mirror of https://github.com/coqui-ai/TTS.git
add Coqpit configs for the TTS models
This commit is contained in:
Eren Gölge
parent
7227e8f1d2
commit
7663bc63c1
|
|
@ -0,0 +1,17 @@
|
|||
import importlib
|
||||
import os
|
||||
from inspect import isclass
|
||||
|
||||
# import all files under configs/
|
||||
configs_dir = os.path.dirname(__file__)
|
||||
for file in os.listdir(configs_dir):
|
||||
path = os.path.join(configs_dir, file)
|
||||
if not file.startswith("_") and not file.startswith(".") and (file.endswith(".py") or os.path.isdir(path)):
|
||||
config_name = file[: file.find(".py")] if file.endswith(".py") else file
|
||||
module = importlib.import_module("TTS.tts.configs." + config_name)
|
||||
for attribute_name in dir(module):
|
||||
attribute = getattr(module, attribute_name)
|
||||
|
||||
if isclass(attribute):
|
||||
# Add the class to this package's variables
|
||||
globals()[attribute_name] = attribute
|
||||
|
|
@ -1,126 +0,0 @@
|
|||
{
|
||||
"attention_heads": 4,
|
||||
"attention_norm": "sigmoid",
|
||||
"attention_type": "original",
|
||||
"audio_config": {
|
||||
"clip_norm": true,
|
||||
"do_trim_silence": true,
|
||||
"fft_size": 1024,
|
||||
"frame_length_ms": null,
|
||||
"frame_shift_ms": null,
|
||||
"griffin_lim_iters": 60,
|
||||
"hop_length": 256,
|
||||
"max_norm": 4,
|
||||
"mel_fmax": 7600,
|
||||
"mel_fmin": 50,
|
||||
"min_level_db": -100,
|
||||
"num_mels": 80,
|
||||
"power": 1.5,
|
||||
"preemphasis": 0,
|
||||
"ref_level_db": 20,
|
||||
"sample_rate": 22050,
|
||||
"signal_norm": true,
|
||||
"spec_gain": 1,
|
||||
"stats_path": "/home/erogol/Data/LJSpeech-1.1/scale_stats.npy",
|
||||
"symmetric_norm": true,
|
||||
"trim_db": 60,
|
||||
"win_length": 1024
|
||||
},
|
||||
"bidirectional_decoder": false,
|
||||
"compute_input_seq_cache": false,
|
||||
"ddc_r": 7,
|
||||
"decoder_diff_spec_alpha": 0.25,
|
||||
"decoder_loss_alpha": 0.5,
|
||||
"decoder_ssim_alpha": 0.5,
|
||||
"double_decoder_consistency": true,
|
||||
"enable_eos_bos_chars": false,
|
||||
"forward_attn_mask": false,
|
||||
"ga_alpha": 5,
|
||||
"grad_clip": 1,
|
||||
"gradual_training": [
|
||||
[
|
||||
0,
|
||||
7,
|
||||
64
|
||||
],
|
||||
[
|
||||
1,
|
||||
5,
|
||||
64
|
||||
],
|
||||
[
|
||||
50000,
|
||||
3,
|
||||
32
|
||||
],
|
||||
[
|
||||
130000,
|
||||
2,
|
||||
32
|
||||
],
|
||||
[
|
||||
290000,
|
||||
1,
|
||||
32
|
||||
]
|
||||
],
|
||||
"location_attn": true,
|
||||
"lr": 0.0001,
|
||||
"memory_size": -1,
|
||||
"noam_schedule": false,
|
||||
"phoneme_cache_path": "/home/erogol/Models/phoneme_cache/",
|
||||
"phoneme_language": "en-us",
|
||||
"postnet_diff_spec_alpha": 0.25,
|
||||
"postnet_loss_alpha": 0.25,
|
||||
"postnet_ssim_alpha": 0.25,
|
||||
"prenet_dropout": false,
|
||||
"prenet_type": "original",
|
||||
"r": 7,
|
||||
"separate_stopnet": true,
|
||||
"seq_len_norm": false,
|
||||
"stopnet": true,
|
||||
"stopnet_pos_weight": 15,
|
||||
"test_sentences_file": null,
|
||||
"text_cleaner": "phoneme_cleaners",
|
||||
"training_config": {
|
||||
"batch_group_size": 4,
|
||||
"batch_size": 32,
|
||||
"checkpoint": true,
|
||||
"datasets": [
|
||||
{
|
||||
