"""
Helper to create a model with different hyperparameters
"""
return ALGOS[args.algo](env=create_env(n_envs, no_log=True),
tensorboard_log=tensorboard_log,
verbose=0,
**kwargs)
study = hyperparam_optimization(args.algo,
create_model,
create_env,
n_trials=args.n_trials,
n_timesteps=n_timesteps,
hyperparams=hyperparams,
n_jobs=args.n_jobs,
seed=args.seed,
sampler_method=args.sampler,
pruner_method=args.pruner,
n_startup_trials=args.n_startup_trials,
n_evaluations=args.n_evaluations,
storage=args.storage,
study_name=args.study_name,
verbose=args.verbose,
deterministic_eval=not is_atari)
data_frame = study.trials_dataframe()
report_name = (f"report_{env_id}_{args.n_trials}-trials-{n_timesteps}"
f"-{args.sampler}-{args.pruner}_{int(time.time())}.csv")
study_name = (f"report_{env_id}_{args.n_trials}-trials-{n_timesteps}"
f"-{args.sampler}-{args.pruner}_{int(time.time())}.pkl")
log_path = os.path.join(args.log_folder, args.algo, report_name)
study_path = os.path.join(args.log_folder, args.algo, study_name)
def create_model(*_args, **kwargs):
"""
Helper to create a model with different hyperparameters
"""
return ALGOS[args.algo](env=create_env(n_envs),
tensorboard_log=tensorboard_log,
verbose=0,
**kwargs)
data_frame = hyperparam_optimization(args.algo,
create_model,
create_env,
n_trials=args.n_trials,
n_timesteps=n_timesteps,
hyperparams=hyperparams,
n_jobs=args.n_jobs,
seed=args.seed,
sampler_method=args.sampler,
pruner_method=args.pruner,
verbose=args.verbose)
report_name = "report_{}_{}-trials-{}-{}-{}.csv".format(
env_id, args.n_trials, n_timesteps, args.sampler, args.pruner)
log_path = os.path.join(args.log_folder, args.algo, report_name)
if args.verbose:
print("Writing report to {}".format(log_path))
os.makedirs(os.path.dirname(log_path), exist_ok=True)
env = VecNormalize(env, **local_normalize_kwargs)
return env
def create_model(*_arg, **kwargs):
net_arch = [dict(pi=[512, 256], vf=[512, 256])]
policy_kwargs = dict(act_fun=tf.nn.relu, net_arch=net_arch)
return PPO2(env=create_env(n_envs, no_log=True),
policy_kwargs=policy_kwargs,
tensorboard_log=tensorboard_log,
verbose=0,
**kwargs)
data_frame = hyperparam_optimization('ppo2',
create_model,
create_env,
n_trials=n_trials,
hyperparams=hyperparams,
n_jobs=n_jobs,
seed=seed,
sampler_method=sampler,
pruner_method=pruner,
verbose=True)
report_name = "report_{}_{}-trials-{}-{}-{}_{}.csv".format(
env_id, n_trials, n_timesteps, sampler, pruner, int(time.time()))
log_path = os.path.join(log_path, 'report', report_name)
os.makedirs(os.path.dirname(log_path), exist_ok=True)
data_frame.to_csv(log_path)
