mirror of https://github.com/coqui-ai/TTS.git
update train scripts for coqpit
This commit is contained in:
Eren Gölge
parent
045f1c3e76
commit
78b3825d0b
|
|
@ -15,7 +15,7 @@ from torch.optim import Adam
|
||||||
from torch.utils.data import DataLoader
|
from torch.utils.data import DataLoader
|
||||||
from torch.utils.data.distributed import DistributedSampler
|
from torch.utils.data.distributed import DistributedSampler
|
||||||
|
|
||||||
from TTS.utils.arguments import parse_arguments, process_args
|
from TTS.utils.arguments import init_training
|
||||||
from TTS.utils.audio import AudioProcessor
|
from TTS.utils.audio import AudioProcessor
|
||||||
from TTS.utils.distribute import init_distributed
|
from TTS.utils.distribute import init_distributed
|
||||||
from TTS.utils.generic_utils import KeepAverage, count_parameters, remove_experiment_folder, set_init_dict
|
from TTS.utils.generic_utils import KeepAverage, count_parameters, remove_experiment_folder, set_init_dict
|
||||||
|
|
@ -131,12 +131,12 @@ def train(model, criterion, optimizer, scheduler, scaler, ap, global_step, epoch
|
||||||
if c.mixed_precision:
|
if c.mixed_precision:
|
||||||
scaler.scale(loss).backward()
|
scaler.scale(loss).backward()
|
||||||
scaler.unscale_(optimizer)
|
scaler.unscale_(optimizer)
|
||||||
grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), c.clip_grad)
|
grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), c.grad_clip)
|
||||||
scaler.step(optimizer)
|
scaler.step(optimizer)
|
||||||
scaler.update()
|
scaler.update()
|
||||||
else:
|
else:
|
||||||
loss.backward()
|
loss.backward()
|
||||||
grad_norm = torch.nn.utils.clip_grad_norm_(model.parameters(), c.clip_grad)
|
grad_norm = torch.nn.utils.grad_clip_norm_(model.parameters(), c.clip_grad)
|
||||||
optimizer.step()
|
optimizer.step()
|
||||||
|
|
||||||
# schedule update
|
# schedule update
|
||||||
|
|
@ -311,7 +311,7 @@ def main(args): # pylint: disable=redefined-outer-name
|
||||||
eval_data, train_data = load_wav_data(c.data_path, c.eval_split_size)
|
eval_data, train_data = load_wav_data(c.data_path, c.eval_split_size)
|
||||||
|
|
||||||
# setup audio processor
|
# setup audio processor
|
||||||
ap = AudioProcessor(**c.audio)
|
ap = AudioProcessor(**c.audio.to_dict())
|
||||||
|
|
||||||
# DISTRUBUTED
|
# DISTRUBUTED
|
||||||
if num_gpus > 1:
|
if num_gpus > 1:
|
||||||
|
|
@ -416,9 +416,7 @@ def main(args): # pylint: disable=redefined-outer-name
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
if __name__ == "__main__":
|
||||||
args = parse_arguments(sys.argv)
|
args, c, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = init_training(sys.argv)
|
||||||
c, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = process_args(args, model_class="vocoder")
|
|
||||||
|
|
||||||
try:
|
try:
|
||||||
main(args)
|
main(args)
|
||||||
except KeyboardInterrupt:
|
except KeyboardInterrupt:
|
||||||
|
|
|
||||||
|
|
@ -11,7 +11,7 @@ import torch
|
||||||
from torch.utils.data import DataLoader
|
from torch.utils.data import DataLoader
|
||||||
|
|
||||||
from TTS.tts.utils.visual import plot_spectrogram
|
from TTS.tts.utils.visual import plot_spectrogram
|
||||||
from TTS.utils.arguments import parse_arguments, process_args
|
from TTS.utils.arguments import init_training
|
||||||
from TTS.utils.audio import AudioProcessor
|
from TTS.utils.audio import AudioProcessor
|
||||||
from TTS.utils.generic_utils import KeepAverage, count_parameters, remove_experiment_folder, set_init_dict
|
