yawning_titan.experiment_helpers.sb3

Functions

init_env	Use the Stable Baselines 3 Monitor wrappper to wrap an environment in order to enable monitoring.
print_metric_stats	Take a metric name and a list of values and prints associated stats.
print_policy_eval_metrics	Output policy evaluation metrics and summary statistics.
train_and_eval	Train and Evaluate an agent.

yawning_titan.experiment_helpers.sb3.init_env(env, experiment_id)

Use the Stable Baselines 3 Monitor wrappper to wrap an environment in order to enable monitoring.

Parameters:

• env – the registered name of an OpenAI gym environment (str)

• experiment_id – a UID for the experiment (str)

Returns:

A Stable Baselines 3 Monitor Wrapped Gym Environment

yawning_titan.experiment_helpers.sb3.train_and_eval(agent_name, environment, training_timesteps, n_eval_episodes)

Train and Evaluate an agent.

Parameters:

• agent_name – the algorithm name (str)

• environment – An initlaised Open AI Gym environment

• training_timesteps – total no. of training timesteps (int)

Returns:

a trained Stable Baselines 3 agent eval_pol: the output from the Stable Baselines 3 ‘evaluate_policy’ function

Return type:

chosen_agent

yawning_titan.experiment_helpers.sb3.print_metric_stats(metric_name, metrics, raw_metrics=False)

Take a metric name and a list of values and prints associated stats.

Parameters:

• metric_name – The metric name (str)

• metrics – A list of ints/float metric readings(list)

yawning_titan.experiment_helpers.sb3.print_policy_eval_metrics(agents, evals, raw_metrics=False)

Output policy evaluation metrics and summary statistics.

Parameters:

• agents – An index of the agents that were evaluated(list)

• evals – The output from Stable Baselines 3 ‘evaluate_policy’ for each of the agents trained