def GLOA_with_disjunctive_bounds(pyomo_model):
job_result = _JobResult()
pyomo_results = SolverFactory('gdpopt').solve(
pyomo_model,
tee=True,
strategy='GLOA',
mip_solver='gams',
mip_solver_args=dict(solver='cplex',
add_options=get_base_gams_options_list()),
nlp_solver='gams',
nlp_solver_args=dict(solver='baron',
add_options=get_base_gams_options_list()),
minlp_solver='gams',
minlp_solver_args=dict(solver='baron',
add_options=get_base_gams_options_list()),
iterlim=300,
calc_disjunctive_bounds=True,
time_limit=options.time_limit)
job_result.solver_run_time = pyomo_results.solver.timing.total
job_result.pyomo_solver_status = pyomo_results.solver.status
job_result.iterations = pyomo_results.solver.iterations
job_result.termination_condition = pyomo_results.solver.termination_condition
job_result.LB = pyomo_results.problem.lower_bound
job_result.UB = pyomo_results.problem.upper_bound
return job_result
def DICOPT(pyomo_model):
job_result = _JobResult()
try:
pyomo_results = SolverFactory('gams').solve(
pyomo_model,
tee=True,
# keepfiles=True,
solver='dicopt',
add_options=get_base_gams_options_list() +
[f'option reslim={options.time_limit};'])
except ValueError as e:
# Handle GAMS interface complaining about using DICOPT for MIPs
if 'GAMS writer passed solver (dicopt) unsuitable for model type (mip)' in str(
e):
pyomo_results = SolverFactory('gams').solve(
pyomo_model,
tee=True,
solver='cplex',
add_options=get_base_gams_options_list() +
[f'option reslim={options.time_limit};'])
else:
raise
job_result.solver_run_time = pyomo_results.solver.user_time
job_result.pyomo_solver_status = pyomo_results.solver.status
job_result.termination_condition = pyomo_results.solver.termination_condition
job_result.LB = pyomo_results.problem.lower_bound
job_result.UB = pyomo_results.problem.upper_bound
return job_result
def _get_job_result(run_dir: Path, model: str, solver: str):
with run_dir.joinpath(solver, model,
job_result_filename).open('r') as result_file:
_stored_result = yaml.safe_load(result_file)
if not _stored_result:
return _JobResult()
job_result = _JobResult(**_stored_result)
if 'termination_condition' in job_result:
if job_result.termination_condition not in [None, "None"]:
job_result.termination_condition = pyomo_tc(
job_result.termination_condition)
else:
job_result.termination_condition = pyomo_tc('unknown')
if 'pyomo_solver_status' in job_result:
job_result.pyomo_solver_status = SolverStatus(
job_result.pyomo_solver_status)
return job_result
def SCIP(pyomo_model):
job_result = _JobResult()
pyomo_results = SolverFactory('gams').solve(
pyomo_model,
tee=True,
# keepfiles=True,
solver='scip',
add_options=get_base_gams_options_list() +
[f'option reslim={options.time_limit};'])
job_result.solver_run_time = pyomo_results.solver.user_time
job_result.pyomo_solver_status = pyomo_results.solver.status
job_result.termination_condition = pyomo_results.solver.termination_condition
job_result.LB = pyomo_results.problem.lower_bound
job_result.UB = pyomo_results.problem.upper_bound
return job_result
def OA(pyomo_model):
job_result = _JobResult()
pyomo_results = SolverFactory('mindtpy').solve(
pyomo_model,
strategy='OA',
tee=True,
mip_solver='gams',
mip_solver_args=dict(solver='cplex',
add_options=get_base_gams_options_list()),
nlp_solver='ipopt',
# nlp_solver_args=dict(solver='ipopth', add_options=get_base_gams_options_list()),
iteration_limit=300,
time_limit=options.time_limit)
job_result.solver_run_time = pyomo_results.solver.timing.total
job_result.pyomo_solver_status = pyomo_results.solver.status
job_result.iterations = pyomo_results.solver.iterations
job_result.termination_condition = pyomo_results.solver.termination_condition
job_result.LB = pyomo_results.problem.lower_bound
job_result.UB = pyomo_results.problem.upper_bound
return job_result
def run_test_case():
# Load test job configuration
with open(runner_config_filename) as file:
runner_options = yaml.safe_load(file)
model_name = runner_options["model name"]
solver_name = runner_options["solver name"]