from TTS.utils.generic_utils import KeepAverage, count_parameters, remove_experiment_folder, set_init_dict
|
||||||
from TTS.utils.radam import RAdam
|
from TTS.utils.radam import RAdam
|
||||||
|
|
@ -307,29 +307,7 @@ def main(args): # pylint: disable=redefined-outer-name
|
||||||
global train_data, eval_data
|
global train_data, eval_data
|
||||||
|
|
||||||
# setup audio processor
|
# setup audio processor
|
||||||
ap = AudioProcessor(**c.audio)
|
ap = AudioProcessor(**c.audio.to_dict())
|
||||||
|
|
||||||
# print(f" > Loading wavs from: {c.data_path}")
|
|
||||||
# if c.feature_path is not None:
|
|
||||||
# print(f" > Loading features from: {c.feature_path}")
|
|
||||||
# eval_data, train_data = load_wav_feat_data(
|
|
||||||
# c.data_path, c.feature_path, c.eval_split_size
|
|
||||||
# )
|
|
||||||
# else:
|
|
||||||
# mel_feat_path = os.path.join(OUT_PATH, "mel")
|
|
||||||
# feat_data = find_feat_files(mel_feat_path)
|
|
||||||
# if feat_data:
|
|
||||||
# print(f" > Loading features from: {mel_feat_path}")
|
|
||||||
# eval_data, train_data = load_wav_feat_data(
|
|
||||||
# c.data_path, mel_feat_path, c.eval_split_size
|
|
||||||
# )
|
|
||||||
# else:
|
|
||||||
# print(" > No feature data found. Preprocessing...")
|
|
||||||
# # preprocessing feature data from given wav files
|
|
||||||
# preprocess_wav_files(OUT_PATH, CONFIG, ap)
|
|
||||||
# eval_data, train_data = load_wav_feat_data(
|
|
||||||
# c.data_path, mel_feat_path, c.eval_split_size
|
|
||||||
# )
|
|
||||||
|
|
||||||
print(f" > Loading wavs from: {c.data_path}")
|
print(f" > Loading wavs from: {c.data_path}")
|
||||||
if c.feature_path is not None:
|
if c.feature_path is not None:
|
||||||
|
|
@ -438,9 +416,7 @@ def main(args): # pylint: disable=redefined-outer-name
|
||||||
|
|
||||||
|
|
||||||
if __name__ == "__main__":
|
if __name__ == "__main__":
|
||||||
args = parse_arguments(sys.argv)
|
args, c, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = init_training(sys.argv)
|
||||||
c, OUT_PATH, AUDIO_PATH, c_logger, tb_logger = process_args(args, model_class="vocoder")
|
|
||||||
|
|
||||||
try:
|
try:
|
||||||
main(args)
|
main(args)
|
||||||
except KeyboardInterrupt:
|
except KeyboardInterrupt:
|
||||||
|
|
|
||||||
|
|
@ -1,164 +0,0 @@
|
||||||
{
|
|
||||||
"run_name": "hifigan",
|
|
||||||
"run_description": "hifigan mean-var scaling",
|
|
||||||
|
|
||||||
// AUDIO PARAMETERS
|
|
||||||
"audio":{
|
|
||||||
"fft_size": 1024, // number of stft frequency levels. Size of the linear spectogram frame.
|
|
||||||
"win_length": 1024, // stft window length in ms.
|
|
||||||
"hop_length": 256, // stft window hop-lengh in ms.
|
|
||||||
"frame_length_ms": null, // stft window length in ms.If null, 'win_length' is used.
|
|
||||||
"frame_shift_ms": null, // stft window hop-lengh in ms. If null, 'hop_length' is used.
|
|
||||||
|
|
||||||
// Audio processing parameters
|
|
||||||
"sample_rate": 22050, // DATASET-RELATED: wav sample-rate. If different than the original data, it is resampled.
|
|
||||||
"preemphasis": 0.0, // pre-emphasis to reduce spec noise and make it more structured. If 0.0, no -pre-emphasis.
|
|
||||||
"ref_level_db": 20, // reference level db, theoretically 20db is the sound of air.
|
|
||||||
"log_func": "np.log10",
|
|
||||||
"do_sound_norm": true,
|
|
||||||
|
|
||||||
// Silence trimming
|
|
||||||
"do_trim_silence": false,// enable trimming of slience of audio as you load it. LJspeech (false), TWEB (false), Nancy (true)
|
|
||||||
"trim_db": 60, // threshold for timming silence. Set this according to your dataset.
