def __init__(
self,
venv: vec_env.VecEnv,
expert_data: Union[Iterable[Mapping], types.Transitions],
expert_batch_size: int,
gen_algo: on_policy_algorithm.OnPolicyAlgorithm,
*,
# FIXME(sam) pass in discrim net directly; don't ask for kwargs indirectly
discrim_kwargs: Optional[Mapping] = None,
**kwargs,
):
"""Generative Adversarial Imitation Learning.
Most parameters are described in and passed to `AdversarialTrainer.__init__`.
Additional parameters that `GAIL` adds on top of its superclass initializer are
as follows:
Args:
discrim_kwargs: Optional keyword arguments to use while constructing the
DiscrimNetGAIL.
"""
discrim_kwargs = discrim_kwargs or {}
discrim = discrim_nets.DiscrimNetGAIL(
venv.observation_space, venv.action_space, **discrim_kwargs
)
super().__init__(
venv, gen_algo, discrim, expert_data, expert_batch_size, **kwargs
)
def __init__(
self,
venv: vec_env.VecEnv,
expert_data: Union[Iterable[Mapping], types.Transitions],
expert_batch_size: int,
gen_algo: on_policy_algorithm.OnPolicyAlgorithm,
*,
# FIXME(sam) pass in discrim net directly; don't ask for kwargs indirectly
discrim_kwargs: Optional[Mapping] = None,
**kwargs,
):
"""Generative Adversarial Imitation Learning.
Most parameters are described in and passed to `AdversarialTrainer.__init__`.
Additional parameters that `GAIL` adds on top of its superclass initializer are
as follows:
Args:
discrim_kwargs: Optional keyword arguments to use while constructing the
DiscrimNetGAIL.
"""
if discrim_kwargs.get("discriminator_hid_sizes", {}):
print(discrim_kwargs.get("discriminator_hid_sizes", {}))
discrim_net = ActObsMLP(
action_space=venv.action_space,
observation_space=venv.observation_space,
hid_sizes=discrim_kwargs.get("discriminator_hid_sizes", {}),
)
discrim = discrim_nets.DiscrimNetGAIL(venv.observation_space,
venv.action_space,
discrim_net=discrim_net)
#discrim=th.load('/home/leonor/Desktop/HIRL_dissertation/hierarchical/output/train_adversarial/GripperReachOpenUR5Sim-v0/20210529_163416_e39a59_servidor_100_demos/checkpoints/final/discrim.pt', map_location=th.device('cpu'))
else:
discrim_kwargs = discrim_kwargs or {}
discrim = discrim_nets.DiscrimNetGAIL(venv.observation_space,
venv.action_space,
**discrim_kwargs)
super().__init__(venv, gen_algo, discrim, expert_data,
expert_batch_size, **kwargs)
def __init__(
self,
venv: vec_env.VecEnv,
expert_data: Union[Iterable[Mapping], types.Transitions],
expert_batch_size: int,
gen_algo: on_policy_algorithm.OnPolicyAlgorithm,
*,
# FIXME(sam) pass in discrim net directly; don't ask for kwargs indirectly
discrim_kwargs: Optional[Mapping] = None,
policy,
_init_setup_model,
**kwargs,
):
"""Generative Adversarial Imitation Learning.
Most parameters are described in and passed to `AdversarialTrainer.__init__`.
Additional parameters that `GAIL` adds on top of its superclass initializer are
as follows:
Args:
discrim_kwargs: Optional keyword arguments to use while constructing the
DiscrimNetGAIL.
"""
env_name = 'ReachObjectUR5Sim-v0'
num_vec = 1
max_episode_steps = 1000
venv = util.make_vec_env(
env_name,
num_vec,
#seed=_seed,
#parallel=parallel,
#log_dir=log_dir,
max_episode_steps=max_episode_steps,
)
discrim_kwargs = discrim_kwargs or {}
discrim = discrim_nets.DiscrimNetGAIL(venv.observation_space,
venv.action_space,
**discrim_kwargs)
super().__init__(
venv=venv,
gen_algo=gen_algo,
discrim=discrim,
expert_data=expert_data,
expert_batch_size=expert_batch_size, #**kwargs
)
self.verbose = False
self.tensorboard_log = None
self.use_sde = False
def _setup_gail_provide_discriminator(venv):
discriminator = discrim_nets.ActObsMLP(venv.action_space,
venv.observation_space,
hid_sizes=(4, 4, 4))
return discrim_nets.DiscrimNetGAIL(venv.observation_space,
venv.action_space, discriminator)
def _setup_gail(venv):
return discrim_nets.DiscrimNetGAIL(venv.observation_space,
venv.action_space)