"meta_file_train": "metadata.csv",
|
||||
"meta_file_val": null,
|
||||
"name": "ljspeech",
|
||||
"path": "/home/erogol/Data/LJSpeech-1.1/"
|
||||
}
|
||||
],
|
||||
"epochs": 1000,
|
||||
"eval_batch_size": 16,
|
||||
"keep_after": 10000,
|
||||
"keep_all_best": false,
|
||||
"loss_masking": true,
|
||||
"max_seq_len": 153,
|
||||
"min_seq_len": 6,
|
||||
"mixed_precision": true,
|
||||
"model": "Tacotron2",
|
||||
"num_loader_workers": 4,
|
||||
"num_val_loader_workers": 4,
|
||||
"output_path": "/home/erogol/Models/LJSpeech/",
|
||||
"print_eval": false,
|
||||
"print_step": 25,
|
||||
"run_description": "tacotron2 with DDC and differential spectral loss.",
|
||||
"run_eval": true,
|
||||
"run_name": "ljspeech-ddc",
|
||||
"save_step": 10000,
|
||||
"tb_model_param_stats": false,
|
||||
"tb_plot_step": 100,
|
||||
"test_delay_epochs": 10,
|
||||
"use_noise_augment": true
|
||||
},
|
||||
"transition_agent": false,
|
||||
"use_forward_attn": false,
|
||||
"use_phonemes": true,
|
||||
"warmup_steps": 4000,
|
||||
"wd": 0.000001,
|
||||
"windowing": false
|
||||
}
|
||||
|
|
@ -1,138 +0,0 @@
|
|||
{
|
||||
"model": "glow_tts",
|
||||
"run_name": "glow-tts-gatedconv",
|
||||
"run_description": "glow-tts model training with gated conv.",
|
||||
|
||||
// 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": 0, // reference level db, theoretically 20db is the sound of air.
|
||||
|
||||
// Griffin-Lim
|
||||
"power": 1.1, // value to sharpen wav signals after GL algorithm.
|
||||
"griffin_lim_iters": 60,// #griffin-lim iterations. 30-60 is a good range. Larger the value, slower the generation.
|
||||
|
||||
// Silence trimming
|
||||
"do_trim_silence": true,// 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": 1.0, // 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/Data/LJSpeech-1.1/scale_stats.npy" // DO NOT USE WITH MULTI_SPEAKER MODEL. scaler stats file computed by 'compute_statistics.py'. If it is defined, mean-std based notmalization is used and other normalization params are ignored
|
||||
},
|
||||
|
||||
// VOCABULARY PARAMETERS
|
||||
// if custom character set is not defined,
|
||||
// default set in symbols.py is used
|
||||
// "characters":{
|
||||
// "pad": "_",
|
||||
// "eos": "~",
|
||||
// "bos": "^",
|
||||
// "characters": "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz!'(),-.:;? ",
|
||||
// "punctuations":"!'(),-.:;? ",
|
||||
// "phonemes":"iyɨʉɯuɪʏʊeøɘəɵɤoɛœɜɞʌɔæɐaɶɑɒᵻʘɓǀɗǃʄǂɠǁʛpbtdʈɖcɟkɡqɢʔɴŋɲɳnɱmʙrʀⱱɾɽɸβfvθðszʃʒʂʐçʝxɣχʁħʕhɦɬɮʋɹɻjɰlɭʎʟˈˌːˑʍwɥʜʢʡɕʑɺɧɚ˞ɫ"
|
||||
// },
|
||||
|
||||
"add_blank": false, // if true add a new token after each token of the sentence. This increases the size of the input sequence, but has considerably improved the prosody of the GlowTTS model.
|
||||
|
||||
// DISTRIBUTED TRAINING
|
||||
"apex_amp_level": null, // APEX amp optimization level. "O1" is currently supported.
|
||||
"distributed":{
|
||||
"backend": "nccl",
|
||||
"url": "tcp:\/\/localhost:54323"
|
||||
},
|
||||
|
||||
"reinit_layers": [], // give a list of layer names to restore from the given checkpoint. If not defined, it reloads all heuristically matching layers.
