def main():
parser = argparse.ArgumentParser()
parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--log_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--input', default='training/train.txt')
parser.add_argument('--variant', default='tacotron')
parser.add_argument('--name', help='Name of the run. Used for logging. Defaults to model name.')
parser.add_argument('--hparams', default='',
help='Hyperparameter overrides as a comma-separated list of name=value pairs')
parser.add_argument('--restore_step', type=int, help='Global step to restore from checkpoint.')
parser.add_argument('--summary_interval', type=int, default=100,
help='Steps between running summary ops.')
parser.add_argument('--checkpoint_interval', type=int, default=1000,
help='Steps between writing checkpoints.')
parser.add_argument('--slack_url', help='Slack webhook URL to get periodic reports.')
parser.add_argument('--tf_log_level', type=int, default=1, help='Tensorflow C++ log level.')
parser.add_argument('--git', action='store_true', help='If set, verify that the client is clean.')
parser.add_argument('--num_steps', type=int, default=100000, help='Maximum number of steps to run training for.')
args = parser.parse_args()
os.environ['TF_CPP_MIN_LOG_LEVEL'] = str(args.tf_log_level)
run_name = args.name or args.variant
log_dir = os.path.join(args.log_dir, 'logs-%s' % run_name)
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'train.log'), run_name, args.slack_url)
log('Initialised log file')
hparams.parse(args.hparams)
train(log_dir, args, args.input)
def setup_log(log_path, checkpoint_path, input_path):
infolog.init(log_path, 'emt4_disc', None)
log('hi')
log('Checkpoint path: {}'.format(checkpoint_path))
log('Loading training data from: {}'.format(input_path))
log('Using model: {}'.format('emt4_disc'))
log(hparams_debug_string())
def prepare_run(args):
modified_hp = hparams.parse(args.hparams)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = str(args.tf_log_level)
run_name = args.name or args.model
log_dir = os.path.join(args.base_dir, 'logs-{}'.format(run_name))
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'), run_name, args.slack_url)
return log_dir, modified_hp
def prepare_run(args):
modified_hp = hparams.parse(args.hparams)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = str(args.tf_log_level)
run_name = args.name or args.model
log_dir = os.path.join(args.base_dir, 'logs-{}'.format(run_name))
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'), run_name)
return log_dir, modified_hp
def prepare_run(args):
modified_hp = hparams.parse(args.hparams)
if args.hparams_json:
import json
with open(args.hparams_json) as hp_json_file:
hp_json = json.dumps(json.load(hp_json_file)['hparams'])
modified_hp = modified_hp.parse_json(hp_json)
os.environ['TF_CPP_MIN_LOG_LEVEL'] = str(args.tf_log_level)
run_name = args.name or args.model
log_dir = os.path.join(args.base_dir, 'logs-{}'.format(run_name))
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'), run_name,
args.slack_url)
return log_dir, modified_hp
def main():
# init log
log_dir = os.path.join('checkpoint', data_configs.speaker)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'),
model_configs.model_name, None)
# train model
if model_configs.model_name == 'TDNN':
model = TDNN(data_configs, model_configs, training_configs)
model.initialize()
model.train()
elif model_configs.model_name == 'TDNN_LSTM':
model = TDNN_LSTM(data_configs, model_configs, training_configs)
model.initialize()
model.train()
def main():
parser = argparse.ArgumentParser()
parser.add_argument('--base_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--log_dir', default=os.path.expanduser('~/tacotron'))
parser.add_argument('--tacotron_input', default='training/train.txt')
parser.add_argument('--wavenet_input',
default='tacotron_output/gta/map.txt')
parser.add_argument('--model', default='tacotron')
parser.add_argument('--variant', default='tacotron')
parser.add_argument(
'--name',
help='Name of the run. Used for logging. Defaults to model name.')
parser.add_argument(
'--hparams',
default='',
help=
'Hyperparameter overrides as a comma-separated list of name=value pairs'
)
parser.add_argument('--restore_step',
type=int,
help='Global step to restore from checkpoint.')
parser.add_argument('--summary_interval',
type=int,
default=100,
help='Steps between running summary ops.')
parser.add_argument('--checkpoint_interval',
type=int,
default=1000,
help='Steps between writing checkpoints.')
parser.add_argument('--slack_url',
help='Slack webhook URL to get periodic reports.')
