def train(trained_models_dir, estimator_model_dir, training_chunk_dir, params): """Train the latest model from gathered data. Args: trained_models_dir: Where to export the completed generation. estimator_model_dir: tf.estimator model directory. training_chunk_dir: Directory where gathered training chunks are. params: An object of hyperparameters for the model. """ model_num, model_name = utils.get_latest_model(trained_models_dir) print('Initializing from model {}'.format(model_name)) new_model_name = utils.generate_model_name(model_num + 1) print('New model will be {}'.format(new_model_name)) save_file = os.path.join(trained_models_dir, new_model_name) tf_records = sorted( tf.gfile.Glob(os.path.join(training_chunk_dir, '*' + _TF_RECORD_SUFFIX))) tf_records = tf_records[-(params.train_window_size // params.examples_per_chunk):] print('Training from: {} to {}'.format(tf_records[0], tf_records[-1])) with utils.logged_timer('Training'): dualnet.train(estimator_model_dir, tf_records, model_num + 1, params) dualnet.export_model(estimator_model_dir, save_file)
def load_model(self): filename = get_latest_model() if filename is None: return False self.saver = tf.train.import_meta_graph(filename) self.saver.restore(self.sess, tf.train.latest_checkpoint('saved_models/')) graph = tf.get_default_graph() variables = graph.get_collection('trainable_variables') #loading weights self.W = [] self.B = [] i = 0 while i < len(variables): self.W.append(variables[i]) i += 1 self.B.append(variables[i]) i += 1 #loading tensors self.X = graph.get_tensor_by_name("X:0") self.Y_ = graph.get_tensor_by_name("Y_:0") self.step = graph.get_tensor_by_name("step:0") self.pkeep = graph.get_tensor_by_name("pkeep:0") self.Y = graph.get_tensor_by_name("Y:0") self.loss = graph.get_tensor_by_name("loss:0") * 100 self.accuracy = graph.get_tensor_by_name("accuracy:0") self.minimize = graph.get_operation_by_name("minimize") return True
def predict(model_name='lgb', pred_start_date='2018-09-01', pred_end_date='2018-09-30', scaler='', use_log=False, num_processes=17, is_train=False): logger.info("开始预测, 当前使用模型:%s" % (model_name)) test_fe_df = feature_engineering(pred_start_date=pred_start_date, pred_end_date=pred_end_date, is_train=False, num_processes=num_processes) logger.info("预测样本数:%s, 分类阈值: %s, 分类每日截断个数:%s" % (test_fe_df.shape[0], CLS_RANKING, NUM_SUBMISSION)) model_save_path = get_latest_model(conf.TRAINED_MODEL_DIR, '%s.model' % model_name) if model_name == 'lgb': submission_df = inference_pipeline_ensemble_tree( test_fe_df, use_log, model_save_path=model_save_path, model_name=model_name) elif model_type == 'stacking': # TODO:增加stacking部分inference pipeline raise NotImplementedError('%s has not been implemented' % model_name) else: # TODO:增加神经网络部分 raise NotImplementedError('%s was not been implemented' % model_name) logger.info("%s预测完成!" % model_name) return submission_df
def __init__(self): model_path = utils.get_latest_model("..", "voxnet") print(model_path) self.model = load_model(model_path) self.model.summary() self.graph = tf.get_default_graph() self.voxel_size_meters = 0.1
def test_get_latest_model(self): with tempfile.TemporaryDirectory() as models_dir: model1 = '000013-model.meta' model2 = '000017-model.meta' f1 = open(os.path.join(models_dir, model1), 'w') f1.close() f2 = open(os.path.join(models_dir, model2), 'w') f2.close() latest_model = utils.get_latest_model(models_dir) self.assertEqual(latest_model, (17, '000017-model'))
def test_get_latest_model(self): with tempfile.TemporaryDirectory() as models_dir: model1 = '000013-model.meta' model2 = '000017-model.meta' f1 = open(os.path.join(models_dir, model1), 'w') f1.close() f2 = open(os.path.join(models_dir, model2), 'w') f2.close() latest_model = utils.get_latest_model(models_dir) self.assertEqual(latest_model, (17, '000017-model'))