# Time limit must be updated before solver library import.
time_limit = runner_options["time_limit"]
options.time_limit = time_limit
# Get model and solver objects
from pysperf.model_library import models
from pysperf.solver_library import solvers
test_model = models[model_name]
test_solver = solvers[solver_name]
job_result = _JobResult()
# Build the model
job_result.model_build_start_time = get_formatted_time_now()
pyomo_model = test_model.build_function()
job_result.model_build_end_time = get_formatted_time_now()
Path(job_model_built_filename).touch()
# Run the solver
job_result.solver_start_time = get_formatted_time_now()
solve_result = test_solver.solve_function(pyomo_model)
job_result.solver_end_time = get_formatted_time_now()
Path(job_solve_done_filename).touch()
# Update results object
job_result.update(solve_result)
# Write result to file
with open(job_result_filename, 'w') as result_file:
if 'termination_condition' in job_result:
job_result.termination_condition = str(
job_result.termination_condition)
if 'pyomo_solver_status' in job_result:
job_result.pyomo_solver_status = str(
job_result.pyomo_solver_status)
yaml.safe_dump(dict(**job_result), result_file)
def create_paver_tracefile(run_number: Optional[int] = None):
this_run_dir = get_run_dir(run_number)
_load_run_config(this_run_dir)
# Create trace file
trace_header = """\
* Trace Record Definition
* GamsSolve
* InputFileName,ModelType,SolverName,NLP,MIP,JulianDate,Direction
* ,NumberOfEquations,NumberOfVariables,NumberOfDiscreteVariables
* ,NumberOfNonZeros,NumberOfNonlinearNonZeros,OptionFile
* ,ModelStatus,SolverStatus,ObjectiveValue,ObjectiveValueEstimate
* ,SolverTime,NumberOfIterations,NumberOfDomainViolations,NumberOfNodes,#empty1
"""
trace_data = []
for model_name, solver_name in this_run_config.jobs_run - this_run_config.jobs_failed:
with this_run_dir.joinpath(
solver_name, model_name,
"pysperf_result.log").open('r') as resultfile:
job_result = _JobResult(**yaml.safe_load(resultfile))
_validate_job_result(job_result)
test_model = models[model_name]
test_solver = solvers[solver_name]
trace_line = [
model_name, # Model Name
'MINLP', # LP, MIP, NLP, etc.
solver_name, # ...
test_solver.nlp, # default NLP solver
test_solver.milp, # default MIP solver
get_julian_datetime(
datetime.strptime(job_result.model_build_start_time,
time_format)), # start day/time of job
0 if test_model.objective_sense == "minimize" else
1, # direction 0=min, 1=max
test_model.constraints, # total number of equations
test_model.variables, # total number of variables
# total number of discrete variables
test_model.binary_variables + test_model.integer_variables,
'', # 'nznum?', # number of nonzeros
'', # 'nlz?', # number of nonlinear nonzeros
0, # 1= optfile included
termination_condition_to_gams_format(
job_result.termination_condition),
# GAMS model return status - see the GAMS return codes section.
# GAMS solver return status - see the GAMS return codes section.
solver_status_to_gams(pyo.SolverStatus.ok),
job_result.UB, # value of objective function
job_result.
UB, # objective function estimate # TODO I think this only works for minimize?
job_result.solver_run_time, # resource time used (sec)
job_result.iterations, # number of solver iterations
0, # dom used
0, # nodes used
'# automatically generated by pysperf'
]
trace_data.append(trace_line)
with outputdir.joinpath("results.trc").open('w') as tracefile:
tracefile.write(textwrap.dedent(trace_header))
tracefile.write('*\n')
csvwriter = csv.writer(tracefile)
csvwriter.writerows(trace_data)