parser.add_argument('--tf_log_level',
type=int,
default=1,
help='Tensorflow C++ log level.')
parser.add_argument('--git',
action='store_true',
help='If set, verify that the client is clean.')
parser.add_argument('--num_steps',
type=int,
default=100000,
help='Maximum number of steps to run training for.')
parser.add_argument(
'--wavenet_num_steps',
type=int,
default=500000,
help='Maximum number of steps to run wavenet training for')
parser.add_argument('--eal_dir', default='')
parser.add_argument('--eal_ckpt', default='')
parser.add_argument('--eal_ft',
default=False,
action='store_true',
help='load the weights, not Adam / BatchNorm history')
parser.add_argument('--eal_trainAlign',
default=False,
action='store_true',
help='train attention mechanism')
parser.add_argument('--eal_trainJoint',
default=False,
action='store_true',
help='train all weight in EAL mode with joint cost')
parser.add_argument('--eal_alignScale',
type=float,
default=1.0,
help='scaler for loss_align')
args = parser.parse_args()
accepted_models = ['tacotron', 'wavenet']
if args.model not in accepted_models:
raise ValueError(
'please enter a valid model to train: {}'.format(accepted_models))
os.environ['TF_CPP_MIN_LOG_LEVEL'] = str(args.tf_log_level)
run_name = args.name or args.variant
log_dir = os.path.join(args.log_dir, 'logs-%s' % run_name)
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'train.log'), run_name, args.slack_url)
log('Initialised log file')
hparams.parse(args.hparams)
if args.model == 'tacotron':
train_tacotron(log_dir, args, args.tacotron_input)
elif args.model == 'wavenet':
wavenet_train(log_dir, args, hparams, args.wavenet_input)
else:
raise ValueError('Model provided {} unknown! {}'.format(
args.model, accepted_models))
from tensorboardX import SummaryWriter
from matplotlib import cm
from warnings import warn
from wavenet_vocoder.util import is_mulaw_quantize, is_mulaw, is_raw, is_scalar_input
from wavenet_vocoder.mixture import discretized_mix_logistic_loss
from wavenet_vocoder.mixture import sample_from_discretized_mix_logistic, sample_from_gaussian
import audio
from hparams import hparams, hparams_debug_string
import infolog
log_dir = "log_wavenet_single"
run_name = "wavenet_single_gaussian"
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'), run_name)
log = infolog.log
global_step = 0
global_test_step = 0
global_epoch = 0
use_cuda = torch.cuda.is_available()
if use_cuda:
cudnn.benchmark = False
def sanity_check(model, c, g):
if model.has_speaker_embedding():
if g is None:
raise RuntimeError(
"WaveNet expects speaker embedding, but speaker-id is not provided"
every_n_iter=1000)
return tf.estimator.EstimatorSpec(mode=mode,
train_op=train_op,
loss=loss,
training_hooks=[logging_hook])
#load_path = None # 새로운 training
load_path = 'hccho-ckpt\\DCT-2019-08-03_08-59-35'
#####
load_path, restore_path, checkpoint_path = prepare_dirs(hp, load_path)
#### log 파일 작성
log_path = os.path.join(load_path, 'train.log')
infolog.init(log_path, hp.model_name)
infolog.set_tf_log(load_path) # Estimator --> log 저장
tf.logging.set_verbosity(tf.logging.INFO) # 이게 있어야 train log가 출력된다.
# load data
inputs, targets, word_to_index, index_to_word, VOCAB_SIZE, INPUT_LENGTH, OUTPUT_LENGTH = load_data(
hp) # (50000, 29), (50000, 12)
hp.add_hparam('VOCAB_SIZE', VOCAB_SIZE)
hp.add_hparam('INPUT_LENGTH', INPUT_LENGTH) # 29
hp.add_hparam('OUTPUT_LENGTH', OUTPUT_LENGTH) # 11
train_input, test_input, train_target, test_target = train_test_split(
inputs, targets, test_size=0.1, random_state=13371447)
datfeeder = DataFeeder(train_input,
def prepare_run(run_name):
log_dir = os.path.join(dirname(__file__), 'logs-{}'.format(run_name))
os.makedirs(log_dir, exist_ok=True)
infolog.init(os.path.join(log_dir, 'Terminal_train_log'), run_name)
return log_dir