|
||||
|
||||
// MODEL PARAMETERS
|
||||
"use_mas": false, // use Monotonic Alignment Search if true. Otherwise use pre-computed attention alignments.
|
||||
|
||||
// TRAINING
|
||||
"batch_size": 32, // Batch size for training. Lower values than 32 might cause hard to learn attention. It is overwritten by 'gradual_training'.
|
||||
"eval_batch_size":16,
|
||||
"r": 1, // Number of decoder frames to predict per iteration. Set the initial values if gradual training is enabled.
|
||||
"loss_masking": true, // enable / disable loss masking against the sequence padding.
|
||||
|
||||
// VALIDATION
|
||||
"run_eval": true,
|
||||
"test_delay_epochs": 0, //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
|
||||
"noam_schedule": true, // use noam warmup and lr schedule.
|
||||
"grad_clip": 5.0, // upper limit for gradients for clipping.
|
||||
"epochs": 10000, // total number of epochs to train.
|
||||
"lr": 1e-3, // Initial learning rate. If Noam decay is active, maximum learning rate.
|
||||
"wd": 0.000001, // Weight decay weight.
|
||||
"warmup_steps": 4000, // Noam decay steps to increase the learning rate from 0 to "lr"
|
||||
"seq_len_norm": false, // Normalize eash sample loss with its length to alleviate imbalanced datasets. Use it if your dataset is small or has skewed distribution of sequence lengths.
|
||||
|
||||
"encoder_type": "gatedconv",
|
||||
|
||||
// TENSORBOARD and LOGGING
|
||||
"print_step": 25, // Number of steps to log training on console.
|
||||
"tb_plot_step": 100, // Number of steps to plot TB training figures.
|
||||
"print_eval": false, // If True, it prints intermediate loss values in evalulation.
|
||||
"save_step": 5000, // Number of training steps expected to save traninpg stats and checkpoints.
|
||||
"checkpoint": true, // If true, it saves checkpoints per "save_step"
|
||||
"keep_all_best": false, // If true, keeps all best_models after keep_after steps
|
||||
"keep_after": 10000, // Global step after which to keep best models if keep_all_best is true
|
||||
"tb_model_param_stats": false, // true, plots param stats per layer on tensorboard. Might be memory consuming, but good for debugging.
|
||||
"apex_amp_level": null,
|
||||
|
||||
// DATA LOADING
|
||||
"text_cleaner": "phoneme_cleaners",
|
||||
"enable_eos_bos_chars": false, // enable/disable beginning of sentence and end of sentence chars.
|
||||
"num_loader_workers": 4, // 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.
|
||||
"batch_group_size": 0, //Number of batches to shuffle after bucketing.
|
||||
"min_seq_len": 3, // DATASET-RELATED: minimum text length to use in training
|
||||
"max_seq_len": 500, // DATASET-RELATED: maximum text length
|
||||
"compute_f0": false, // compute f0 values in data-loader
|
||||
"compute_input_seq_cache": false, // if true, text sequences are computed before starting training. If phonemes are enabled, they are also computed at this stage.
|
||||
|
||||
// PATHS
|
||||
"output_path": "/home/erogol/Models/LJSpeech/",
|
||||
|
||||
// PHONEMES
|
||||
"phoneme_cache_path": "/home/erogol/Models/phoneme_cache/", // phoneme computation is slow, therefore, it caches results in the given folder.
|
||||
"use_phonemes": true, // use phonemes instead of raw characters. It is suggested for better pronounciation.
|
||||
"phoneme_language": "en-us", // depending on your target language, pick one from https://github.com/bootphon/phonemizer#languages
|
||||
|
||||
// MULTI-SPEAKER and GST
|
||||
"use_speaker_embedding": false, // use speaker embedding to enable multi-speaker learning.
|
||||
"style_wav_for_test": null, // path to style wav file to be used in TacotronGST inference.
|
||||
"use_gst": false, // TACOTRON ONLY: use global style tokens
|
||||
|
||||
// DATASETS
|
||||
"datasets": // List of datasets. They all merged and they get different speaker_ids.
