def test_set_reward_signal_nan_inf(self, float_case: float) -> None:
"""Test (not) keeping the reward signal for nan/inf."""
predictor = TemporalDifference()
state = State()
state.add_resolved_dependency(
("tensorflow", "2.3.0", "https://pypi.org/simple"))
state.add_resolved_dependency(
("flask", "0.12", "https://pypi.org/simple"))
state.add_unresolved_dependency(
("termial-random", "0.0.2", "https://pypi.org/simple"))
predictor._policy = {
("flask", "0.12", "https://pypi.org/simple"): [0.2, 1],
}
predictor._steps_taken = 2
predictor._steps_reward = 1.2
predictor._next_state = state
assert (predictor.set_reward_signal(
state, ("tensorflow", "2.0.0", "https://pypi.org/simple"),
float_case) is None)
assert predictor._policy == {
("flask", "0.12", "https://pypi.org/simple"): [1.4, 2],
("tensorflow", "2.3.0", "https://pypi.org/simple"): [1.2, 1],
}
assert predictor._steps_taken == 0
assert predictor._steps_reward == 0.0
assert predictor._next_state is None
def test_pre_run(self) -> None:
"""Test initialization done before running."""
predictor = TemporalDifference()
predictor._policy = {("tensorflow", "2.0.0", "https://pypi.org/simple"): [1.0, 2]}
predictor._temperature_history = [(0.212, True, 0.23, 100)]
predictor._temperature = 12.3
context = flexmock(limit=42)
with predictor.assigned_context(context):
predictor.pre_run()
assert predictor._policy == {}
assert predictor._temperature_history == []
assert isinstance(predictor._temperature, float)
assert predictor._temperature == float(context.limit)
def test_set_reward_signal_unseen(self) -> None:
"""Test keeping the reward signal for an unseen step."""
reward = 42.24
package_tuple = ("tensorflow", "2.0.0", "https://thoth-station.ninja")
state = flexmock()
state.should_receive("iter_resolved_dependencies").and_return([package_tuple]).once()
predictor = TemporalDifference()
predictor._policy = {
("numpy", "1.0.0", "https://pypi.org/simple"): [30.30, 92],
}
predictor._steps_taken = 1
predictor.set_reward_signal(state, None, reward)
assert predictor._policy == {
package_tuple: [42.24, 1],
("numpy", "1.0.0", "https://pypi.org/simple"): [30.30, 92],
}
def test_do_exploitation(self) -> None:
"""Tests on exploitation computation."""
predictor = TemporalDifference()
predictor._policy = {
("tensorflow", "2.1.0", "https://thoth-station.ninja"):
[2020.21, 666],
("tensorflow", "2.0.0", "https://thoth-station.ninja"):
[16.61, 1992],
("numpy", "1.0.0", "https://pypi.org/simple"): [30.30, 92],
}
state = flexmock()
state.should_receive("iter_unresolved_dependencies").and_return([
("spacy", "2.2.4", "https://pypi.org/simple"),
("numpy", "1.0.0", "https://pypi.org/simple"),
("tensorflow", "2.1.0", "https://thoth-station.ninja"),
]).once()
state.should_receive("get_random_unresolved_dependency").times(0)
assert predictor._do_exploitation(state) == (
"tensorflow",
"2.1.0",
"https://thoth-station.ninja",
)