def argsparser():
parser = argparse.ArgumentParser("Tensorflow Implementation of GAIL")
parser.add_argument('--env_id', help='environment ID', default='DeepMimic')
parser.add_argument('--seed', help='RNG seed', type=int, default=0)
parser.add_argument('--checkpoint_dir', help='the directory to save model', default='checkpoint_tmp')
parser.add_argument('--log_dir', help='the directory to save log file', default='log')
parser.add_argument('--load_model_path', help='if provided, load the model', type=str, default=None)
# Task
parser.add_argument('--task', type=str, choices=['train', 'evaluate', 'sample'], default='train')
# for evaluatation
boolean_flag(parser, 'stochastic_policy', default=False, help='use stochastic/deterministic policy to evaluate')
boolean_flag(parser, 'save_sample', default=False, help='save the trajectories or not')
# Mujoco Dataset Configuration
parser.add_argument('--traj_limitation', type=int, default=-1)
# Optimization Configuration
parser.add_argument('--g_step', help='number of steps to train policy in each epoch', type=int, default=3)
# Network Configuration (Using MLP Policy)
parser.add_argument('--policy_hidden_size', type=int, default=100)
# Algorithms Configuration
parser.add_argument('--max_kl', type=float, default=0.01)
parser.add_argument('--policy_entcoeff', help='entropy coefficiency of policy', type=float, default=0)
# Traing Configuration
parser.add_argument('--save_per_iter', help='save model every xx iterations', type=int, default=100)
parser.add_argument('--num_timesteps', help='number of timesteps per episode', type=int, default=1e6)
parser.add_argument('--pretrained_weight_path', help='path of pretrained weights', type=str, default=None)
return parser.parse_args()
def bc_argparser(description="Behavioral Cloning Experiment"):
"""Create an argparse.ArgumentParser for behavioral cloning-related tasks"""
parser = argparse(description)
parser.add_argument('--note', help='w/e', type=str, default=None)
parser.add_argument('--env_id', help='environment identifier', default='Hopper-v2')
parser.add_argument('--horizon', help='maximum number of timesteps in an episode',
type=int, default=None)
parser.add_argument('--seed', help='RNG seed', type=int, default=0)
parser.add_argument('--checkpoint_dir', help='directory to save the models',
default=None)
parser.add_argument('--log_dir', help='directory to save the log files',
default='data/logs')
parser.add_argument('--summary_dir', help='directory to save the summaries',
default='data/summaries')
parser.add_argument('--task', help='task to carry out', type=str,
choices=['clone',
'evaluate_bc_policy'],
default='clone')
parser.add_argument('--expert_path', help='.npz archive containing the demos',
type=str, default=None)
parser.add_argument('--num_demos', help='number of expert demo trajs for imitation',
type=int, default=None)
parser.add_argument('--save_frequency', help='save model every xx iterations',
type=int, default=100)
parser.add_argument('--num_iters', help='cummulative number of iterations since launch',
type=int, default=int(1e6))
parser.add_argument('--batch_size', help='minibatch size', type=int, default=64)
parser.add_argument('--lr', help='adam learning rate', type=float, default=3e-4)
parser.add_argument('--clip_norm', type=float, default=None)
boolean_flag(parser, 'render', help='whether to render the interaction traces', default=False)
parser.add_argument('--num_trajs', help='number of trajectories to evaluate/gather',
type=int, default=10)
parser.add_argument('--exact_model_path', help='exact path of the model',
type=str, default=None)
parser.add_argument('--model_ckpt_dir', help='checkpoint directory containing the models',
type=str, default=None)
parser.add_argument('--demos_dir', type=str, help='directory to save the demonstrations',
default='data/expert_demonstrations')
boolean_flag(parser, 'rmsify_obs', default=True)
parser.add_argument('--hid_widths', nargs='+', type=int, default=[64, 64])
parser.add_argument('--hid_nonlin', type=str, default='leaky_relu',
choices=['relu', 'leaky_relu', 'prelu', 'elu', 'selu', 'tanh'])
parser.add_argument('--hid_w_init', type=str, default='he_normal',
choices=['he_normal', 'he_uniform', 'xavier_normal', 'xavier_uniform'])
return parser
def argsparser():
parser = argparse.ArgumentParser("Tensorflow Implementation of GAIL")
parser.add_argument('--env_id',
help='environment ID',
default='CartPole-v0')
parser.add_argument('--seed', help='RNG seed', type=int, default=0)
parser.add_argument('--expert_path',
type=str,
default='expert_data/cartpole')
parser.add_argument('--checkpoint_dir',
help='the directory to save model',
default='checkpoint')
parser.add_argument('--log_dir',
help='the directory to save log file',
default='log')
parser.add_argument('--load_model_path',
help='if provided, load the model',
type=str,
default=None)
# Task
parser.add_argument('--task',
type=str,
choices=['train', 'evaluate', 'sample'],
default='train')
# for evaluatation
boolean_flag(parser,
'stochastic_policy',
default=False,
help='use stochastic/deterministic policy to evaluate')
boolean_flag(parser,
'save_sample',
default=False,
help='save the trajectories or not')
# Mujoco Dataset Configuration
parser.add_argument('--traj_limitation', type=int, default=-1)
parser.add_argument('--loss_percent', type=float, default=0.0)
# Optimization Configuration
parser.add_argument('--g_step',
help='number of steps to train policy in each epoch',
type=int,
default=1)
parser.add_argument(
'--d_step',
help='number of steps to train discriminator in each epoch',
type=int,
default=1)
# Network Configuration (Using MLP Policy)
parser.add_argument('--policy_hidden_size', type=int, default=100)
parser.add_argument('--adversary_hidden_size', type=int, default=100)
# Algorithms Configuration
parser.add_argument('--algo',
type=str,
choices=['trpo', 'state', 'agail'],
default='trpo')
parser.add_argument('--max_kl', type=float, default=0.01)
parser.add_argument('--policy_entcoeff',
help='entropy coefficiency of policy',
type=float,
default=0)
parser.add_argument('--adversary_entcoeff',
help='entropy coefficiency of discriminator',
type=float,
default=1e-3)
# Traing Configuration
parser.add_argument('--save_per_iter',
help='save model every xx iterations',
type=int,
default=100)
parser.add_argument('--num_timesteps',
help='number of timesteps per episode',
type=int,
default=1e6)
# Behavior Cloning
boolean_flag(parser,
'pretrained',
default=False,
help='Use BC to pretrain')
parser.add_argument('--BC_max_iter',
help='Max iteration for training BC',
type=int,
default=None)
return parser.parse_args()
