def main():
args = get_args()
print(args)
create_data_folders()
args.gradients = ['discount']
study_regress(args)
plot_results(args)
env.reinit()
memory = ReplayBuffer()
# Initialise the policy/actor
policy = PolicyNet(params.lr_actor, params.init_alpha, params.lr_alpha, params.target_entropy_alpha)
pw = PolicyWrapper(policy, params.policy_type, params.env_name, params.team_name, params.max_episode_steps)
pw.duration_flag = False
# Initialise the critics
critic = DoubleQNet(params.lr_critic,params.gamma, params.tau)
plot_policy(policy, env, True, params.env_name, params.study_name, '_ante_', j, plot=False)
simulation.train(memory, pw, critic, policy_loss_file, critic_loss_file)
plot_policy(policy, env, True, params.env_name, params.study_name, '_post_', j, plot=False)
plot_critic(env, params.env_name, critic.q1, policy, params.study_name, '_q1_post_', j)
plot_critic(env, params.env_name, critic.q2, policy, params.study_name, '_q2_post_', j)
critic.q1.save_model('data/critics/{}#{}#SAC{}.pt'.format(params.env_name, params.team_name, str(j)))
critic.q2.save_model('data/critics/{}#{}#SAC{}.pt'.format(params.env_name, params.team_name, str(j)))
simulation.env.close()
chrono.stop()
if __name__ == '__main__':
args = get_args()
print(args)
create_data_folders()
main(args)
plot_results(args)