def train():
args = ['--n_workers=2', 'UnityRemote-v0']
environment_definitions['state_shape'] = (7, )
environment_definitions['action_shape'] = (5, )
environment_definitions['actions'] = [('fx', 0.1), ('fx', -0.1),
('fz', 0.1), ('fz', -0.1),
('noop', 0.0)]
environment_definitions['action_meaning'] = [
'tx_right', 'tx_left', 'tz_toward', 'tz_backward', 'NOOP'
]
environment_definitions['state_wrapper'] = state_wrapper
run_train(environment_definitions, args)
def train():
args = ['--n_workers=8', '--steps_per_update=30', 'UnityRemote-v0']
make_env_def()
run_train(environment_definitions, args)
def train():
args = ['--n_workers=4', '--preprocessing=external', 'UnityRemote-v0']
make_env_def()
run_train(environment_definitions, args)
def train():
args = ['--n_workers=4', 'UnityRemote-v0']
make_env_def()
run_train(environment_definitions, args)