Python GaussianMLPPolicy.get_param_values 예제들

예제 #1

파일: test_trpo.py 프로젝트: gntoni/garage

def test_trpo_deterministic_nan():
    env = TheanoEnv(DummyEnv())
    policy = GaussianMLPPolicy(env_spec=env.spec, hidden_sizes=(1, ))
    policy._l_log_std.param.set_value([np.float32(np.log(1e-8))])
    baseline = ZeroBaseline(env_spec=env.spec)
    algo = TRPO(env=env,
                policy=policy,
                baseline=baseline,
                n_itr=10,
                batch_size=1000,
                max_path_length=100,
                step_size=0.01)
    algo.train()
    assert not np.isnan(np.sum(policy.get_param_values()))

예제 #2

파일: test_trpo.py 프로젝트: gntoni/garage

def test_trpo_relu_nan():
    env = TheanoEnv(DummyEnv())
    policy = GaussianMLPPolicy(env_spec=env.spec,
                               hidden_nonlinearity=naive_relu,
                               hidden_sizes=(1, ))
    baseline = ZeroBaseline(env_spec=env.spec)
    algo = TRPO(env=env,
                policy=policy,
                baseline=baseline,
                n_itr=1,
                batch_size=1000,
                max_path_length=100,
                step_size=0.001)
    algo.train()
    assert not np.isnan(np.sum(policy.get_param_values()))