[Feature Request] Call train from Python code

[younik]

[Feature Request] Call train from Python code

[araffin]

Hello,

I have checked that there is no similar issue in the repo

is that what you want: https://github.com/DLR-RM/rl-baselines3-zoo#passing-arguments-in-an-interactive-session ?

See discussion in #297
if you want more granularity, I would recommend using the ExperimentManager class directly:

rl-baselines3-zoo/rl_zoo3/train.py

Lines 221 to 275 in e4667b6

	exp_manager = ExperimentManager(
	args,
	args.algo,
	env_id,
	args.log_folder,
	args.tensorboard_log,
	args.n_timesteps,
	args.eval_freq,
	args.eval_episodes,
	args.save_freq,
	args.hyperparams,
	args.env_kwargs,
	args.eval_env_kwargs,
	args.trained_agent,
	args.optimize_hyperparameters,
	args.storage,
	args.study_name,
	args.n_trials,
	args.max_total_trials,
	args.n_jobs,
	args.sampler,
	args.pruner,
	args.optimization_log_path,
	n_startup_trials=args.n_startup_trials,
	n_evaluations=args.n_evaluations,
	truncate_last_trajectory=args.truncate_last_trajectory,
	uuid_str=uuid_str,
	seed=args.seed,
	log_interval=args.log_interval,
	save_replay_buffer=args.save_replay_buffer,
	verbose=args.verbose,
	vec_env_type=args.vec_env,
	n_eval_envs=args.n_eval_envs,
	no_optim_plots=args.no_optim_plots,
	device=args.device,
	config=args.conf_file,
	show_progress=args.progress,
	)
	# Prepare experiment and launch hyperparameter optimization if needed
	results = exp_manager.setup_experiment()
	if results is not None:
	model, saved_hyperparams = results
	if args.track:
	# we need to save the loaded hyperparameters
	args.saved_hyperparams = saved_hyperparams
	assert run is not None # make mypy happy
	run.config.setdefaults(vars(args))
	# Normal training
	if model is not None:
	exp_manager.learn(model)
	exp_manager.save_trained_model(model)
	else:
	exp_manager.hyperparameters_optimization()

[younik]

Oh, that's correct, thank you. I didn't find it, sorry about that.