ml-agents/ml-agents/tests/trainers/test_trainer_controller.py

import json
import os
from unittest.mock import *

import yaml
import pytest

from mlagents.trainers.trainer_controller import TrainerController
from mlagents.trainers.ppo.trainer import PPOTrainer
from mlagents.trainers.bc.offline_trainer import OfflineBCTrainer
from mlagents.trainers.bc.online_trainer import OnlineBCTrainer
from mlagents.envs.exception import UnityEnvironmentException


@pytest.fixture
def dummy_config():
    return yaml.load(
        '''
        default:
            trainer: ppo
            batch_size: 32
            beta: 5.0e-3
            buffer_size: 512
            epsilon: 0.2
            gamma: 0.99
            hidden_units: 128
            lambd: 0.95
            learning_rate: 3.0e-4
            max_steps: 5.0e4
            normalize: true
            num_epoch: 5
            num_layers: 2
            time_horizon: 64
            sequence_length: 64
            summary_freq: 1000
            use_recurrent: false
            memory_size: 8
            use_curiosity: false
            curiosity_strength: 0.0
            curiosity_enc_size: 1
        ''')


@pytest.fixture
def dummy_online_bc_config():
    return yaml.load(
        '''
        default:
            trainer: online_bc
            brain_to_imitate: ExpertBrain
            batches_per_epoch: 16
            batch_size: 32
            beta: 5.0e-3
            buffer_size: 512
            epsilon: 0.2
            gamma: 0.99
            hidden_units: 128
            lambd: 0.95
            learning_rate: 3.0e-4
            max_steps: 5.0e4
            normalize: true
            num_epoch: 5
            num_layers: 2
            time_horizon: 64
            sequence_length: 64
            summary_freq: 1000
            use_recurrent: false
            memory_size: 8
            use_curiosity: false
            curiosity_strength: 0.0
            curiosity_enc_size: 1
        ''')


@pytest.fixture
def dummy_offline_bc_config():
    return yaml.load(
        '''
        default:
            trainer: offline_bc
            demo_path: '''
        + os.path.dirname(os.path.abspath(__file__)) + '''/test.demo
            batches_per_epoch: 16
            batch_size: 32
            beta: 5.0e-3
            buffer_size: 512
            epsilon: 0.2
            gamma: 0.99
            hidden_units: 128
            lambd: 0.95
            learning_rate: 3.0e-4
            max_steps: 5.0e4
            normalize: true
            num_epoch: 5
            num_layers: 2
            time_horizon: 64
            sequence_length: 64
            summary_freq: 1000
            use_recurrent: false
            memory_size: 8
            use_curiosity: false
            curiosity_strength: 0.0
            curiosity_enc_size: 1
        ''')

@pytest.fixture
def dummy_offline_bc_config_with_override():
    base = dummy_offline_bc_config()
    base['testbrain'] = {}
    base['testbrain']['normalize'] = False
    return base

@pytest.fixture
def dummy_bad_config():
    return yaml.load(
        '''
        default:
            trainer: incorrect_trainer
            brain_to_imitate: ExpertBrain
            batches_per_epoch: 16
            batch_size: 32
            beta: 5.0e-3
            buffer_size: 512
            epsilon: 0.2
            gamma: 0.99
            hidden_units: 128
            lambd: 0.95
            learning_rate: 3.0e-4
            max_steps: 5.0e4
            normalize: true
            num_epoch: 5
            num_layers: 2
            time_horizon: 64
            sequence_length: 64
            summary_freq: 1000
            use_recurrent: false
            memory_size: 8
        ''')

@pytest.fixture
def basic_trainer_controller(brain_info):
    return TrainerController(
        model_path='test_model_path',
        summaries_dir='test_summaries_dir',
        run_id='test_run_id',
        save_freq=100,
        meta_curriculum=None,
        load=True,
        train=True,
        keep_checkpoints=False,
        lesson=None,
        external_brains={'testbrain': brain_info},
        training_seed=99
    )

@patch('numpy.random.seed')
@patch('tensorflow.set_random_seed')
def test_initialization_seed(numpy_random_seed, tensorflow_set_seed):
    seed = 27
    TrainerController('', '', '1', 1, None, True, False, False, None, {}, seed)
    numpy_random_seed.assert_called_with(seed)
    tensorflow_set_seed.assert_called_with(seed)


def assert_bc_trainer_constructed(trainer_cls, input_config, tc, expected_brain_info, expected_config):
    def mock_constructor(self, brain, trainer_params, training, load, seed, run_id):
        assert(brain == expected_brain_info)
        assert(trainer_params == expected_config)
        assert(training == tc.train_model)
        assert(load == tc.load_model)
        assert(seed == tc.seed)
        assert(run_id == tc.run_id)

    with patch.object(trainer_cls, "__init__", mock_constructor):
        tc.initialize_trainers(input_config)
        assert('testbrain' in tc.trainers)
        assert(isinstance(tc.trainers['testbrain'], trainer_cls))


def assert_ppo_trainer_constructed(input_config, tc, expected_brain_info,
                                   expected_config, expected_reward_buff_cap=0):
    def mock_constructor(self, brain, reward_buff_cap, trainer_parameters, training, load, seed, run_id):
        assert(brain == expected_brain_info)
        assert(trainer_parameters == expected_config)
        assert(reward_buff_cap == expected_reward_buff_cap)
        assert(training == tc.train_model)
        assert(load == tc.load_model)
        assert(seed == tc.seed)
        assert(run_id == tc.run_id)

    with patch.object(PPOTrainer, "__init__", mock_constructor):
        tc.initialize_trainers(input_config)
        assert('testbrain' in tc.trainers)
        assert(isinstance(tc.trainers['testbrain'], PPOTrainer))