|
|
||||||
|
|
||||||
// MelSpectrogram parameters
|
|
||||||
"num_mels": 80, // size of the mel spec frame.
|
|
||||||
"mel_fmin": 50.0, // minimum freq level for mel-spec. ~50 for male and ~95 for female voices. Tune for dataset!!
|
|
||||||
"mel_fmax": 7600.0, // maximum freq level for mel-spec. Tune for dataset!!
|
|
||||||
"spec_gain": 1.0, // scaler value appplied after log transform of spectrogram.
|
|
||||||
|
|
||||||
// Normalization parameters
|
|
||||||
"signal_norm": true, // normalize spec values. Mean-Var normalization if 'stats_path' is defined otherwise range normalization defined by the other params.
|
|
||||||
"min_level_db": -100, // lower bound for normalization
|
|
||||||
"symmetric_norm": true, // move normalization to range [-1, 1]
|
|
||||||
"max_norm": 4.0, // scale normalization to range [-max_norm, max_norm] or [0, max_norm]
|
|
||||||
"clip_norm": true, // clip normalized values into the range.
|
|
||||||
"stats_path": "/home/erogol/.local/share/tts/tts_models--en--ljspeech--speedy-speech-wn/scale_stats.npy"
|
|
||||||
},
|
|
||||||
|
|
||||||
// DISTRIBUTED TRAINING
|
|
||||||
"distributed":{
|
|
||||||
"backend": "nccl",
|
|
||||||
"url": "tcp:\/\/localhost:54324"
|
|
||||||
},
|
|
||||||
|
|
||||||
// MODEL PARAMETERS
|
|
||||||
"use_pqmf": false,
|
|
||||||
|
|
||||||
// LOSS PARAMETERS
|
|
||||||
"use_stft_loss": false,
|
|
||||||
"use_subband_stft_loss": false,
|
|
||||||
"use_mse_gan_loss": true,
|
|
||||||
"use_hinge_gan_loss": false,
|
|
||||||
"use_feat_match_loss": true, // use only with melgan discriminators
|
|
||||||
"use_l1_spec_loss": true,
|
|
||||||
|
|
||||||
// loss weights
|
|
||||||
"stft_loss_weight": 0,
|
|
||||||
"subband_stft_loss_weight": 0,
|
|
||||||
"mse_G_loss_weight": 1,
|
|
||||||
"hinge_G_loss_weight": 0,
|
|
||||||
"feat_match_loss_weight": 10,
|
|
||||||
"l1_spec_loss_weight": 45,
|
|
||||||
|
|
||||||
// multiscale stft loss parameters
|
|
||||||
// "stft_loss_params": {
|
|
||||||
// "n_ffts": [1024, 2048, 512],
|
|
||||||
// "hop_lengths": [120, 240, 50],
|
|
||||||
// "win_lengths": [600, 1200, 240]
|
|
||||||
// },
|
|
||||||
|
|
||||||
"l1_spec_loss_params": {
|
|
||||||
"use_mel": true,
|
|
||||||
"sample_rate": 22050,
|
|
||||||
"n_fft": 1024,
|
|
||||||
"hop_length": 256,
|
|
||||||
"win_length": 1024,
|
|
||||||
"n_mels": 80,
|
|
||||||
"mel_fmin": 0.0,
|
|
||||||
"mel_fmax": null
|
|
||||||
},
|
|
||||||
|
|
||||||
"target_loss": "avg_G_loss", // loss value to pick the best model to save after each epoch
|
|
||||||
|
|
||||||
// DISCRIMINATOR
|
|
||||||
"discriminator_model": "hifigan_discriminator",
|
|
||||||
//"discriminator_model_params":{
|
|
||||||
// "peroids": [2, 3, 5, 7, 11],
|
|
||||||
// "base_channels": 16,
|
|
||||||
// "max_channels":512,
|
|
||||||
// "downsample_factors":[4, 4, 4]
|
|
||||||
//},
|
|
||||||
"steps_to_start_discriminator": 0, // steps required to start GAN trainining.1
|
|
||||||
"diff_samples_for_G_and_D": false, // draw a new sample from the dataset for the D pass.