def validate(trained_models_dir, holdout_dir, estimator_model_dir, params): """Validate the latest model on the holdout dataset. Args: trained_models_dir: Directories where the completed generations/models are. holdout_dir: Directories where holdout data are. estimator_model_dir: tf.estimator model directory. params: A MiniGoParams instance of hyperparameters for the model. """ model_num, _ = utils.get_latest_model(trained_models_dir) # Get the holdout game data nums_names = utils.get_models(trained_models_dir) # Model N was trained on games up through model N-1, so the validation set # should only be for models through N-1 as well, thus the (model_num) term. models = [num_name for num_name in nums_names if num_name[0] < model_num] # pair is a tuple of (model_num, model_name), like (13, 000013-modelname) holdout_dirs = [ os.path.join(holdout_dir, pair[1]) for pair in models[-params.holdout_generation:] ] tf_records = [] with utils.logged_timer('Building lists of holdout files'): for record_dir in holdout_dirs: if os.path.exists(record_dir): # make sure holdout dir exists tf_records.extend( tf.gfile.Glob( os.path.join(record_dir, '*' + _TF_RECORD_SUFFIX))) if not tf_records: print('No holdout dataset for validation! ' 'Please check your holdout directory: {}'.format(holdout_dir)) return print('The length of tf_records is {}.'.format(len(tf_records))) first_tf_record = os.path.basename(tf_records[0]) last_tf_record = os.path.basename(tf_records[-1]) with utils.logged_timer('Validating from {} to {}'.format( first_tf_record, last_tf_record)): dualnet.validate(estimator_model_dir, tf_records, params)
def validate(trained_models_dir, holdout_dir, estimator_model_dir, params): """Validate the latest model on the holdout dataset. Args: trained_models_dir: Directories where the completed generations/models are. holdout_dir: Directories where holdout data are. estimator_model_dir: tf.estimator model directory. params: A MiniGoParams instance of hyperparameters for the model. """ model_num, _ = utils.get_latest_model(trained_models_dir) # Get the holdout game data nums_names = utils.get_models(trained_models_dir) # Model N was trained on games up through model N-1, so the validation set # should only be for models through N-1 as well, thus the (model_num) term. models = [num_name for num_name in nums_names if num_name[0] < model_num] # pair is a tuple of (model_num, model_name), like (13, 000013-modelname) holdout_dirs = [os.path.join(holdout_dir, pair[1]) for pair in models[-params.holdout_generation:]] tf_records = [] with utils.logged_timer('Building lists of holdout files'): for record_dir in holdout_dirs: if os.path.exists(record_dir): # make sure holdout dir exists tf_records.extend( tf.gfile.Glob(os.path.join(record_dir, '*'+_TF_RECORD_SUFFIX))) if not tf_records: print('No holdout dataset for validation! ' 'Please check your holdout directory: {}'.format(holdout_dir)) return print('The length of tf_records is {}.'.format(len(tf_records))) first_tf_record = os.path.basename(tf_records[0]) last_tf_record = os.path.basename(tf_records[-1]) with utils.logged_timer('Validating from {} to {}'.format( first_tf_record, last_tf_record)): dualnet.validate(estimator_model_dir, tf_records, params)
