def launch(args):
# Experiment parameters
exp_kwargs = {
'goal_tolerance_parameters': {
'inc_tol_obs': True,
'initial_tol': 0.020,
'final_tol': 0.001,
'N_ts': 200000,
'function': 'decay',
'set_tol': 0
},
'relative_q': True,
'joint_representation': 'trig',
'resample_joints': False,
'normalize_obs': False
}
# create the ddpg_agent, test relative decay curriculum
env = gym.make(args.env_name, **exp_kwargs)
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env)
# create the ddpg agent to interact with the environment
ddpg_trainer = ddpg_agent(args, env, env_params)
ddpg_trainer.learn()
def launch(args):
env = gym.make(args.env_name)
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env)
# create the ddpg_agent
if args.alg == 'gac':
# create the ddpg agent to interact with the environment
print("Start GAC...")
gac_trainer = gac_agent(args, env, env_params)
gac_trainer.learn()
elif args.alg == 'sac':
# create the ddpg agent to interact with the environment
print("Start SAC...")
sac_trainer = sac_agent(args, env, env_params)
sac_trainer.learn()
elif args.alg == 'td3':
print('Start TD3...')
td3_trainer = td3_agent(args, env, env_params)
td3_trainer.learn()
else:
# create the ddpg agent to interact with the environment
print("Start DDPG...")
ddpg_trainer = ddpg_agent(args, env, env_params)
ddpg_trainer.learn()
def launch(args):
# create the ddpg_agent
env = gym.make(args.env_name)
# set random seeds for reproduce
env.seed(args.seed)
random.seed(args.seed)
np.random.seed(args.seed)
torch.manual_seed(args.seed)
if args.cuda:
torch.cuda.manual_seed(args.seed)
# get the environment parameters
env_params = get_env_params(env)
if args.algo == 'ddpg':
eval = ddpg_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'sac':
eval = sac_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'forward':
eval = dynamics_learner(args, env, env_params)
eval.eval_dynamics(PATH)
elif args.algo == 'inverse':
eval = inverse_dynamics_learner(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'double':
eval = double_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'share':
eval = double_agent_share(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'shaping_share':
eval = double_agent_shaping_share(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'attention':
eval = double_agent_attention(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'sgg':
eval = sgg_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'alt':
eval = alternate_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'br':
eval = sac_br_agent(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'br_ppo':
eval = sac_br_agent_ppo(args, env, env_params)
eval.visualize(PATH)
elif args.algo == 'skill':
eval = sac_skill_agent(args, env, env_params)
eval.visualize(PATH)
else:
raise NotImplementedError
def launch(args):
# create the ddpg_agent
env = gym.make(args.env_name)
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env)
# create the ddpg agent to interact with the environment
ddpg_trainer = ddpg_agent(args, env, env_params)
ddpg_trainer.learn()
def launch(args):
# create the ddpg_agent
env = SharedBlockTouchSensorsEnvSparse()
env = TimeLimit(env, 200)
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env)
# create the ddpg agent to interact with the environment
ddpg_trainer = ddpg_agent(args, env, env_params)
ddpg_trainer.learn()
def launch(args):
# create the ddpg_agent
env = gym.make('gym_multiRL:MultiRL{}'.format(args.env_name))
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env, args)
# create the ddpg agent to interact with the environment
# if args.actor_loss_type=='mgda':
# ddpg_trainer = ddpg_agent_mgda(args, env, env_params)
# else:
ddpg_trainer = ddpg_agent(args, env, env_params)
ddpg_trainer.learn()
def launch(args):
# create the ddpg_agent
env = gym.make(args.env_name)
# set random seeds for reproduce
env.seed(args.seed + MPI.COMM_WORLD.Get_rank())
random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
np.random.seed(args.seed + MPI.COMM_WORLD.Get_rank())
torch.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
if args.cuda:
torch.cuda.manual_seed(args.seed + MPI.COMM_WORLD.Get_rank())
# get the environment parameters
env_params = get_env_params(env)
# create the ddpg agent to interact with the environment
reset_dir('./experiments/log')
tb_wrt = SummaryWriter('./experiments/log')
ddpg_trainer = ddpg_agent(args,
env,
env_params,
tb_wrt,
is_her=args.is_her)
ddpg_trainer.learn()