@patch('mlagents.envs.BrainInfo')
def test_initialize_trainer_parameters_uses_defaults(BrainInfoMock):
    brain_info_mock = BrainInfoMock()
    tc = basic_trainer_controller(brain_info_mock)

    full_config = dummy_offline_bc_config()
    expected_config = full_config['default']
    expected_config['summary_path'] = tc.summaries_dir + '/test_run_id_testbrain'
    expected_config['model_path'] = tc.model_path + '/testbrain'
    expected_config['keep_checkpoints'] = tc.keep_checkpoints

    assert_bc_trainer_constructed(OfflineBCTrainer, full_config, tc, brain_info_mock, expected_config)


@patch('mlagents.envs.BrainInfo')
def test_initialize_trainer_parameters_override_defaults(BrainInfoMock):
    brain_info_mock = BrainInfoMock()
    tc = basic_trainer_controller(brain_info_mock)

    full_config = dummy_offline_bc_config_with_override()
    expected_config = full_config['default']
    expected_config['summary_path'] = tc.summaries_dir + '/test_run_id_testbrain'
    expected_config['model_path'] = tc.model_path + '/testbrain'
    expected_config['keep_checkpoints'] = tc.keep_checkpoints

    # Override value from specific brain config
    expected_config['normalize'] = False

    assert_bc_trainer_constructed(OfflineBCTrainer, full_config, tc, brain_info_mock, expected_config)


@patch('mlagents.envs.BrainInfo')
def test_initialize_online_bc_trainer(BrainInfoMock):
    brain_info_mock = BrainInfoMock()
    tc = basic_trainer_controller(brain_info_mock)

    full_config = dummy_online_bc_config()
    expected_config = full_config['default']
    expected_config['summary_path'] = tc.summaries_dir + '/test_run_id_testbrain'
    expected_config['model_path'] = tc.model_path + '/testbrain'
    expected_config['keep_checkpoints'] = tc.keep_checkpoints

    assert_bc_trainer_constructed(OnlineBCTrainer, full_config, tc, brain_info_mock, expected_config)


@patch('mlagents.envs.BrainInfo')
def test_initialize_ppo_trainer(BrainInfoMock):
    brain_info_mock = BrainInfoMock()
    tc = basic_trainer_controller(brain_info_mock)

    full_config = dummy_config()
    expected_config = full_config['default']
    expected_config['summary_path'] = tc.summaries_dir + '/test_run_id_testbrain'
    expected_config['model_path'] = tc.model_path + '/testbrain'
    expected_config['keep_checkpoints'] = tc.keep_checkpoints

    assert_ppo_trainer_constructed(full_config, tc, brain_info_mock, expected_config)


@patch('mlagents.envs.BrainInfo')
def test_initialize_invalid_trainer_raises_exception(BrainInfoMock):
    brain_info_mock = BrainInfoMock()
    tc = basic_trainer_controller(brain_info_mock)
    bad_config = dummy_bad_config()

    try:
        tc.initialize_trainers(bad_config)
        assert(1 == 0, "Initialize trainers with bad config did not raise an exception.")
    except UnityEnvironmentException:
        pass


def trainer_controller_with_start_learning_mocks():
    trainer_mock = MagicMock()
    trainer_mock.get_step = 0
    trainer_mock.get_max_steps = 5
    trainer_mock.parameters = {'some': 'parameter'}
    trainer_mock.write_tensorboard_text = MagicMock()

    brain_info_mock = MagicMock()
    tc = basic_trainer_controller(brain_info_mock)
    tc.initialize_trainers = MagicMock()
    tc.trainers = {'testbrain': trainer_mock}
    tc.take_step = MagicMock()

    def take_step_sideeffect(env, curr_info):
        tc.trainers['testbrain'].get_step += 1
        if tc.trainers['testbrain'].get_step > 10:
            raise KeyboardInterrupt
    tc.take_step.side_effect = take_step_sideeffect

    tc._export_graph = MagicMock()
    tc._save_model = MagicMock()
    return tc, trainer_mock


@patch('tensorflow.reset_default_graph')
def test_start_learning_trains_forever_if_no_train_model(tf_reset_graph):
    tc, trainer_mock = trainer_controller_with_start_learning_mocks()
    tc.train_model = False

    trainer_config = dummy_config()
    tf_reset_graph.return_value = None

    env_mock = MagicMock()
    env_mock.close = MagicMock()
    env_mock.reset = MagicMock()

    tc.start_learning(env_mock, trainer_config)
    tf_reset_graph.assert_called_once()
    tc.initialize_trainers.assert_called_once_with(trainer_config)
    env_mock.reset.assert_called_once()
    assert (tc.take_step.call_count == 11)
    tc._export_graph.assert_not_called()
    tc._save_model.assert_not_called()
    env_mock.close.assert_called_once()


