Coverage for sparkle/CLI/construct_portfolio_selector.py: 71%
103 statements
« prev ^ index » next coverage.py v7.6.4, created at 2024-11-05 14:48 +0000
1#!/usr/bin/env python3
2"""Sparkle command to construct a portfolio selector."""
3import sys
4import argparse
5from pathlib import Path
7import runrunner as rrr
8from runrunner.base import Runner
10from sparkle.CLI.help import global_variables as gv
11from sparkle.structures import PerformanceDataFrame, FeatureDataFrame
12from sparkle.CLI.help import logging as sl
13from sparkle.platform.settings_objects import SettingState
14from sparkle.CLI.help import argparse_custom as ac
15from sparkle.CLI.help.reporting_scenario import Scenario
16from sparkle.platform import CommandName, COMMAND_DEPENDENCIES
17from sparkle.CLI.initialise import check_for_initialise
20def parser_function() -> argparse.ArgumentParser:
21 """Define the command line arguments."""
22 parser = argparse.ArgumentParser(
23 description="Command to construct a portfolio selector over all known features "
24 "solver performances.")
25 parser.add_argument(*ac.RecomputePortfolioSelectorArgument.names,
26 **ac.RecomputePortfolioSelectorArgument.kwargs)
27 parser.add_argument(*ac.SelectorTimeoutArgument.names,
28 **ac.SelectorTimeoutArgument.kwargs)
29 parser.add_argument(*ac.SparkleObjectiveArgument.names,
30 **ac.SparkleObjectiveArgument.kwargs)
31 parser.add_argument(*ac.SelectorAblationArgument.names,
32 **ac.SelectorAblationArgument.kwargs)
33 parser.add_argument(*ac.RunOnArgument.names,
34 **ac.RunOnArgument.kwargs)
35 parser.add_argument(*ac.SettingsFileArgument.names,
36 **ac.SettingsFileArgument.kwargs)
37 return parser
40def judge_exist_remaining_jobs(feature_data_csv: Path,
41 performance_data_csv: Path) -> bool:
42 """Return whether there are remaining feature or performance computation jobs."""
43 feature_data = FeatureDataFrame(feature_data_csv)
44 if feature_data.has_missing_vectors():
45 return True
46 performance_data = PerformanceDataFrame(performance_data_csv)
47 return performance_data.has_missing_values
50def main(argv: list[str]) -> None:
51 """Main method of construct portfolio selector."""
52 # Log command call
53 sl.log_command(sys.argv)
55 # Define command line arguments
56 parser = parser_function()
58 # Process command line arguments
59 args = parser.parse_args(argv)
60 selector_timeout = args.selector_timeout
61 flag_recompute_portfolio = args.recompute_portfolio_selector
62 solver_ablation = args.solver_ablation
64 check_for_initialise(
65 COMMAND_DEPENDENCIES[CommandName.CONSTRUCT_PORTFOLIO_SELECTOR]
66 )
67 if ac.set_by_user(args, "settings_file"):
68 gv.settings().read_settings_ini(
69 args.settings_file, SettingState.CMD_LINE
70 ) # Do first, so other command line options can override settings from the file
71 if ac.set_by_user(args, "objectives"):
72 gv.settings().set_general_sparkle_objectives(
73 args.objectives, SettingState.CMD_LINE
74 )
75 if args.run_on is not None:
76 gv.settings().set_run_on(
77 args.run_on.value, SettingState.CMD_LINE)
78 run_on = gv.settings().get_run_on()
80 print("Start constructing Sparkle portfolio selector ...")
81 selector = gv.settings().get_general_sparkle_selector()
83 flag_judge_exist_remaining_jobs = judge_exist_remaining_jobs(
84 gv.settings().DEFAULT_feature_data_path,
85 gv.settings().DEFAULT_performance_data_path)
87 if flag_judge_exist_remaining_jobs:
88 print("There remain unperformed feature computation jobs or performance "
89 "computation jobs!")
90 print("Please first execute all unperformed jobs before constructing Sparkle "
91 "portfolio selector")
92 print("Sparkle portfolio selector is not successfully constructed!")