|
||||
[
|
||||
{
|
||||
"name": "ljspeech",
|
||||
"path": "/home/erogol/Data/LJSpeech-1.1/",
|
||||
"meta_file_train": "metadata.csv",
|
||||
"meta_file_val": null
|
||||
// "path_for_attn": "/home/erogol/Data/LJSpeech-1.1/alignments/"
|
||||
}
|
||||
]
|
||||
}
|
||||
|
||||
|
||||
|
|
@ -1,151 +0,0 @@
|
|||
{
|
||||
"model": "glow_tts",
|
||||
"run_name": "glow-tts-residual_bn_conv",
|
||||
"run_description": "glow-tts model training with residual BN conv.",
|
||||
|
||||
// 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": 0, // reference level db, theoretically 20db is the sound of air.
|
||||
|
||||
// Griffin-Lim
|
||||
"power": 1.1, // value to sharpen wav signals after GL algorithm.
|
||||
"griffin_lim_iters": 60,// #griffin-lim iterations. 30-60 is a good range. Larger the value, slower the generation.
|
||||
|
||||
// Silence trimming
|
||||
"do_trim_silence": true,// 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.00
|
||||
|
||||
// Normalization parameters
|
||||
"signal_norm": false, // 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": 1.0, // scale normalization to range [-max_norm, max_norm] or [0, max_norm]
|
||||
"clip_norm": true, // clip normalized values into the range.
|
||||
"stats_path": null // DO NOT USE WITH MULTI_SPEAKER MODEL. scaler stats file computed by 'compute_statistics.py'. If it is defined, mean-std based notmalization is used and other normalization params are ignored
|
||||
},
|
||||
|
||||
// VOCABULARY PARAMETERS
|
||||
// if custom character set is not defined,
|
||||
// default set in symbols.py is used
|
||||
// "characters":{
|
||||
// "pad": "_",
|
||||
// "eos": "~",
|
||||
// "bos": "^",
|
||||
// "characters": "ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz!'(),-.:;? ",
|
||||
// "punctuations":"!'(),-.:;? ",
|
||||
// "phonemes":"iyɨʉɯuɪʏʊeøɘəɵɤoɛœɜɞʌɔæɐaɶɑɒᵻʘɓǀɗǃʄǂɠǁʛpbtdʈɖcɟkɡqɢʔɴŋɲɳnɱmʙrʀⱱɾɽɸβfvθðszʃʒʂʐçʝxɣχʁħʕhɦɬɮʋɹɻjɰlɭʎʟˈˌːˑʍwɥʜʢʡɕʑɺɧɚ˞ɫ"
|
||||
// },
|
||||
|
||||
"add_blank": false, // if true add a new token after each token of the sentence. This increases the size of the input sequence, but has considerably improved the prosody of the GlowTTS model.
|
||||
|
||||
// DISTRIBUTED TRAINING
|
||||
"distributed":{
|
||||
"backend": "nccl",
|
||||
"url": "tcp:\/\/localhost:54321"
|
||||
},
|
||||
|
||||
"reinit_layers": [], // give a list of layer names to restore from the given checkpoint. If not defined, it reloads all heuristically matching layers.
|
||||
|
||||
// MODEL PARAMETERS
|
||||
// "use_mas": false, // use Monotonic Alignment Search if true. Otherwise use pre-computed attention alignments.
|
||||
"hidden_channels_encoder": 192,
|
||||
"hidden_channels_decoder": 192,
|
||||
"hidden_channels_duration_predictor": 256,
|
||||
"use_encoder_prenet": true,
|
||||
"encoder_type": "rel_pos_transformer",
|
||||
"encoder_params": {
|
||||
"kernel_size":3,
|
||||
"dropout_p": 0.1,
|
||||
"num_layers": 6,
|
||||
"num_heads": 2,
|
||||
"hidden_channels_ffn": 768,
|
||||
"input_length": null
|
||||
},
|
||||
|
||||
// TRAINING
|
||||
"batch_size": 32, // Batch size for training. Lower values than 32 might cause hard to learn attention. It is overwritten by 'gradual_training'.
|
||||
"eval_batch_size":16,
|
||||
"r": 1, // Number of decoder frames to predict per iteration. Set the initial values if gradual training is enabled.
|
||||
"loss_masking": true, // enable / disable loss masking against the sequence padding.
|
||||
"mixed_precision": true,
|
||||
"data_dep_init_iter": 10,
|
||||
|
||||
// VALIDATION
|
||||
"run_eval": true,
|
||||
"test_delay_epochs": 0, //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
|
||||
"noam_schedule": true, // use noam warmup and lr schedule.
