 0.5 <= linear.random_sample() < 6.5
assert isinstance(linear.random_sample(), float)
assert linear.random_sample(123) == linear.random_sample(123)
# No default
linear = hp_module.Float('linear', min_value=0.5, max_value=9.5, step=0.1)
assert linear.default == 0.5
def test_bayesian_oracle_with_zero_y(tmp_dir):
hp_list = [hp_module.Choice('a', [1, 2], default=1),
hp_module.Int('b', 3, 10, default=3),
hp_module.Float('c', 0, 1, 0.1, default=0),
hp_module.Fixed('d', 7),
hp_module.Choice('e', [9, 0], default=9)]
oracle = bo_module.BayesianOptimizationOracle()
for i in range(100):
oracle.populate_space(str(i), hp_list)
oracle.result(str(i), 0)
def test_save_before_result(tmp_dir):
hp_list = [hp_module.Choice('a', [1, 2], default=1),
hp_module.Int('b', 3, 10, default=3),
hp_module.Float('c', 0, 1, 0.1, default=0),
hp_module.Fixed('d', 7),
hp_module.Choice('e', [9, 0], default=9)]
oracle = bo_module.BayesianOptimizationOracle()
oracle.populate_space(str(1), hp_list)
oracle.save(os.path.join(tmp_dir, 'temp_oracle'))
oracle.result(str(1), 0)
def test_sampling_arg():
f = hp_module.Float('f', 1e-20, 1e10, sampling='log')
f = hp_module.Float.from_config(f.get_config())
assert f.sampling == 'log'
i = hp_module.Int('i', 0, 10, sampling='linear')
i = hp_module.Int.from_config(i.get_config())
assert i.sampling == 'linear'
with pytest.raises(ValueError, match='`sampling` must be one of'):
hp_module.Int('j', 0, 10, sampling='invalid')
def test_float_proto():
hp = hp_module.Float('a', -10, 10, sampling='linear', default=3)
proto = hp.to_proto()
assert proto.name == 'a'
assert proto.min_value == -10.
assert proto.max_value == 10.
assert proto.sampling == kerastuner_pb2.Sampling.LINEAR
assert proto.default == 3.
# Zero is the default, gets converted to `None` in `from_proto`.
assert proto.step == 0.
new_hp = hp_module.Float.from_proto(proto)
assert new_hp.get_config() == hp.get_config()
def test_bayesian_dynamic_space(tmp_dir):
hp_list = [hp_module.Choice('a', [1, 2], default=1)]
oracle = bo_module.BayesianOptimizationOracle()
for i in range(10):
oracle.populate_space(str(i), hp_list)
oracle.result(str(i), i)
hp_list.append(hp_module.Int('b', 3, 10, default=3))
assert 'b' in oracle.populate_space('1_0', hp_list)['values']
hp_list.append(hp_module.Float('c', 0, 1, 0.1, default=0))
assert 'c' in oracle.populate_space('1_1', hp_list)['values']
hp_list.append(hp_module.Fixed('d', 7))
assert 'd' in oracle.populate_space('1_2', hp_list)['values']
hp_list.append(hp_module.Choice('e', [9, 0], default=9))
assert 'e' in oracle.populate_space('1_3', hp_list)['values']
def test_reverse_log_sampling_random_state():
f = hp_module.Float('f', 1e-3, 1e3, sampling='reverse_log')
rand_sample = f.random_sample()
assert rand_sample >= f.min_value
assert rand_sample <= f.max_value
val = 1e-3
prob = hp_module.value_to_cumulative_prob(val, f)
assert prob == 0
new_val = hp_module.cumulative_prob_to_value(prob, f)
assert np.isclose(val, new_val)
val = 1
prob = hp_module.value_to_cumulative_prob(val, f)
assert prob > 0 and prob < 1
new_val = hp_module.cumulative_prob_to_value(prob, f)
assert np.isclose(val, new_val)
def test_log_sampling_random_state():
f = hp_module.Float("f", 1e-3, 1e3, sampling="log")
rand_sample = f.random_sample()
assert rand_sample >= f.min_value
assert rand_sample <= f.max_value
val = 1e-3
prob = hp_module.value_to_cumulative_prob(val, f)
assert prob == 0
new_val = hp_module.cumulative_prob_to_value(prob, f)
assert np.isclose(val, new_val)
val = 1
prob = hp_module.value_to_cumulative_prob(val, f)
assert prob == 0.5
new_val = hp_module.cumulative_prob_to_value(prob, f)
assert np.isclose(val, new_val)
val = 1e3
prob = hp_module.value_to_cumulative_prob(val, f)
assert prob == 1
new_val = hp_module.cumulative_prob_to_value(prob, f)
assert np.isclose(val, new_val)
def test_sampling_arg():
f = hp_module.Float("f", 1e-20, 1e10, sampling="log")
f = hp_module.Float.from_config(f.get_config())
assert f.sampling == "log"
i = hp_module.Int("i", 0, 10, sampling="linear")
i = hp_module.Int.from_config(i.get_config())
assert i.sampling == "linear"
with pytest.raises(ValueError, match="`sampling` must be one of"):
hp_module.Int("j", 0, 10, sampling="invalid")
with pytest.raises(
ValueError,
match="`sampling` `min_value` 1 is greater than the `max_value` 0",
):
hp_module.Int("k", 1, 0, sampling="linear")
with pytest.raises(
ValueError,
match="`sampling` `min_value` 1 is greater than the `max_value` 0",
):
hp_module.Int("k", 1, 0, sampling="linear")
