def main():
rng = random.PRNGKey(0)
expert_rollouts = run_expert_rollouts(rng)
expert_rollouts_flat = np.reshape(expert_rollouts,
(-1, expert_rollouts.shape[-1]))
state_min, state_max = get_estop_bounds(expert_rollouts)
# Debug plot!
plt.figure()
for i in range(17):
ax = plt.subplot(2, 9, i + 2)
plt.hist(expert_rollouts_flat[:, i], bins=256)
ax.yaxis.set_ticklabels([])
plt.title(i)
plt.axvline(state_min[i], c="r")
plt.axvline(state_max[i], c="r")
plt.show()
debug_run(env_spec,
train_config,
seed=0,
state_min=state_min,
state_max=state_max)
from research.estop.gym.ddpg_training import debug_run, make_default_ddpg_train_config from research.estop.gym.ant import env_name, reward_adjustment from research.estop.gym.gym_wrappers import build_env_spec env_spec = build_env_spec(env_name, reward_adjustment) debug_run(env_spec, make_default_ddpg_train_config(env_spec), respect_gym_done=True)
from research.estop.gym.ddpg_training import debug_run, make_default_ddpg_train_config from research.estop.gym.ant import env_name, reward_adjustment from research.estop.gym.gym_wrappers import build_env_spec env_spec = build_env_spec(env_name, reward_adjustment) debug_run(env_spec, make_default_ddpg_train_config(env_spec))