|
||||
"grad_clip": 5.0, // upper limit for gradients for clipping.
|
||||
"epochs": 10000, // total number of epochs to train.
|
||||
"lr": 1e-3, // Initial learning rate. If Noam decay is active, maximum learning rate.
|
||||
"wd": 0.000001, // Weight decay weight.
|
||||
"warmup_steps": 4000, // Noam decay steps to increase the learning rate from 0 to "lr"
|
||||
"seq_len_norm": false, // Normalize eash sample loss with its length to alleviate imbalanced datasets. Use it if your dataset is small or has skewed distribution of sequence lengths.
|
||||
|
||||
// TENSORBOARD and LOGGING
|
||||
"print_step": 25, // Number of steps to log training on console.
|
||||
"tb_plot_step": 100, // Number of steps to plot TB training figures.
|
||||
"print_eval": false, // If True, it prints intermediate loss values in evalulation.
|
||||
"save_step": 5000, // Number of training steps expected to save traninpg stats and checkpoints.
|
||||
"checkpoint": true, // If true, it saves checkpoints per "save_step"
|
||||
"keep_all_best": false, // If true, keeps all best_models after keep_after steps
|
||||
"keep_after": 10000, // Global step after which to keep best models if keep_all_best is true
|
||||
"tb_model_param_stats": false, // true, plots param stats per layer on tensorboard. Might be memory consuming, but good for debugging.
|
||||
|
||||
// DATA LOADING
|
||||
"text_cleaner": "phoneme_cleaners",
|
||||
"enable_eos_bos_chars": false, // enable/disable beginning of sentence and end of sentence chars.
|
||||
"num_loader_workers": 4, // 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.
|
||||
"batch_group_size": 0, //Number of batches to shuffle after bucketing.
|
||||
"min_seq_len": 3, // DATASET-RELATED: minimum text length to use in training
|
||||
"max_seq_len": 500, // DATASET-RELATED: maximum text length
|
||||
"compute_f0": false, // compute f0 values in data-loader
|
||||
"use_noise_augment": true, //add a random noise to audio signal for augmentation at training .
|
||||
"compute_input_seq_cache": true,
|
||||
|
||||
// PATHS
|
||||
"output_path": "/home/erogol/Models/LJSpeech/",
|
||||
|
||||
// PHONEMES
|
||||
"phoneme_cache_path": "/home/erogol/Models/phoneme_cache/", // phoneme computation is slow, therefore, it caches results in the given folder.
|
||||
"use_phonemes": true, // use phonemes instead of raw characters. It is suggested for better pronounciation.
|
||||
"phoneme_language": "en-us", // depending on your target language, pick one from https://github.com/bootphon/phonemizer#languages
|
||||
|
||||
// MULTI-SPEAKER and GST
|
||||
"use_speaker_embedding": false, // use speaker embedding to enable multi-speaker learning.
|
||||
"use_external_speaker_embedding_file": false,
|
||||
"style_wav_for_test": null, // path to style wav file to be used in TacotronGST inference.
|
||||
"use_gst": false, // TACOTRON ONLY: use global style tokens
|
||||
|
||||
// DATASETS
|
||||
"datasets": // List of datasets. They all merged and they get different speaker_ids.
