def collect_trajectories(env: gym.Env,
agent: Agent,
n_games: int = 10) -> np.ndarray:
for _ in range(n_games):
state = env.reset()
done: bool = False
state_history: list[np.ndarray] = []
while not done:
state_history.append(state)
action = agent.choose_action(state)
next_state, _, done, _ = env.step(action)
state = next_state
return np.vstack(state_history)
# Init. Agent
agent = Agent(env=env, n_games=n_games, training=False)
agent.load_models(data_path)
for i in tqdm(range(n_games), desc=f'Testing', total=n_games):
score_history: List[np.float32] = [] * n_games
for _ in tqdm(range(n_games), desc=f'Testing', total=n_games):
score = 0
done = False
# Initial Reset of Environment
state = env.reset()
while not done:
action = agent.choose_action(state)
next_state, reward, done, _ = env.step(action)
agent.memory.add(state, action, reward, next_state, done)
state = copy.deepcopy(next_state)
score += reward
score_history.append(score)
print(f'Test Analysis:\n'
f'Mean:{np.mean(score_history)}\n'
f'Variance:{np.std(score_history)}')
test_data.append({'Test Score': score_history})
env=env,
batch_size=64,
layer1_size=256,
layer2_size=128,
n_actions=3)
#agent.load_models()
# np.random.seed(1)
score_history = []
for i in range(50):
obs = env.reset()
done = False
score = 0
while not done:
act = agent.choose_action(obs)
print(act)
new_state, reward, done, info = env.step(act)
agent.remember(obs, act, reward, new_state, int(done))
agent.learn()
score += reward
obs = new_state
#env.render()
score_history.append(score)
if i % 10 == 0:
agent.save_models()
env.render()
print('episode ', i, 'score %.2f' % score,
'trailing 25 games avg %.3f' % np.mean(score_history[-25:]))