# # Unity ML-Agents Toolkit # ## ML-Agent Learning (SAC) # Contains an implementation of SAC as described in https://arxiv.org/abs/1801.01290 # and implemented in https://github.com/hill-a/stable-baselines import logging from collections import defaultdict from typing import List, Dict import os import numpy as np from mlagents.envs.brain import AllBrainInfo from mlagents.envs.action_info import ActionInfoOutputs from mlagents.envs.timers import timed from mlagents.trainers.sac.policy import SACPolicy from mlagents.trainers.rl_trainer import RLTrainer, AllRewardsOutput LOGGER = logging.getLogger("mlagents.trainers") BUFFER_TRUNCATE_PERCENT = 0.8 class SACTrainer(RLTrainer): """ The SACTrainer is an implementation of the SAC algorithm, with support for discrete actions and recurrent networks. """ def __init__( self, brain, reward_buff_cap, trainer_parameters, training, load, seed, run_id ): """ Responsible for collecting experiences and training SAC model. :param trainer_parameters: The parameters for the trainer (dictionary). :param training: Whether the trainer is set for training. :param load: Whether the model should be loaded. :param seed: The seed the model will be initialized with :param run_id: The The identifier of the current run """ super().__init__(brain, trainer_parameters, training, run_id, reward_buff_cap) self.param_keys = [ "batch_size", "buffer_size", "buffer_init_steps", "hidden_units", "learning_rate", "init_entcoef", "max_steps", "normalize", "num_update", "num_layers", "time_horizon", "sequence_length", "summary_freq", "tau", "use_recurrent", "summary_path", "memory_size", "model_path", "reward_signals", "vis_encode_type", ] self.check_param_keys() self.step = 0 self.train_interval = ( trainer_parameters["train_interval"] if "train_interval" in trainer_parameters else 1 ) self.reward_signal_updates_per_train = ( trainer_parameters["reward_signals"]["reward_signal_num_update"] if "reward_signal_num_update" in trainer_parameters["reward_signals"] else trainer_parameters["num_update"] ) self.checkpoint_replay_buffer = ( trainer_parameters["save_replay_buffer"] if "save_replay_buffer" in trainer_parameters else False ) self.policy = SACPolicy(seed, brain, trainer_parameters, self.is_training, load) # Load the replay buffer if load if load and self.checkpoint_replay_buffer: try: self.load_replay_buffer() except (AttributeError, FileNotFoundError): LOGGER.warning( "Replay buffer was unable to load, starting from scratch." ) LOGGER.debug( "Loaded update buffer with {} sequences".format( len(self.training_buffer.update_buffer["actions"]) ) ) for _reward_signal in self.policy.reward_signals.keys(): self.collected_rewards[_reward_signal] = {} self.episode_steps = {} def save_model(self) -> None: """ Saves the model. Overrides the default save_model since we want to save the replay buffer as well. """ self.policy.save_model(self.get_step) if self.checkpoint_replay_buffer: self.save_replay_buffer() def save_replay_buffer(self) -> None: """ Save the training buffer's update buffer to a pickle file. """ filename = os.path.join(self.policy.model_path, "last_replay_buffer.hdf5") LOGGER.info("Saving Experience Replay Buffer to {}".format(filename)) with open(filename, "wb") as file_object: self.training_buffer.update_buffer.save_to_file(file_object) def load_replay_buffer(self) -> None: """ Loads the last saved replay buffer from a file. """ filename = os.path.join(self.policy.model_path, "last_replay_buffer.hdf5") LOGGER.info("Loading Experience Replay Buffer from {}".format(filename)) with open(filename, "rb+") as file_object: self.training_buffer.update_buffer.load_from_file(file_object) LOGGER.info( "Experience replay buffer has {} experiences.".format( len(self.training_buffer.update_buffer["actions"]) ) ) def add_policy_outputs( self, take_action_outputs: ActionInfoOutputs, agent_id: str, agent_idx: int ) -> None: """ Takes the output of the last action and store it into the training buffer. """ actions = take_action_outputs["action"] self.training_buffer[agent_id]["actions"].append(actions[agent_idx]) def add_rewards_outputs( self, rewards_out: AllRewardsOutput, values: Dict[str, np.ndarray], agent_id: str, agent_idx: int, agent_next_idx: int, ) -> None: """ Takes the value output of the last action and store it into the training buffer. """ self.training_buffer[agent_id]["environment_rewards"].append( rewards_out.environment[agent_next_idx] ) def process_experiences( self, current_info: AllBrainInfo, new_info: AllBrainInfo ) -> None: """ Checks agent histories for processing condition, and processes them as necessary. :param current_info: Dictionary of all current brains and corresponding BrainInfo. :param new_info: Dictionary of all next brains and corresponding BrainInfo. """ info = new_info[self.brain_name] if self.is_training: self.policy.update_normalization(info.vector_observations) for l in range(len(info.agents)): agent_actions = self.training_buffer[info.agents[l]]["actions"] if ( info.local_done[l] or len(agent_actions) >= self.trainer_parameters["time_horizon"] ) and len(agent_actions) > 0: agent_id = info.agents[l] # Bootstrap using last brain info. Set last element to duplicate obs and remove dones. if info.max_reached[l]: bootstrapping_info = self.training_buffer[agent_id].last_brain_info idx = bootstrapping_info.agents.index(agent_id) for i, obs in