|
||||
[
|
||||
{
|
||||
"name": "ljspeech",
|
||||
"path": "/home/erogol/Data/LJSpeech-1.1/",
|
||||
"meta_file_train": "metadata.csv",
|
||||
"meta_file_val": null
|
||||
// "path_for_attn": "/home/erogol/Data/LJSpeech-1.1/alignments/"
|
||||
}
|
||||
]
|
||||
}
|
||||
|
||||
|
||||
|
|
@ -0,0 +1,83 @@
|
|||
from dataclasses import asdict, dataclass, field
|
||||
from typing import List
|
||||
|
||||
from coqpit import MISSING, Coqpit, check_argument
|
||||
|
||||
from TTS.config import BaseAudioConfig, BaseDatasetConfig, BaseTrainingConfig
|
||||
|
||||
|
||||
@dataclass
|
||||
class GSTConfig(Coqpit):
|
||||
"""Defines Global Style Toke module"""
|
||||
|
||||
gst_style_input_wav: str = None
|
||||
gst_style_input_weights: dict = None
|
||||
gst_embedding_dim: int = 256
|
||||
gst_use_speaker_embedding: bool = False
|
||||
gst_num_heads: int = 4
|
||||
gst_num_style_tokens: int = 10
|
||||
|
||||
def check_values(
|
||||
self,
|
||||
):
|
||||
"""Check config fields"""
|
||||
c = asdict(self)
|
||||
super().check_values()
|
||||
check_argument("gst_style_input_weights", c, restricted=False)
|
||||
check_argument("gst_style_input_wav", c, restricted=False)
|
||||
check_argument("gst_embedding_dim", c, restricted=True, min_val=0, max_val=1000)
|
||||
check_argument("gst_use_speaker_embedding", c, restricted=False)
|
||||
check_argument("gst_num_heads", c, restricted=True, min_val=2, max_val=10)
|
||||
check_argument("gst_num_style_tokens", c, restricted=True, min_val=1, max_val=1000)
|
||||
|
||||
|
||||
@dataclass
|
||||
class CharactersConfig:
|
||||
"""Defines character or phoneme set used by the model"""
|
||||
|
||||
pad: str = None
|
||||
eos: str = None
|
||||
bos: str = None
|
||||
characters: str = None
|
||||
punctuations: str = None
|
||||
phonemes: str = None
|
||||
|
||||
def check_values(
|
||||
self,
|
||||
):
|
||||
"""Check config fields"""
|
||||
c = asdict(self)
|
||||
check_argument("pad", c, "characters", restricted=True)
|
||||
check_argument("eos", c, "characters", restricted=True)
|
||||
check_argument("bos", c, "characters", restricted=True)
|
||||
check_argument("characters", c, "characters", restricted=True)
|
||||
check_argument("phonemes", c, restricted=True)
|
||||
check_argument("punctuations", c, "characters", restricted=True)
|
||||
|
||||
|
||||
@dataclass
|
||||
class BaseTTSConfig(BaseTrainingConfig):
|
||||
"""Shared parameters among all the tts models."""
|
||||
|
||||
audio: BaseAudioConfig = field(default_factory=BaseAudioConfig)
|
||||
# phoneme settings
|
||||
use_phonemes: bool = False
|
||||
phoneme_language: str = None
|
||||
compute_input_seq_cache: bool = False
|
||||
text_cleaner: str = MISSING
|
||||
enable_eos_bos_chars: bool = False
|
||||
test_sentences_file: str = ""
|
||||
phoneme_cache_path: str = None
|
||||
# vocabulary parameters
|
||||
characters: CharactersConfig = None
|
||||
# training params
|
||||
batch_group_size: int = 0
|
||||
loss_masking: bool = None
|
||||
# dataloading
|
||||
min_seq_len: int = 1
|
||||
max_seq_len: int = float("inf")
|
||||
compute_f0: bool = False
|
||||
use_noise_augment: bool = False
|
||||
add_blank: bool = False
|
||||
# dataset
|
||||
datasets: List[BaseDatasetConfig] = field(default_factory=lambda: [BaseDatasetConfig()])
|
||||
|
|
@ -0,0 +1,53 @@
|
|||
from dataclasses import dataclass, field
|
||||
|
||||
from .shared_configs import BaseTTSConfig
|
||||
|
||||
|
||||
@dataclass
|
||||
class SpeedySpeechConfig(BaseTTSConfig):
|
||||
"""Defines parameters for Speedy Speech (feed-forward encoder-decoder) based models."""
