ast_toolbox.mcts.AdaptiveStressTesting module

class ast_toolbox.mcts.AdaptiveStressTesting.ASTAction(action)

Bases: object

get()

Get the true action.

Returns:action – The true actions used in the env.

class ast_toolbox.mcts.AdaptiveStressTesting.ASTParams(max_steps, log_interval, log_tabular, log_dir=None, n_itr=100)

Bases: object

Structure that stores internal parameters for AST.

Parameters:max_steps (int, optional) – The maximum search depth.

class ast_toolbox.mcts.AdaptiveStressTesting.ASTState(t_index, parent, action)

Bases: object

The AST state.

Parameters:

• t_index (int) – The index of the timestep.
• parent (ast_toolbox.mcts.AdaptiveStressTesting.ASTState) – The parent state.
• action (ast_toolbox.mcts.AdaptiveStressTesting.ASTAction) – The action leading to this state.

class ast_toolbox.mcts.AdaptiveStressTesting.AdaptiveStressTest(p, env, top_paths)

Bases: object

The AST wrapper for MCTS using the actions in env.action_space.

Parameters:

• p (ast_toolbox.mcts.AdaptiveStressTesting.ASTParams) – The AST parameters
• env (ast_toolbox.envs.go_explore_ast_env.GoExploreASTEnv.) – The environment.
• top_paths (ast_toolbox.mcts.BoundedPriorityQueues, optional) – The bounded priority queue to store top-rewarded trajectories.

explore_action(s, tree)

Randomly sample an action for the exploration.

Parameters:

• s (ast_toolbox.mcts.AdaptiveStressTesting.ASTState) – The current state.
• tree (dict) – The searching tree.

Returns:

action (ast_toolbox.mcts.AdaptiveStressTesting.ASTAction) – The sampled action.

get_reward()

Get the current AST reward.

Returns:reward (bool) – The AST reward.

initialize()

Initialize training variables.

Returns:env_reset – The reset result from the env.

isterminal()

Check whether the current path is finished.

Returns:isterinal (bool) – Whether the current path is finished.

logging()

Logging the training information.

random_action()

Randomly sample an action for the rollout.

Returns:action (ast_toolbox.mcts.AdaptiveStressTesting.ASTAction) – The sampled action.

reset_step_count()

Reset the env step count.

transition_model()

Generate the transition model used in MCTS.

Returns:transition_model (ast_toolbox.mcts.MDP.TransitionModel) – The transition model.

update(action)

Update the environment as well as the assosiated parameters.

Parameters:action (ast_toolbox.mcts.AdaptiveStressTesting.ASTAction) – The AST action. Returns:

• obs (numpy.ndarry) – The observation from the env step.
• reward (float) – The reward from the env step.
• done (bool) – The terminal indicator from the env step.
• info (dict) – The env info from the env step.

ast_toolbox.mcts.AdaptiveStressTesting.get_action_sequence(s)

Get the action sequence that leads to the state.

Parameters:s (ast_toolbox.mcts.AdaptiveStressTesting.ASTState) – The target state. Returns:actions (list[ast_toolbox.mcts.AdaptiveStressTesting.ASTAction]) – The action sequences leading to the target state.