def main(_): """Run the reinforcement learning loop.""" tf.logging.set_verbosity(tf.logging.INFO) params = _set_params(FLAGS) # A dummy model for debug/testing purpose with fewer games and iterations if FLAGS.test: params = model_params.DummyMiniGoParams() base_dir = FLAGS.base_dir + str(FLAGS.board_size) + '_size_dummy/' else: # Set directories for models and datasets base_dir = FLAGS.base_dir + str(FLAGS.board_size) + '_size/' dirs = utils.MiniGoDirectory(base_dir) # Run selfplay only if user specifies the argument. if FLAGS.selfplay: selfplay_model_name = FLAGS.selfplay_model_name or utils.get_latest_model( dirs.trained_models_dir)[1] max_games = FLAGS.selfplay_max_games or params.max_games_per_generation run_selfplay(selfplay_model_name, max_games, dirs, params) return # Run the RL pipeline # if no models have been trained, start from bootstrap model if not os.path.isdir(dirs.trained_models_dir): print('No trained model exists! Starting from Bootstrap...') print('Creating random initial weights...') bootstrap(dirs.estimator_model_dir, dirs.trained_models_dir, params) else: print('A MiniGo base directory has been found! ') print('Start from the last checkpoint...') _, best_model_so_far = utils.get_latest_model(dirs.trained_models_dir) for rl_iter in range(params.max_iters_per_pipeline): print('RL_iteration: {}'.format(rl_iter)) # Self-play with the best model to generate training data run_selfplay( best_model_so_far, params.max_games_per_generation, dirs, params) # gather selfplay data for training print('Gathering game output...') gather(dirs.selfplay_dir, dirs.training_chunk_dir, params) # train the next generation model model_num, _ = utils.get_latest_model(dirs.trained_models_dir) print('Training on gathered game data...') train(dirs.trained_models_dir, dirs.estimator_model_dir, dirs.training_chunk_dir, model_num + 1, params) # validate the latest model if needed if FLAGS.validation: print('Validating on the holdout game data...') validate(dirs.trained_models_dir, dirs.holdout_dir, dirs.estimator_model_dir, params) _, current_model = utils.get_latest_model(dirs.trained_models_dir) if FLAGS.evaluation: # Perform evaluation if needed print('Evaluate models between {} and {}'.format( best_model_so_far, current_model)) black_model = os.path.join(dirs.trained_models_dir, best_model_so_far) white_model = os.path.join(dirs.trained_models_dir, current_model) _ensure_dir_exists(dirs.evaluate_dir) with utils.logged_timer('Loading weights'): black_net = dualnet.DualNetRunner(black_model, params) white_net = dualnet.DualNetRunner(white_model, params) best_model_so_far = evaluate( best_model_so_far, black_net, current_model, white_net, dirs.evaluate_dir, params) print('Winner of evaluation: {}!'.format(best_model_so_far)) else: best_model_so_far = current_model
def main(_): """Run the reinforcement learning loop.""" tf.logging.set_verbosity(tf.logging.INFO) params = _set_params(FLAGS) # A dummy model for debug/testing purpose with fewer games and iterations if FLAGS.test: params = model_params.DummyMiniGoParams() base_dir = FLAGS.base_dir + str(FLAGS.board_size) + '_size_dummy/' else: # Set directories for models and datasets base_dir = FLAGS.base_dir + str(FLAGS.board_size) + '_size/' dirs = utils.MiniGoDirectory(base_dir) # Run selfplay only if user specifies the argument. if FLAGS.selfplay: selfplay_model_name = FLAGS.selfplay_model_name or utils.get_latest_model( dirs.trained_models_dir)[1] max_games = FLAGS.selfplay_max_games or params.max_games_per_generation run_selfplay(selfplay_model_name, max_games, dirs, params) return # Run the RL pipeline # if no models have been trained, start from bootstrap model if not os.path.isdir(dirs.trained_models_dir): print('No trained model exists! Starting from Bootstrap...') print('Creating random initial weights...') bootstrap(dirs.estimator_model_dir, dirs.trained_models_dir, params) else: print('A MiniGo base directory has been found! ') print('Start from the last checkpoint...') _, best_model_so_far = utils.get_latest_model(dirs.trained_models_dir) for rl_iter in range(params.max_iters_per_pipeline): print('RL_iteration: {}'.format(rl_iter)) # Self-play with the best model to generate training data