def parse_args():
parser = yamlargparse.ArgumentParser(description="SVS training")
parser.add_argument(
"-c",
"--config",
help="config file path",
action=yamlargparse.ActionConfigFile,
)
parser.add_argument("--test_align",
help="alignment data dir used for validation.")
parser.add_argument("--test_pitch",
help="pitch data dir used for validation.")
parser.add_argument("--test_wav", help="wave data dir used for validation")
parser.add_argument("--model_file", help="model file for prediction.")
parser.add_argument("--prediction_path",
help="prediction result output (e.g. wav, png).")
parser.add_argument(
"--model_type",
default="GLU_Transformer",
help="Type of model (New_Transformer or GLU_Transformer or LSTM)",
)
parser.add_argument(
"--num_frames",
default=500,
type=int,
help="number of frames in one utterance",
)
parser.add_argument(
"--db_joint",
type=bool,
default=False,
help="Combine multiple datasets & add singer embedding",
)
parser.add_argument(
"--Hz2semitone",
type=bool,
default=False,
help="Transfer f0 value into semitone",
)
parser.add_argument(
"--semitone_size",
type=int,
default=59,
help=
"Semitone size of your dataset, can be found in data/semitone_set.txt",
)
parser.add_argument(
"--semitone_min",
type=str,
default="F_1",
help=
"Minimum semitone of your dataset, can be found in data/semitone_set.txt",
)
parser.add_argument(
"--semitone_max",
type=str,
default="D_6",
help=
"Maximum semitone of your dataset, can be found in data/semitone_set.txt",
)
parser.add_argument("--char_max_len",
default=100,
type=int,
help="max length for character")
parser.add_argument("--num_workers",
default=4,
type=int,
help="number of cpu workers")
parser.add_argument("--decode_sample",
default=-1,
type=int,
help="samples to decode")
parser.add_argument("--frame_length", default=0.06, type=float)
parser.add_argument("--frame_shift", default=0.03, type=float)
parser.add_argument("--sampling_rate", default=44100, type=int)
parser.add_argument("--preemphasis", default=0.97, type=float)
parser.add_argument("--n_mels", default=80, type=int)
parser.add_argument("--power", default=1.2, type=float)
parser.add_argument("--max_db", default=100, type=int)
parser.add_argument("--ref_db", default=20, type=int)
parser.add_argument("--nfft", default=2048, type=int)
parser.add_argument("--phone_size", default=67, type=int)
parser.add_argument("--singer_size", default=10, type=int)
parser.add_argument("--feat_dim", default=1324, type=int)
parser.add_argument("--embedding_size", default=256, type=int)
parser.add_argument("--hidden_size", default=256, type=int)
parser.add_argument("--glu_num_layers",
default=1,
type=int,
help="number of glu layers")
parser.add_argument("--dropout", default=0.1, type=float)
parser.add_argument("--dec_num_block", default=6, type=int)
parser.add_argument("--num_rnn_layers", default=2, type=int)
parser.add_argument("--dec_nhead", default=4, type=int)
parser.add_argument("--local_gaussian", default=False, type=bool)
parser.add_argument("--seed", default=666, type=int)
parser.add_argument(
"--use_tfb",
dest="use_tfboard",
help="whether use tensorboard",
action="store_true",
)
parser.add_argument("--loss", default="l1", type=str)
parser.add_argument("--perceptual_loss", default=-1, type=float)
parser.add_argument("--use_pos_enc", default=0, type=int)
parser.add_argument("--gradient_accumulation_steps", default=1, type=int)
parser.add_argument("--use_asr_post", default=False, type=bool)
parser.add_argument("--sing_quality",
default="conf/sing_quality.csv",
type=str)
parser.add_argument("--standard", default=-1, type=int)
parser.add_argument("--stats_file", default="", type=str)
parser.add_argument("--stats_mel_file", default="", type=str)
parser.add_argument("--collect_stats", default=False, type=bool)
parser.add_argument("--normalize", default=False, type=bool)
parser.add_argument("--num_saved_model", default=5, type=int)
parser.add_argument("--accumulation_steps", default=1, type=int)
parser.add_argument("--auto_select_gpu", default=True, type=bool)
parser.add_argument("--gpu_id", default=1, type=int)
parser.add_argument("--double_mel_loss", default=False, type=float)
parser.add_argument("--vocoder_category", default="griffin", type=str)
args = parser.parse_args()
return args
def get_parser():
"""
Loading parser to parse yaml configurations.