@patch('tensorflow.reset_default_graph')
def test_start_learning_trains_until_max_steps_then_saves(tf_reset_graph):
    tc, trainer_mock = trainer_controller_with_start_learning_mocks()
    trainer_config = dummy_config()
    tf_reset_graph.return_value = None

    brain_info_mock = MagicMock()
    env_mock = MagicMock()
    env_mock.close = MagicMock()
    env_mock.reset = MagicMock(return_value=brain_info_mock)

    tc.start_learning(env_mock, trainer_config)
    tf_reset_graph.assert_called_once()
    tc.initialize_trainers.assert_called_once_with(trainer_config)
    env_mock.reset.assert_called_once()
    assert(tc.take_step.call_count == trainer_mock.get_max_steps + 1)
    env_mock.close.assert_called_once()
    tc._save_model.assert_called_once_with(steps=6)


def test_start_learning_updates_meta_curriculum_lesson_number():
    tc, trainer_mock = trainer_controller_with_start_learning_mocks()
    trainer_config = dummy_config()

    brain_info_mock = MagicMock()
    env_mock = MagicMock()
    env_mock.close = MagicMock()
    env_mock.reset = MagicMock(return_value=brain_info_mock)

    meta_curriculum_mock = MagicMock()
    meta_curriculum_mock.set_all_curriculums_to_lesson_num = MagicMock()
    tc.meta_curriculum = meta_curriculum_mock
    tc.lesson = 5

    tc.start_learning(env_mock, trainer_config)
    meta_curriculum_mock.set_all_curriculums_to_lesson_num.assert_called_once_with(tc.lesson)


def trainer_controller_with_take_step_mocks():
    trainer_mock = MagicMock()
    trainer_mock.get_step = 0
    trainer_mock.get_max_steps = 5
    trainer_mock.parameters = {'some': 'parameter'}
    trainer_mock.write_tensorboard_text = MagicMock()

    brain_info_mock = MagicMock()
    tc = basic_trainer_controller(brain_info_mock)
    tc.trainers = {'testbrain': trainer_mock}

    return tc, trainer_mock


def test_take_step_resets_env_on_global_done():
    tc, trainer_mock = trainer_controller_with_take_step_mocks()

    brain_info_mock = MagicMock()
    action_data_mock_out = [None, None, None, None, None]
    trainer_mock.take_action = MagicMock(return_value=action_data_mock_out)
    trainer_mock.add_experiences = MagicMock()
    trainer_mock.process_experiences = MagicMock()
    trainer_mock.update_policy = MagicMock()
    trainer_mock.write_summary = MagicMock()
    trainer_mock.trainer.increment_step_and_update_last_reward = MagicMock()
    env_mock = MagicMock()
    step_data_mock_out = MagicMock()
    env_mock.step = MagicMock(return_value=step_data_mock_out)
    env_mock.close = MagicMock()
    env_mock.reset = MagicMock(return_value=brain_info_mock)
    env_mock.global_done = True

    tc.take_step(env_mock, brain_info_mock)
    env_mock.reset.assert_called_once()


def test_take_step_adds_experiences_to_trainer_and_trains():
    tc, trainer_mock = trainer_controller_with_take_step_mocks()

    curr_info_mock = MagicMock()
    trainer_action_output_mock = [
        'action',
        'memory',
        'actiontext',
        'value',
        'output',
    ]
    trainer_mock.take_action = MagicMock(return_value=trainer_action_output_mock)
    trainer_mock.is_ready_update = MagicMock(return_value=True)

    env_mock = MagicMock()
    env_step_output_mock = MagicMock()
    env_mock.step = MagicMock(return_value=env_step_output_mock)
    env_mock.close = MagicMock()
    env_mock.reset = MagicMock(return_value=curr_info_mock)
    env_mock.global_done = False

    tc.take_step(env_mock, curr_info_mock)
    env_mock.reset.assert_not_called()
    trainer_mock.take_action.assert_called_once_with(curr_info_mock)
    env_mock.step.assert_called_once_with(
        vector_action={'testbrain': trainer_action_output_mock[0]},
        memory={'testbrain': trainer_action_output_mock[1]},
        text_action={'testbrain': trainer_action_output_mock[2]},
        value={'testbrain': trainer_action_output_mock[3]}
    )
    trainer_mock.add_experiences.assert_called_once_with(
        curr_info_mock, env_step_output_mock, trainer_action_output_mock[4]
    )
    trainer_mock.process_experiences.assert_called_once_with(curr_info_mock, env_step_output_mock)
    trainer_mock.update_policy.assert_called_once()
    trainer_mock.write_summary.assert_called_once()
    trainer_mock.increment_step_and_update_last_reward.assert_called_once()