def main():
logging.debug("Loading configuration file.")
config = json.load(open(args.learning_config, "r"))
logging.debug("Loading Initial/Generative models.")
dbm = DBM(config["layers"], **config["initial_model_args"])
gen_dbm = DBM(config["layers"], **config["generative_model_args"])
logging.debug("Loading learning data.")
learning_data = Data(gen_dbm.sampling(sampling_num=config["data_size"]))
logging.info("Optimizer: %s" % config["optimizer"])
optimizer = getattr(mltools.optimizer,
config["optimizer"])(**config["optimizer_args"])
setting_log = {
"learning_epoch": args.learning_epoch,
"learning_configfile": args.learning_config,
"traindata_size": len(learning_data),
}
setting_log.update(config)
learning_log = LearningLog(setting_log)
logging.info("Train started.")
dbm.train(learning_data,
args.learning_epoch,
gen_dbm,
learning_log,
optimizer,
minibatch_size=config["minibatch_size"],
test_interval=config["test_interval"])
logging.info("Train ended.")
timestamp = time.strftime("%Y-%m-%d_%H-%M-%S")
suffix = "_" + args.filename_suffix if args.filename_suffix is not None else ""
model_filepath = os.path.join(args.output_directory,
"{}_model.json".format(timestamp + suffix))
log_filepath = os.path.join(args.output_directory,
"{}_log.json".format(timestamp + suffix))
dbm.save(model_filepath)
logging.info("Model parameters were dumped to: {}".format(model_filepath))
learning_log.save(log_filepath)
logging.info("Learning log was dumped to: {}".format(log_filepath))
import tensorflow as tf
import datetime
import os
from DRBM import DRBM
from tensorflow.keras.utils import to_categorical
from mltools import LearningLog
parser = argparse.ArgumentParser("DRBM learning script.", add_help=False)
parser.add_argument("learning_config", action="store", type=str, help="path of learning configuration file.")
parser.add_argument("learning_epoch", action="store", type=int, help="numbers of epochs.")
parser.add_argument("-d", "--output_directory", action="store", type=str, default="./results/", help="directory to output parameter & log")
parser.add_argument("-s", "--filename_suffix", action="store", type=str, default=None, help="filename suffix")
args = parser.parse_args()
config = json.load(open(args.learning_config, "r"))
ll = LearningLog(config)
dtype = config["dtype"]
gen_drbm = DRBM(*config["generative-layers"], **config["generative-args"], dtype=dtype, random_bias=True)
x_train, y_train = gen_drbm.stick_break(config["datasize"])
y_train = to_categorical(y_train, dtype=dtype)
train_ds = tf.data.Dataset.from_tensor_slices((x_train, y_train))
optimizer = tf.keras.optimizers.Adamax(learning_rate=0.002, epsilon=1e-8)
drbm = DRBM(*config["training-layers"], **config["training-args"], dtype=dtype)
drbm.fit_generative(args.learning_epoch, config["datasize"], config["minibatch-size"], optimizer, train_ds, gen_drbm, ll)
now = datetime.datetime.now().strftime("%Y-%m-%d_%H-%M-%S")
filename = [
parser.add_argument("-d",
"--output_directory",
action="store",
type=str,
default="./results/",
help="directory to output parameter & log")
parser.add_argument("-s",
"--filename_suffix",
action="store",
type=str,
default=None,
help="filename suffix")
args = parser.parse_args()
config = json.load(open(args.learning_config, "r"))
ll = LearningLog(config)
y, x = np.split(np.loadtxt("./urban.txt", delimiter=","), [1], 1)
x_mean = np.mean(x, 0)
x_std = np.std(x, 0)
x = (x - x_mean) / x_std
y = to_categorical(y)
dtype = config["dtype"]
x = x.astype(dtype)
y = y.astype(dtype)
x_train, x_test, y_train, y_test = train_test_split(x,
y,