def main(early_stopping,optimizer_class_o,problem_o,num_evals_o,random_seed_o,path_o):
config.set_early_stopping(early_stopping)
optimizer_class = OPTIM_MAP[optimizer_class_o]
hyperparams = OPTIM_PARAMETERS[optimizer_class_o]
sampler = OPTIM_SAMPLERS_TUNED_PRIOR[optimizer_class_o]
runner = StandardRunner
if optimizer_class_o in ['sgdmcwclr', 'adamwclrdecay']:
optimizer_class.set_max_epochs(config.get_testproblem_default_setting(problem_o)['num_epochs'])
runner = LearningRateScheduleRunner
tuner = RandomSearch(optimizer_class, hyperparams, sampler, runner=runner, ressources=num_evals_o)
tuner.tune(problem_o, rerun_best_setting=False, output_dir=path_o, random_seed=random_seed_o, weight_decay=0)
def main():
args = parse_arguments()
# config.set_framework('pytorch')
config.set_early_stopping(args.early_stopping)
optimizer_class = OPTIM_MAP[args.optim]
hyperparams = OPTIM_PARAMETERS[args.optim]
sampler = OPTIM_SAMPLERS_TUNED_PRIOR[args.optim]
runner = StandardRunner
if args.optim in ['sgdmcwclr', 'adamwclrdecay']:
optimizer_class.set_max_epochs(
config.get_testproblem_default_setting(args.problem)['num_epochs'])
runner = LearningRateScheduleRunner
tuner = RandomSearch(optimizer_class,
hyperparams,
sampler,
runner=runner,
ressources=args.num_evals)
tuner.tune(args.problem,
rerun_best_setting=False,
output_dir=args.log_path,
random_seed=args.random_seed,
weight_decay=0)
def init_default_problem_params(testproblem):
defaults = {}
tesproblem_default = config.get_testproblem_default_setting(testproblem)
defaults[testproblem] = tesproblem_default
return defaults
def _use_default(testproblem, key):
return global_config.get_testproblem_default_setting(testproblem)[key]