Coverage for mlos_bench/mlos_bench/tests/optimizers/toy_optimization_loop_test.py: 100%

1# 

2# Copyright (c) Microsoft Corporation. 

3# Licensed under the MIT License. 

4# 

5"""Toy optimization loop to test the optimizers on mock benchmark environment.""" 

6 

7import logging 

8 

9import pytest 

10 

11from mlos_bench.environments.base_environment import Environment 

12from mlos_bench.environments.mock_env import MockEnv 

13from mlos_bench.optimizers.base_optimizer import Optimizer 

14from mlos_bench.optimizers.convert_configspace import tunable_values_to_configuration 

15from mlos_bench.optimizers.mlos_core_optimizer import MlosCoreOptimizer 

16from mlos_bench.optimizers.mock_optimizer import MockOptimizer 

17from mlos_bench.tunables.tunable_groups import TunableGroups 

18from mlos_core.data_classes import Suggestion 

19from mlos_core.optimizers.bayesian_optimizers.smac_optimizer import SmacOptimizer 

20from mlos_core.util import config_to_series 

21 

22# For debugging purposes output some warnings which are captured with failed tests. 

23DEBUG = True 

24logger = logging.debug 

25if DEBUG: 

26 logger = logging.warning 

27 

28 

29def _optimize(env: Environment, opt: Optimizer) -> tuple[float, TunableGroups]: 

30 """Toy optimization loop.""" 

31 assert opt.not_converged() 

32 

33 while opt.not_converged(): 

34 

35 with env as env_context: 

36 

37 tunables = opt.suggest() 

38 

39 logger("tunables: %s", str(tunables)) 

40 # pylint: disable=protected-access 

41 if isinstance(opt, MlosCoreOptimizer) and isinstance(opt._opt, SmacOptimizer): 

42 config = tunable_values_to_configuration(tunables) 

43 config_series = config_to_series(config) 

44 logger("config: %s", str(config)) 

45 try: 

46 logger( 

47 "prediction: %s", 

48 opt._opt.surrogate_predict(suggestion=Suggestion(config=config_series)), 

49 ) 

50 except RuntimeError: 

51 pass 

52 

53 assert env_context.setup(tunables) 

54 

55 (status, _ts, output) = env_context.run() 

56 assert status.is_succeeded() 

57 assert output is not None 

58 score = output["score"] 

59 assert isinstance(score, float) 

60 assert 60 <= score <= 120 

61 logger("score: %s", str(score)) 

62 

63 opt.register(tunables, status, output) 

64 

65 (best_score, best_tunables) = opt.get_best_observation() 

66 assert best_score is not None and len(best_score) == 1 

67 assert isinstance(best_tunables, TunableGroups) 

68 return (best_score["score"], best_tunables) 

69 

70 

71def test_mock_optimization_loop(mock_env_no_noise: MockEnv, mock_opt: MockOptimizer) -> None: 

72 """Toy optimization loop with mock environment and optimizer.""" 

73 (score, tunables) = _optimize(mock_env_no_noise, mock_opt) 

74 assert score == pytest.approx(64.9, 0.01) 

75 assert tunables.get_param_values() == { 

76 "vmSize": "Standard_B2ms", 

77 "idle": "halt", 

78 "kernel_sched_migration_cost_ns": 117026, 

79 "kernel_sched_latency_ns": 149827706, 

80 } 

81 

82 

83def test_mock_optimization_loop_no_defaults( 

84 mock_env_no_noise: MockEnv, 

85 mock_opt_no_defaults: MockOptimizer, 

86) -> None: 

87 """Toy optimization loop with mock environment and optimizer.""" 

88 (score, tunables) = _optimize(mock_env_no_noise, mock_opt_no_defaults) 

89 assert score == pytest.approx(60.97, 0.01) 

90 assert tunables.get_param_values() == { 

91 "vmSize": "Standard_B2s", 

92 "idle": "halt", 

93 "kernel_sched_migration_cost_ns": 49123, 

94 "kernel_sched_latency_ns": 234760738, 

95 } 

96 

97 

98def test_flaml_optimization_loop(mock_env_no_noise: MockEnv, flaml_opt: MlosCoreOptimizer) -> None: 

99 """Toy optimization loop with mock environment and FLAML optimizer.""" 

100 (score, tunables) = _optimize(mock_env_no_noise, flaml_opt) 

101 assert score == pytest.approx(60.15, 0.01) 

102 assert tunables.get_param_values() == { 

103 "vmSize": "Standard_B2s", 

104 "idle": "halt", 

105 "kernel_sched_migration_cost_ns": -1, 

106 "kernel_sched_latency_ns": 13718105, 

107 } 

108 

109 

110# @pytest.mark.skip(reason="SMAC is not deterministic") 

111def test_smac_optimization_loop(mock_env_no_noise: MockEnv, smac_opt: MlosCoreOptimizer) -> None: 

112 """Toy optimization loop with mock environment and SMAC optimizer.""" 

113 (score, tunables) = _optimize(mock_env_no_noise, smac_opt) 

114 expected_score = 70.33 

115 expected_tunable_values = { 

116 "vmSize": "Standard_B2s", 

117 "idle": "mwait", 

118 "kernel_sched_migration_cost_ns": 297669, 

119 "kernel_sched_latency_ns": 290365137, 

120 } 

121 assert score == pytest.approx(expected_score, 0.01) 

122 assert tunables.get_param_values() == expected_tunable_values