f1 = sklearn.metrics.f1_score(y_true, preds, average=None)
conf_mat = sklearn.metrics.confusion_matrix(y_true, preds)
return prec, rec, f1, conf_mat
env = DrivingSimulatorEnv(defs_obs=[('/training/image/compressed',
CompressedImage)],
func_compile_obs=compile_obs,
defs_reward=[
('/rl/has_obstacle_nearby', Bool),
('/rl/distance_to_longestpath', Float32),
('/rl/car_velocity', Float32),
('/rl/last_on_opposite_path', Int16),
('/rl/on_pedestrian', Bool)
],
func_compile_reward=compile_reward,
defs_action=[('/autoDrive_KeyboardMode', Char)],
rate_action=10.0,
window_sizes={
'obs': 2,
'reward': 3
},
buffer_sizes={
'obs': 2,
'reward': 3
},
step_delay_target=0.4)
# def f_net(inputs):
# l2 = 1e-3
# state = inputs[0]
# conv = hrl.utils.Network.conv2ds(state, shape=[(32, 4, 4), (64, 4, 4), (64, 2, 2)], out_flatten=True,
"/home/pirate03/PycharmProjects/hobotrl/playground/initialD/imitaion_learning/DrSim_resnet_rename_pi_q_opt_add_logitsadam",
"""save tmp model""")
FLAGS = tf.app.flags.FLAGS
# What is the result's name?? Need check
env = DrivingSimulatorEnv(
defs_obs=[('/training/image/compressed', CompressedImage),
('/decision_result', Int16)],
func_compile_obs=compile_obs,
defs_reward=[
('/rl/has_obstacle_nearby', Bool),
('/rl/distance_to_longestpath', Float32),
('/rl/car_velocity', Float32),
('/rl/last_on_opposite_path', Int16),
('/rl/on_pedestrian', Bool)],
func_compile_reward=compile_reward,
defs_action=[('/autoDrive_KeyboardMode', Char)],
rate_action=10.0,
window_sizes={'obs': 2, 'reward': 3},
buffer_sizes={'obs': 2, 'reward': 3},
step_delay_target=0.5
)
env.observation_space = Box(low=0, high=255, shape=(640, 640, 3))
env.action_space = Discrete(3)
env.reward_range = (-np.inf, np.inf)
env.metadata = {}
# env = FrameStack(env, 1)
ACTIONS = [(Char(ord(mode)),) for mode in ['s', 'd', 'a']]