# # Unity ML-Agents Toolkit # ## ML-Agent Learning """Launches trainers for each External Brains in a Unity Environment.""" import os import sys import json import logging from typing import Dict, List, Optional, Set, NamedTuple from collections import defaultdict import numpy as np from mlagents.tf_utils import tf from time import time from mlagents.trainers.env_manager import EnvManager, EnvironmentStep from mlagents_envs.exception import ( UnityEnvironmentException, UnityCommunicationException, ) from mlagents.trainers.sampler_class import SamplerManager from mlagents_envs.timers import hierarchical_timer, get_timer_tree, timed from mlagents.trainers.trainer import Trainer from mlagents.trainers.meta_curriculum import MetaCurriculum from mlagents.trainers.trainer_util import TrainerFactory from mlagents.trainers.agent_processor import AgentProcessor class AgentManager(NamedTuple): processor: AgentProcessor class TrainerController(object): def __init__( self, trainer_factory: TrainerFactory, model_path: str, summaries_dir: str, run_id: str, save_freq: int, meta_curriculum: Optional[MetaCurriculum], train: bool, training_seed: int, sampler_manager: SamplerManager, resampling_interval: Optional[int], ): """ :param model_path: Path to save the model. :param summaries_dir: Folder to save training summaries. :param run_id: The sub-directory name for model and summary statistics :param save_freq: Frequency at which to save model :param meta_curriculum: MetaCurriculum object which stores information about all curricula. :param train: Whether to train model, or only run inference. :param training_seed: Seed to use for Numpy and Tensorflow random number generation. :param sampler_manager: SamplerManager object handles samplers for resampling the reset parameters. :param resampling_interval: Specifies number of simulation steps after which reset parameters are resampled. """ self.trainers: Dict[str, Trainer] = {} self.brain_name_to_identifier: Dict[str, Set] = defaultdict(set) self.managers: Dict[str, AgentManager] = {} self.trainer_factory = trainer_factory self.model_path = model_path self.summaries_dir = summaries_dir self.logger = logging.getLogger("mlagents.trainers") self.run_id = run_id self.save_freq = save_freq self.train_model = train self.meta_curriculum = meta_curriculum self.training_start_time = time() self.sampler_manager = sampler_manager self.resampling_interval = resampling_interval np.random.seed(training_seed) tf.set_random_seed(training_seed) def _get_measure_vals(self): brain_names_to_measure_vals = {} if self.meta_curriculum: for ( brain_name, curriculum, ) in self.meta_curriculum.brains_to_curriculums.items(): # Skip brains that are in the metacurriculum but no trainer yet. if brain_name not in self.trainers: continue if curriculum.measure == "progress": measure_val = ( self.trainers[brain_name].get_step / self.trainers[brain_name].get_max_steps ) brain_names_to_measure_vals[brain_name] = measure_val elif curriculum.measure == "reward": measure_val = np.mean(self.trainers[brain_name].reward_buffer) brain_names_to_measure_vals[brain_name] = measure_val else: for brain_name, trainer in self.trainers.items(): measure_val = np.mean(trainer.reward_buffer) brain_names_to_measure_vals[brain_name] = measure_val return brain_names_to_measure_vals def _save_model(self): """ Saves current model to checkpoint folder. """ for brain_name in self.trainers.keys(): for name_behavior_id in self.brain_name_to_identifier[brain_name]: self.trainers[brain_name].save_model(name_behavior_id) self.logger.info("Saved Model") def _save_model_when_interrupted(self): self.logger.info( "Learning was interrupted. Please wait while the graph is generated." ) self._save_model() def _write_timing_tree(self) -> None: timing_path = f"{self.summaries_dir}/{self.run_id}_timers.json" try: with open(timing_path, "w") as f: json.dump(get_timer_tree(), f, indent=2) except FileNotFoundError: self.logger.warning( f"Unable to save to {timing_path}. Make sure the directory exists" ) def _export_graph(self): """ Exports latest saved models to .nn format for Unity embedding. """ for brain_name in self.trainers.keys(): for name_behavior_id in self.brain_name_to_identifier[brain_name]: self.trainers[brain_name].export_model(name_behavior_id) @staticmethod def _create_model_path(model_path): try: if not os.path.exists(model_path): os.makedirs(model_path) except Exception: raise UnityEnvironmentException( "The folder {} containing the " "generated model could not be " "accessed. Please make sure the " "permissions are set correctly.".format(model_path) ) def _reset_env(self, env: EnvManager) -> List[EnvironmentStep]: """Resets the environment. Returns: A Data structure corresponding to the initial reset state of the environment. """ sampled_reset_param = self.sampler_manager.sample_all() new_meta_curriculum_config = ( self.meta_curriculum.get_config() if self.meta_curriculum else {} ) sampled_reset_param.update(new_meta_curriculum_config) return env.reset(config=sampled_reset_param) def _should_save_model(self, global_step: int) -> bool: return ( global_step % self.save_freq == 0 and global_step != 0 and self.train_model ) def _not_done_training(self) -> bool: return ( any(t.get_step <= t.get_max_steps for k, t in self.trainers.items()) or not self.train_model ) or len(self.trainers) == 0 def write_to_tensorboard(self, global_step: int) -> None: for brain_name, trainer in self.trainers.items(): # Write training statistics to Tensorboard. delta_train_start = time() - self.training_start_time if ( self.meta_curriculum and brain_name in self.meta_curriculum.brains_to_curriculums ): lesson_num = self.meta_curriculum.brains_to_curriculums[ brain_name ].lesson_num trainer.stats_reporter.add_stat("Environment/Lesson", lesson_num) trainer.write_summary(global_step, delta_train_start) def start_learning(self, env_manager: EnvManager) -> None: self._create_model_path(self.model_path) tf.reset_default_graph() global_step = 0 last_brain_behavior_ids: Set[str] = set() try: self._reset_env(env_manager) while self._not_done_training(): external_brain_behavior_ids = set(env_manager.external_brains.keys()) new_behavior_ids = external_brain_behavior_ids - last_brain_behavior_ids for name_behavior_id in new_behavior_ids: try: brain_name, _ = name_behavior_id.split("?") except ValueError: brain_name = name_behavior_id try: trainer = self.trainers[brain_name] except KeyError: trainer = self.trainer_factory.generate(brain_name) self.trainers[brain_name] = trainer self.logger.info(trainer) if self.train_model: trainer.write_tensorboard_text( "Hyperparameters", trainer.parameters ) policy = trainer.create_policy( env_manager.external_brains[name_behavior_id] ) trainer.add_policy(name_behavior_id, policy) env_manager.set_policy(name_behavior_id, policy) self.brain_name_to_identifier[brain_name].add(name_behavior_id) agent_manager = AgentManager( processor=AgentProcessor( trainer, policy, name_behavior_id, trainer.stats_reporter, trainer.parameters.get("time_horizon", sys.maxsize), ) ) self.managers[name_behavior_id] = agent_manager last_brain_behavior_ids = external_brain_behavior_ids n_steps = self.advance(env_manager) for _ in range(n_steps): global_step += 1 self.reset_env_if_ready(env_manager, global_step) if self._should_save_model(global_step): # Save Tensorflow model self._save_model() self.write_to_tensorboard(global_step) # Final save Tensorflow model if global_step != 0 and self.train_model: self._save_model() except (KeyboardInterrupt, UnityCommunicationException): if self.train_model: self._save_model_when_interrupted() pass if self.train_model: self._export_graph() self._write_timing_tree() def end_trainer_episodes( self, env: EnvManager, lessons_incremented: Dict[str, bool] ) -> None: self._reset_env(env) # Reward buffers reset takes place only for curriculum learning # else no reset. for trainer in self.trainers.values(): trainer.end_episode() for brain_name, changed in lessons_incremented.items(): if changed: self.trainers[brain_name].reward_buffer.clear() def reset_env_if_ready(self, env: EnvManager, steps: int) -> None: if self.meta_curriculum: # Get the sizes of the reward buffers. reward_buff_sizes = { k: len(t.reward_buffer) for (k, t) in self.trainers.items() } # Attempt to increment the lessons of the brains who # were ready. lessons_incremented = self.meta_curriculum.increment_lessons( self._get_measure_vals(), reward_buff_sizes=reward_buff_sizes ) else: lessons_incremented = {} # If any lessons were incremented or the environment is # ready to be reset meta_curriculum_reset = any(lessons_incremented.values()) # Check if we are performing generalization training and we have finished the # specified number of steps for the lesson generalization_reset = ( not self.sampler_manager.is_empty() and (steps != 0) and (self.resampling_interval) and (steps % self.resampling_interval == 0) ) if meta_curriculum_reset or generalization_reset: self.end_trainer_episodes(env, lessons_incremented) @timed def advance(self, env: EnvManager) -> int: with hierarchical_timer("env_step"): new_step_infos = env.step() for step_info in new_step_infos: for name_behavior_id in step_info.name_behavior_ids: if name_behavior_id not in self.managers: self.logger.warning( "Agent manager was not created for behavior id {}.".format( name_behavior_id ) ) continue _processor = self.managers[name_behavior_id].processor _processor.add_experiences( step_info.previous_all_brain_info[name_behavior_id], step_info.current_all_brain_info[name_behavior_id], step_info.brain_name_to_action_info[name_behavior_id].outputs, ) for brain_name, trainer in self.trainers.items(): if self.train_model and trainer.get_step <= trainer.get_max_steps: n_steps = len(new_step_infos) trainer.increment_step(n_steps) for name_behavior_id in self.brain_name_to_identifier[brain_name]: trainer.get_policy(name_behavior_id).increment_step(n_steps) if trainer.is_ready_update(): # Perform gradient descent with experience buffer with hierarchical_timer("update_policy"): trainer.update_policy() for name_behavior_id in self.brain_name_to_identifier[brain_name]: env.set_policy( name_behavior_id, trainer.get_policy(name_behavior_id) ) else: # Avoid memory leak during inference # Eventually this whole block will take place in advance() # But currently this only calls clear_update_buffer() in RLTrainer # and nothing in the base class trainer.advance() return len(new_step_infos)