93 sys.exit(-1)
95 # Selector (AutoFolio) cannot handle cutoff time less than 2, adjust if needed
96 cutoff_time = max(gv.settings().get_general_target_cutoff_time(), 2)
98 # Determine the objective function
99 objective = gv.settings().get_general_sparkle_objectives()[0]
101 performance_data = PerformanceDataFrame(gv.settings().DEFAULT_performance_data_path)
102 feature_data = FeatureDataFrame(gv.settings().DEFAULT_feature_data_path)
104 if feature_data.has_missing_value():
105 print("WARNING: Missing values in the feature data, will be imputed as the mean "
106 "value of all other non-missing values! Imputing all missing values...")
107 feature_data.impute_missing_values()
109 # TODO: Allow user to specify subsets of data to be used
111 # Selector is named after the solvers it can predict, sort for permutation invariance
112 solvers = sorted([s.name for s in gv.settings().DEFAULT_solver_dir.iterdir()])
113 selection_scenario_path = (
114 gv.settings().DEFAULT_selection_output
115 / gv.settings().DEFAULT_general_sparkle_selector.name
116 / "_".join(solvers))
118 # Update latest scenario
119 gv.latest_scenario().set_selection_scenario_path(selection_scenario_path)
120 gv.latest_scenario().set_latest_scenario(Scenario.SELECTION)
121 # Set to default to overwrite possible old path
122 gv.latest_scenario().set_selection_test_case_directory()
124 selector_path = selection_scenario_path / "portfolio_selector"
125 sbatch_options = gv.settings().get_slurm_extra_options(as_args=True)
126 if selector_path.exists() and not flag_recompute_portfolio:
127 print("Portfolio selector already exists. Set the recompute flag to re-create.")
128 sys.exit()
130 selector_path.parent.mkdir(exist_ok=True, parents=True)
132 selector_run = selector.construct(selector_path,
133 performance_data,
134 feature_data,
135 objective,
136 cutoff_time,
137 selector_timeout,
138 run_on=run_on,
139 sbatch_options=sbatch_options,
140 base_dir=sl.caller_log_dir)
141 if run_on == Runner.LOCAL:
142 print("Sparkle portfolio selector constructed!")
143 else:
144 print("Sparkle portfolio selector constructor running...")
146 dependencies = [selector_run]
147 if solver_ablation:
148 for solver in performance_data.solvers:
149 solver_name = Path(solver).name
150 ablate_solver_dir = selection_scenario_path / f"ablate_{solver_name}"
151 ablate_solver_selector = ablate_solver_dir / "portfolio_selector"
152 if (ablate_solver_selector.exists() and not flag_recompute_portfolio):
153 print(f"Portfolio selector without {solver_name} already exists. "
154 "Set the recompute flag to re-create.")
155 continue
156 ablate_solver_dir.mkdir(exist_ok=True, parents=True)
157 ablated_performance_data = performance_data.copy()
158 ablated_performance_data.remove_solver(solver)
159 ablated_run = selector.construct(ablate_solver_selector,
160 ablated_performance_data,
161 feature_data,
162 objective,
163 cutoff_time,
164 selector_timeout,
165 run_on=run_on,
166 sbatch_options=sbatch_options,
167 base_dir=sl.caller_log_dir)
168 dependencies.append(ablated_run)
169 if run_on == Runner.LOCAL:
170 print(f"Portfolio selector without {solver_name} constructed!")
171 else:
172 print(f"Portfolio selector without {solver_name} constructor running...")
174 # Compute the marginal contribution
175 with_actual = "--actual" if solver_ablation else ""
176 cmd = (f"sparkle/CLI/compute_marginal_contribution.py --perfect {with_actual} "
177 f"{ac.SparkleObjectiveArgument.names[0]} {objective}")
178 marginal_contribution = rrr.add_to_queue(
179 runner=run_on,
180 cmd=cmd,
181 name=CommandName.COMPUTE_MARGINAL_CONTRIBUTION,
182 base_dir=sl.caller_log_dir,
183 dependencies=dependencies,
184 sbatch_options=sbatch_options)
185 dependencies.append(marginal_contribution)
186 if run_on == Runner.LOCAL:
187 marginal_contribution.wait()
188 print("Selector marginal contribution computing done!")
189 else:
190 print(f"Running selector construction. Waiting for Slurm job(s) with id(s): "
191 f"{', '.join([d.run_id for d in dependencies])}")
193 # Write used settings to file
194 gv.settings().write_used_settings()
195 # Write used scenario to file
196 gv.latest_scenario().write_scenario_ini()
197 sys.exit(0)
200if __name__ == "__main__":
201 main(sys.argv[1:])