|
|
||||||
|
|
||||||
// GENERATOR
|
|
||||||
"generator_model": "hifigan_generator",
|
|
||||||
"generator_model_params": {
|
|
||||||
"upsample_factors":[8,8,2,2],
|
|
||||||
"upsample_kernel_sizes": [16,16,4,4],
|
|
||||||
"upsample_initial_channel": 512,
|
|
||||||
"resblock_kernel_sizes": [3,7,11],
|
|
||||||
"resblock_dilation_sizes": [[1,3,5], [1,3,5], [1,3,5]],
|
|
||||||
"resblock_type": "1"
|
|
||||||
},
|
|
||||||
|
|
||||||
// DATASET
|
|
||||||
"data_path": "/home/erogol/gdrive/Datasets/LJSpeech-1.1/wavs/",
|
|
||||||
"feature_path": null,
|
|
||||||
// "feature_path": "/home/erogol/gdrive/Datasets/non-binary-voice-files/tacotron-DCA/",
|
|
||||||
"seq_len": 8192,
|
|
||||||
"pad_short": 2000,
|
|
||||||
"conv_pad": 0,
|
|
||||||
"use_noise_augment": false,
|
|
||||||
"use_cache": true,
|
|
||||||
"reinit_layers": [], // give a list of layer names to restore from the given checkpoint. If not defined, it reloads all heuristically matching layers.
|
|
||||||
|
|
||||||
// TRAINING
|
|
||||||
"batch_size": 16, // Batch size for training. Lower values than 32 might cause hard to learn attention. It is overwritten by 'gradual_training'.
|
|
||||||
|
|
||||||
// VALIDATION
|
|
||||||
"run_eval": true,
|
|
||||||
"test_delay_epochs": 10, //Until attention is aligned, testing only wastes computation time.
|
|
||||||
"test_sentences_file": null, // set a file to load sentences to be used for testing. If it is null then we use default english sentences.
|
|
||||||
|
|
||||||
// OPTIMIZER
|
|
||||||
"epochs": 10000, // total number of epochs to train.
|
|
||||||
"wd": 0.0, // Weight decay weight.
|
|
||||||
"gen_clip_grad": -1, // Generator gradient clipping threshold. Apply gradient clipping if > 0
|
|
||||||
"disc_clip_grad": -1, // Discriminator gradient clipping threshold.
|
|
||||||
"lr_gen": 0.0002, // Initial learning rate. If Noam decay is active, maximum learning rate.
|
|
||||||
"lr_disc": 0.0002,
|
|
||||||
"optimizer": "AdamW",
|
|
||||||
"optimizer_params":{
|
|
||||||
"betas": [0.8, 0.99],
|
|
||||||
"weight_decay": 0.0
|
|
||||||
},
|
|
||||||
"lr_scheduler_gen": "ExponentialLR", // one of the schedulers from https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
|
|
||||||
"lr_scheduler_gen_params": {
|
|
||||||
"gamma": 0.999,
|
|
||||||
"last_epoch": -1
|
|
||||||
},
|
|
||||||
"lr_scheduler_disc": "ExponentialLR", // one of the schedulers from https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate
|
|
||||||
"lr_scheduler_disc_params": {
|
|
||||||
"gamma": 0.999,
|
|
||||||
"last_epoch": -1
|
|
||||||
},
|
|
||||||
|
|
||||||
// TENSORBOARD and LOGGING
|
|
||||||
"print_step": 25, // Number of steps to log traning on console.
|
|
||||||
"print_eval": false, // If True, it prints loss values for each step in eval run.
|
|
||||||
"save_step": 25000, // Number of training steps expected to plot training stats on TB and save model checkpoints.
|
|
||||||
"checkpoint": true, // If true, it saves checkpoints per "save_step"
|
|
||||||
"tb_model_param_stats": true, // true, plots param stats per layer on tensorboard. Might be memory consuming, but good for debugging.
|
|
||||||
|
|
||||||
// DATA LOADING
|
|
||||||
"num_loader_workers": 8, // number of training data loader processes. Don't set it too big. 4-8 are good values.
|
|
||||||
"num_val_loader_workers": 4, // number of evaluation data loader processes.
|
|
||||||
"eval_split_size": 10,
|
|
||||||
|
|
||||||
// PATHS
|
|
||||||
"output_path": "/home/erogol/gdrive/Trainings/LJSpeech/"
|
|
||||||
}
|
|
||||||
|
|
||||||
|
|
||||||
Loading…
Reference in New Issue