|
||||
|
||||
model: str = "speedy_speech"
|
||||
# model specific params
|
||||
positional_encoding: bool = True
|
||||
hidden_channels: int = 128
|
||||
encoder_type: str = "residual_conv_bn"
|
||||
encoder_params: dict = field(
|
||||
default_factory=lambda: {
|
||||
"kernel_size": 4,
|
||||
"dilations": [1, 2, 4, 1, 2, 4, 1, 2, 4, 1, 2, 4, 1],
|
||||
"num_conv_blocks": 2,
|
||||
"num_res_blocks": 13,
|
||||
}
|
||||
)
|
||||
decoder_type: str = "residual_conv_bn"
|
||||
decoder_params: dict = field(
|
||||
default_factory=lambda: {
|
||||
"kernel_size": 4,
|
||||
"dilations": [1, 2, 4, 8, 1, 2, 4, 8, 1, 2, 4, 8, 1, 2, 4, 8, 1],
|
||||
"num_conv_blocks": 2,
|
||||
"num_res_blocks": 17,
|
||||
}
|
||||
)
|
||||
|
||||
# multi-speaker settings
|
||||
use_speaker_embedding: bool = False
|
||||
use_external_speaker_embedding_file: bool = False
|
||||
external_speaker_embedding_file: str = False
|
||||
|
||||
# optimizer parameters
|
||||
noam_schedule: bool = False
|
||||
warmup_steps: int = 4000
|
||||
lr: float = 1e-4
|
||||
wd: float = 1e-6
|
||||
grad_clip: float = 5.0
|
||||
|
||||
# loss params
|
||||
ssim_alpha: float = 1.0
|
||||
huber_alpha: float = 1.0
|
||||
l1_alpha: float = 1.0
|
||||
|
||||
# overrides
|
||||
min_seq_len: int = 13
|
||||
max_seq_len: int = 200
|
||||
r: int = 1
|
||||
|
|
@ -0,0 +1,10 @@
|
|||
from dataclasses import dataclass
|
||||
|
||||
from TTS.tts.configs.tacotron_config import TacotronConfig
|
||||
|
||||
|
||||
@dataclass
|
||||
class Tacotron2Config(TacotronConfig):
|
||||
"""Defines parameters for Tacotron2 based models."""
|
||||
|
||||
model: str = "tacotron2"
|
||||
|
|
@ -0,0 +1,70 @@
|
|||
from dataclasses import asdict, dataclass
|
||||
from typing import List
|
||||
|
||||
from coqpit import check_argument
|
||||
|
||||
from .shared_configs import BaseTTSConfig, GSTConfig
|
||||
|
||||
|
||||
@dataclass
|
||||
class TacotronConfig(BaseTTSConfig):
|
||||
"""Defines parameters for Tacotron based models."""
|
||||
|
||||
model: str = "tacotron"
|
||||
gst: GSTConfig = None
|
||||
gst_style_input: str = None
|
||||
# model specific params
|
||||
r: int = 2
|
||||
gradual_training: List = None
|
||||
memory_size: int = -1
|
||||
prenet_type: str = "original"
|
||||
prenet_dropout: bool = True
|
||||
prenet_dropout_at_inference: bool = False
|
||||
stopnet: bool = True
|
||||
separate_stopnet: bool = True
|
||||
stopnet_pos_weight: float = 10.0
|
||||
|
||||
# attention layers
|
||||
attention_type: str = "original"
|
||||
attention_heads: int = None
|
||||
attention_norm: str = "sigmoid"
|
||||
windowing: bool = False
|
||||
use_forward_attn: bool = False
|
||||
forward_attn_mask: bool = False
|
||||
transition_agent: bool = False
|
||||
location_attn: bool = True
|
||||
|
||||
# advance methods
|
||||
bidirectional_decoder: bool = False
|
||||
double_decoder_consistency: bool = False
|
||||
ddc_r: int = 6
|
||||
|
||||
# multi-speaker settings
|
||||
use_speaker_embedding: bool = False
|
||||
use_external_speaker_embedding_file: bool = False
|
||||
external_speaker_embedding_file: str = False
|
||||
|
||||
# optimizer parameters
|
||||
noam_schedule: bool = False
|
||||
warmup_steps: int = 4000
|
||||
lr: float = 1e-4
|
||||
wd: float = 1e-6
|
||||
grad_clip: float = 5.0
|
||||
seq_len_norm: bool = False
|
||||
loss_masking: bool = True
|
||||
|
||||
# loss params
|
||||
decoder_loss_alpha: float = 0.25
|
||||
postnet_loss_alpha: float = 0.25
|
||||
postnet_diff_spec_alpha: float = 0.25
|
||||
decoder_diff_spec_alpha: float = 0.25
|
||||
decoder_ssim_alpha: float = 0.25
|
||||
postnet_ssim_alpha: float = 0.25
|
||||
ga_alpha: float = 5.0
|
||||
|
||||
|
||||
@dataclass
|
||||
class Tacotron2Config(TacotronConfig):
|
||||
"""Defines parameters for Tacotron2 based models."""
|
||||
|
||||
model: str = "tacotron2"
|
||||
Loading…
Reference in New Issue