run_selfplay( best_model_so_far, params.max_games_per_generation, dirs, params) # gather selfplay data for training print('Gathering game output...') gather(dirs.selfplay_dir, dirs.training_chunk_dir, params) # train the next generation model model_num, _ = utils.get_latest_model(dirs.trained_models_dir) print('Training on gathered game data...') train(dirs.trained_models_dir, dirs.estimator_model_dir, dirs.training_chunk_dir, model_num + 1, params) # validate the latest model if needed if FLAGS.validation: print('Validating on the holdout game data...') validate(dirs.trained_models_dir, dirs.holdout_dir, dirs.estimator_model_dir, params) _, current_model = utils.get_latest_model(dirs.trained_models_dir) if FLAGS.evaluation: # Perform evaluation if needed print('Evaluate models between {} and {}'.format( best_model_so_far, current_model)) black_model = os.path.join(dirs.trained_models_dir, best_model_so_far) white_model = os.path.join(dirs.trained_models_dir, current_model) _ensure_dir_exists(dirs.evaluate_dir) with utils.logged_timer('Loading weights'): black_net = dualnet.DualNetRunner(black_model, params) white_net = dualnet.DualNetRunner(white_model, params) best_model_so_far = evaluate( best_model_so_far, black_net, current_model, white_net, dirs.evaluate_dir, params) print('Winner of evaluation: {}!'.format(best_model_so_far)) else: best_model_so_far = current_model
def main(_): """Run the reinforcement learning loop.""" tf.logging.set_verbosity(tf.logging.INFO) params = _set_params_from_board_size(FLAGS.board_size) # A dummy model for debug/testing purpose with fewer games and iterations if FLAGS.debug: params = model_params.DummyMiniGoParams() # Set directories for models and datasets base_dir = FLAGS.base_dir + str(FLAGS.board_size) + '_board_size/' dirs = utils.MiniGoDirectory(base_dir) # if no models have been trained, start from bootstrap model if os.path.isdir(base_dir) is False: print('No trained model exists! Starting from Bootstrap...') print('Creating random initial weights...') bootstrap(dirs.estimator_model_dir, dirs.trained_models_dir, params) else: print('A MiniGo base directory has been found! ') print('Start from the last checkpoint...') _, best_model_so_far = utils.get_latest_model(dirs.trained_models_dir) for rl_iter in range(params.max_iters_per_pipeline): print('RL_iteration: {}'.format(rl_iter)) # Self-play to generate at least params.max_games_per_generation games selfplay(best_model_so_far, dirs.trained_models_dir, dirs.selfplay_dir, dirs.holdout_dir, dirs.sgf_dir, params) games = tf.gfile.Glob( os.path.join(dirs.selfplay_dir, best_model_so_far, '*.zz')) while len(games) < params.max_games_per_generation: selfplay(best_model_so_far, dirs.trained_models_dir, dirs.selfplay_dir, dirs.holdout_dir, dirs.sgf_dir, params) if FLAGS.validation: params = model_params.DummyValidationParams() selfplay(best_model_so_far, dirs.trained_models_dir, dirs.selfplay_dir, dirs.holdout_dir, dirs.sgf_dir, params) games = tf.gfile.Glob( os.path.join(dirs.selfplay_dir, best_model_so_far, '*.zz')) print('Gathering game output...') gather(dirs.selfplay_dir, dirs.training_chunk_dir, params) print('Training on gathered game data...') train(dirs.trained_models_dir, dirs.estimator_model_dir, dirs.training_chunk_dir, params) if FLAGS.validation: print('Validating on the holdout game data...') validate(dirs.trained_models_dir, dirs.holdout_dir, dirs.estimator_model_dir, params) _, current_model = utils.get_latest_model(dirs.trained_models_dir) if FLAGS.evaluation: # Perform evaluation if needed print('Evaluating the latest model...') best_model_so_far = evaluate(dirs.trained_models_dir, best_model_so_far, current_model, dirs.evaluate_dir, params) print('Winner: {}!'.format(best_model_so_far)) else: best_model_so_far = current_model