metadata
license: mit
This is a pre-trained model (DDPG agent) for XArm6Reach simulation task, defined in here.
You can load and try it with test_xarm6_reach.py in the repository.
Hyperparameters:
OrderedDict([
('batch_size', 2048),
('buffer_size', 1000000),
('gamma', 0.95),
('learning_rate', 0.001),
('learning_starts', 100),
('policy', 'MultiInputPolicy'),
('policy_kwargs', dict(net_arch=[512, 512, 512], n_critics=2)),
('replay_buffer_class', HerReplayBuffer),
('replay_buffer_kwargs', dict( goal_selection_strategy='future', n_sampled_goal=4)),
('tau', 0.05),
])