def main():
env = MiningCraftingEnv(stochastic_reset=False)
d_s = env.d_s
n_a = env.high_a[0]
given_conditions = False
# Learning conditions from data or fixed conditions?
if given_conditions:
primitive_skills = [
PrimitiveSkill(
Effect([i], [0], [1], n_a), i,
FixedCondition(env.conditions[i],
[1.] * len(env.conditions[i])))
for i in range(n_a)
]
else:
primitive_skills = [
PrimitiveSkill(Effect([i], [0], [1], n_a), i, None)
for i in range(n_a)
]
runner = Runner(env, primitive_skills, verbose=1)
goals = [
MiningCraftingEnv.goal_stick, MiningCraftingEnv.goal_stone_pick,
MiningCraftingEnv.goal_coal, MiningCraftingEnv.goal_furnace,
MiningCraftingEnv.goal_smelt_iron, MiningCraftingEnv.goal_iron_pick,
MiningCraftingEnv.goal_gold_ore, MiningCraftingEnv.goal_goldware,
MiningCraftingEnv.goal_necklace, MiningCraftingEnv.goal_earrings
]
n_success = 8
for i in range(len(goals)):
success = 0
for ite in range(20):
goal = SparseState(goals[i][0], goals[i][1], d_s)
if runner.run(goal, 20):
success += 1
if success >= n_success:
print("####### Success at skill {} after {} "
"episodes\n\n".format(i, ite + 1))
break
if success < n_success:
print("####### Failed at skill {}\n\n".format(i))
print("\n\n\n\n%%%%%%% END")
print_skills(env, runner.il.skill_base)
def main():
n_runs = 1
env_name = "mining"
if env_name == "mining":
env = MiningCraftingEnv(stochastic_reset=False)
elif env_name == "baking":
env = BakingEnv(stochastic_reset=False)
elif env_name == "random":
env = RandomGraphEnv(stochastic_reset=False, noise_prob=0.2)
else:
raise RuntimeError("Unknown environment name")
n_mcts_iter = 100
goal = env.curriculum[-1]
s_goal = SparseState(goal[0], goal[1], env.d_s)
agent = Mcts(iteration_limit=n_mcts_iter)
trajectory_length, time_ = run_n_episodes(env, agent, s_goal, n_runs)
print("#### FINAL ####")
print("number of runs:", n_runs)
print("number mcts iter:", n_mcts_iter)
print("trajectory length:", trajectory_length)
print("time:", time_)
print(seq_a)
# Compute stats
time_buffer = time_buffer[1:len(time_buffer)] + [time.time() - t]
success_buffer = success_buffer[1:len(success_buffer)] + [success]
length_buffer = length_buffer[1:len(length_buffer)] + [len(seq_a)]
stats = compute_stats(success_buffer, length_buffer, time_buffer)
return stats[0], stats[1]
if __name__ == "__main__":
n_runs = 1
env_name = "mining"
if env_name == "mining":
env = MiningCraftingEnv(stochastic_reset=False)
elif env_name == "baking":
env = BakingEnv(stochastic_reset=False)
elif env_name == "random":
env = RandomGraphEnv(stochastic_reset=False, noise_prob=0.2)
else:
raise RuntimeError("Unknown environment name")
goal = env.curriculum[-1]
s_goal = SparseState(goal[0], goal[1], env.d_s)
n_a = env.high_a[0]
primitive_skills = [
PrimitiveSkill(
Effect([i], [0], [1], n_a), i,
FixedCondition(env.conditions[i], [1.] * len(env.conditions[i])))
for i in range(n_a)
n_ep_buffer = n_ep_buffer[1:len(n_ep_buffer)] + [n_train_ep]
# Print learned skills
if n_ep == n_eps - 1:
print_skills(env, agent.skill_base)
stats = compute_stats(success_buffer, length_buffer, time_buffer,
n_ep_buffer)
return stats[0], stats[1], stats[2]
if __name__ == "__main__":
n_runs = 1
env_name = "mining"
if env_name == "mining":
env, ep_length = MiningCraftingEnv(stochastic_reset=False), 20
elif env_name == "baking":
env, ep_length = BakingEnv(stochastic_reset=False), 40
elif env_name == "drawer":
env, ep_length = DrawerEnv(stochastic_reset=False), 10
else:
raise RuntimeError("Unknown environment name")
goal = env.curriculum[-1]
ds = env.d_s
n_a = env.high_a[0]
n_max_success_per_goal = 5
s_goal = SparseState(goal[0], goal[1], ds)
primitive_skills = [PrimitiveSkill(Effect([i], [0], [1], n_a), i, None) for
i in range(n_a)]
trajectory_length, time_, n_train_ep = run_n_episodes(