"""
parser = yamlargparse.ArgumentParser(
prog='train_forcast',
description=
'configurations realted to training process of forcasting mechanism')
parser.add_argument(
'--info.run_id',
default='',
help='the unique identifier for logging and metadata creation')
parser.add_argument('--info.m',
default=10,
help='use past m values for prediction')
parser.add_argument('--info.n', default=5, help='predict next n values')
parser.add_argument(
'--info.operation_type',
choices=[const.TRAIN_OP, const.DEPLOY_OP],
help='choosing whether to perform training or deployment')
parser.add_argument(
'--info.model_type',
choices=[
const.LIN_REG, const.RAN_FOR_REG, const.DEC_TREE_REG,
const.MULT_OP_REG
],
help='choosing model type in case of training operation')
parser.add_argument('--info.model_file',
default='',
help='the relative path to the stored model file')
parser.add_argument(
'--info.output_dir',
default='output',
help='the relative path to the directory for storing results')
parser.add_argument(
'--train_test_split.type',
choices=[const.SPLIT_BY_DATE, const.SPLIT_BY_FILES],
help='determines the way in which train-test split should be done')
parser.add_argument(
'--train_test_split.date',
default='',
help=
'the date string in \'YYYY-mm-dd\' format, indicating the date at which split should be made'
)
parser.add_argument(
'--train_test_split.train',
default='',
help='the relative path to the .tsv file containing train data')
parser.add_argument(
'--train_test_split.test',
default='',
help='the relative path to the .tsv file containing test data')
parser.add_argument(
'--visualize.train_data',
action=yamlargparse.ActionYesNo,
default=False,
help='determines if the training visualizations are to be stored')
parser.add_argument(
'--visualize.train_fname',
default='',
help=
'the relative path to the .pdf file storing train data visualizations')
parser.add_argument('--random_forest_regression.max_depth',
default=20,
help='choosing hyperparams for random forest')
parser.add_argument('--random_forest_regression.random_state',
default=7,
help='choosing hyperparams for random forest')
parser.add_argument('--decison_tree_regression.max_depth',
default=20,
help='choosing hyperparams for decision tree')
parser.add_argument('--multi_output_regression.n_estimators',
default=100,
help='choosing hyperparams for multioutput regression')
parser.add_argument('--cfg',
action=yamlargparse.ActionConfigFile,
required=True)
return parser
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
"""
import yamlargparse
import logging
from SVS.model.infer import infer
if __name__ == "__main__":
parser = yamlargparse.ArgumentParser(description="SVS training")
parser.add_argument(
"-c",
"--config",
help="config file path",
action=yamlargparse.ActionConfigFile,
)
parser.add_argument("--test_align",
help="alignment data dir used for validation.")
parser.add_argument("--test_pitch",
help="pitch data dir used for validation.")
parser.add_argument("--test_wav", help="wave data dir used for validation")
parser.add_argument("--model_file", help="model file for prediction.")
parser.add_argument("--prediction_path",
help="prediction result output (e.g. wav, png).")
parser.add_argument(
#!/usr/bin/env python3
#
# Copyright 2019 Hitachi, Ltd. (author: Yusuke Fujita)
# Licensed under the MIT license.
#
import yamlargparse
from eend import system_info
parser = yamlargparse.ArgumentParser(description='decoding')
parser.add_argument('-c',
'--config',
help='config file path',
action=yamlargparse.ActionConfigFile)
parser.add_argument('data_dir', help='kaldi-style data dir')
parser.add_argument('model_file', help='best.nnet')
parser.add_argument('out_dir', help='output directory.')
parser.add_argument('--backend',
default='chainer',
choices=['chainer', 'pytorch'],
help='backend framework')
parser.add_argument('--model_type', default='LSTM', type=str)
parser.add_argument('--gpu', type=int, default=-1)
parser.add_argument('--num-speakers', type=int, default=4)
parser.add_argument('--hidden-size',
default=256,
type=int,
help='number of lstm output nodes')
parser.add_argument('--num-lstm-layers',
default=1,
type=int,
help='number of lstm layers')
#!/usr/bin/env python3
# Copyright 2019 Hitachi, Ltd. (author: Yusuke Fujita)
# Licensed under the MIT license.
#
import yamlargparse
from eend import system_info
parser = yamlargparse.ArgumentParser(description='EEND training')
parser.add_argument('-c',
'--config',
help='config file path',
action=yamlargparse.ActionConfigFile)
parser.add_argument('train_data_dir',
help='kaldi-style data dir used for training.')
parser.add_argument('valid_data_dir',
help='kaldi-style data dir used for validation.')
parser.add_argument('model_save_dir',
help='output directory which model file will be saved in.')
parser.add_argument('--backend',
default='chainer',
choices=['chainer', 'pytorch'],
help='backend framework')
parser.add_argument('--model-type',
default='Transformer',
help='Type of model (Transformer or BLSTM)')
parser.add_argument('--initmodel',
'-m',
default='',
help='Initialize the model from given file')
parser.add_argument('--resume',
#!/usr/bin/env python3
# Copyright 2020 The Johns Hopkins University (author: Jiatong Shi)
import yamlargparse
import sys
sys.path.append(
"/export/c04/jiatong/project/ai_system/CALL-proto/backend/duration_model")
parser = yamlargparse.ArgumentParser(description='Duration training')
parser.add_argument('-c',
'--config',
help='config file path',
action=yamlargparse.ActionConfigFile)
parser.add_argument('--train', help='train data')
parser.add_argument('--val', help="validation data")
parser.add_argument('--model-save-dir',
help='output directory which model file will be saved in.')
parser.add_argument('--model-type',
default='Transformer',
help='Type of model (Transformer or LSTM)')
parser.add_argument('--initmodel',
'-m',
default='',
help='Initialize the model from given file')
parser.add_argument('--resume',
default=False,
type=bool,
help='Resume the optimization from snapshot')