enumerate(bootstrapping_info.visual_observations): self.training_buffer[agent_id]["next_visual_obs%d" % i][ -1 ] = obs[idx] if self.policy.use_vec_obs: self.training_buffer[agent_id]["next_vector_in"][ -1 ] = bootstrapping_info.vector_observations[idx] self.training_buffer[agent_id]["done"][-1] = False self.training_buffer.append_update_buffer( agent_id, batch_size=None, training_length=self.policy.sequence_length, ) self.training_buffer[agent_id].reset_agent() if info.local_done[l]: self.stats["Environment/Episode Length"].append( self.episode_steps.get(agent_id, 0) ) self.episode_steps[agent_id] = 0 for name, rewards in self.collected_rewards.items(): if name == "environment": self.cumulative_returns_since_policy_update.append( rewards.get(agent_id, 0) ) self.stats["Environment/Cumulative Reward"].append( rewards.get(agent_id, 0) ) self.reward_buffer.appendleft(rewards.get(agent_id, 0)) rewards[agent_id] = 0 else: self.stats[ self.policy.reward_signals[name].stat_name ].append(rewards.get(agent_id, 0)) rewards[agent_id] = 0 def is_ready_update(self) -> bool: """ Returns whether or not the trainer has enough elements to run update model :return: A boolean corresponding to whether or not update_model() can be run """ return ( len(self.training_buffer.update_buffer["actions"]) >= self.trainer_parameters["batch_size"] and self.step >= self.trainer_parameters["buffer_init_steps"] ) @timed def update_policy(self) -> None: """ If train_interval is met, update the SAC policy given the current reward signals. If reward_signal_train_interval is met, update the reward signals from the buffer. """ if self.step % self.train_interval == 0: self.trainer_metrics.start_policy_update_timer( number_experiences=len(self.training_buffer.update_buffer["actions"]), mean_return=float(np.mean(self.cumulative_returns_since_policy_update)), ) self.update_sac_policy() self.update_reward_signals() self.trainer_metrics.end_policy_update() def update_sac_policy(self) -> None: """ Uses demonstration_buffer to update the policy. The reward signal generators are updated using different mini batches. If we want to imitate http://arxiv.org/abs/1809.02925 and similar papers, where the policy is updated N times, then the reward signals are updated N times, then reward_signal_updates_per_train is greater than 1 and the reward signals are not updated in parallel. """ self.cumulative_returns_since_policy_update: List[float] = [] n_sequences = max( int(self.trainer_parameters["batch_size"] / self.policy.sequence_length), 1 ) num_updates = self.trainer_parameters["num_update"] batch_update_stats: Dict[str, list] = defaultdict(list) for _ in range(num_updates): LOGGER.debug("Updating SAC policy at step {}".format(self.step)) buffer = self.training_buffer.update_buffer if ( len(self.training_buffer.update_buffer["actions"]) >= self.trainer_parameters["batch_size"] ): sampled_minibatch = buffer.sample_mini_batch( self.trainer_parameters["batch_size"], sequence_length=self.policy.sequence_length, ) # Get rewards for each reward for name, signal in self.policy.reward_signals.items(): sampled_minibatch[ "{}_rewards".format(name) ] = signal.evaluate_batch(sampled_minibatch).scaled_reward update_stats = self.policy.update( sampled_minibatch, n_sequences, update_target=True ) for stat_name, value in update_stats.items(): batch_update_stats[stat_name].append(value) # Truncate update buffer if neccessary. Truncate more than we need to to avoid truncating # a large buffer at each update. if ( len(self.training_buffer.update_buffer["actions"]) > self.trainer_parameters["buffer_size"] ): self.training_buffer.truncate_update_buffer( int(self.trainer_parameters["buffer_size"] * BUFFER_TRUNCATE_PERCENT) ) for stat, stat_list in batch_update_stats.items(): self.stats[stat].append(np.mean(stat_list)) if self.policy.bc_module: update_stats = self.policy.bc_module.update() for stat, val in update_stats.items(): self.stats[stat].append(val) def update_reward_signals(self) -> None: """ Iterate through the reward signals and update them. Unlike in PPO, do it separate from the policy so that it can be done at a different interval. This function should only be used to simulate http://arxiv.org/abs/1809.02925 and similar papers, where the policy is updated N times, then the reward signals are updated N times. Normally, the reward signal and policy are updated in parallel. """ buffer = self.training_buffer.update_buffer num_updates = self.reward_signal_updates_per_train n_sequences = max( int(self.trainer_parameters["batch_size"] / self.policy.sequence_length), 1 ) batch_update_stats: Dict[str, list] = defaultdict(list) for _ in range(num_updates): # Get minibatches for reward signal update if needed reward_signal_minibatches = {} for name, signal in self.policy.reward_signals.items(): LOGGER.debug("Updating {} at step {}".format(name, self.step)) # Some signals don't need a minibatch to be sampled - so we don't! if signal.update_dict: reward_signal_minibatches[name] = buffer.sample_mini_batch( self.trainer_parameters["batch_size"], sequence_length=self.policy.sequence_length, ) update_stats = self.policy.update_reward_signals( reward_signal_minibatches, n_sequences ) for stat_name, value in update_stats.items(): batch_update_stats[stat_name].append(value) for stat, stat_list in batch_update_stats.items(): self.stats[stat].append(np.mean(stat_list))