def main(unused_argv): tf.compat.v1.logging.set_verbosity(tf.compat.v1.logging.INFO) if FLAGS.mode == 'train': runners.run_train(FLAGS) elif FLAGS.mode == 'eval': runners.run_eval(FLAGS)
def main(unused_argv): del unused_argv logging.set_verbosity(logging.INFO) logging.info("Arguments: {}".format(FLAGS.flag_values_dict())) if FLAGS.mode == 'train': runners.run_train(FLAGS) elif FLAGS.mode == 'eval': runners.run_eval(FLAGS)
def main(unused_argv): tf.logging.set_verbosity(tf.logging.INFO) if FLAGS.data_dimension is None: if FLAGS.dataset_type == "pianoroll": FLAGS.data_dimension = PIANOROLL_DEFAULT_DATA_DIMENSION elif FLAGS.dataset_type == "speech": FLAGS.data_dimension = SPEECH_DEFAULT_DATA_DIMENSION if FLAGS.mode == "train": runners.run_train(FLAGS) elif FLAGS.mode == "eval": runners.run_eval(FLAGS)
LOGPROB_MEAN_MIN = -10.0 LOGPROB_STD_MAX = 5 ## RUN TRAIN #====================================== if config.mode == "train": print(config.trainingset_path) fh = logging.FileHandler( os.path.join(config.logdir, config.log_filename + ".log")) tf.logging.set_verbosity(tf.logging.INFO) # get TF logger logger = logging.getLogger('tensorflow') logger.addHandler(fh) runners.run_train(config) else: with open(config.testset_path, "rb") as f: Vs_test = pickle.load(f) dataset_size = len(Vs_test) ## RUN TASK-SPECIFIC SUBMODEL #====================================== step = None if config.mode in ["save_logprob", "traj_reconstruction"]: tf.Graph().as_default() global_step = tf.train.get_or_create_global_step() inputs, targets, mmsis, time_starts, time_ends, lengths, model = runners.create_dataset_and_model( config, shuffle=False, repeat=False)