recipes/multilingual/cml_yourtts/train_yourtts.py

import os

import torch
from trainer import Trainer, TrainerArgs

from TTS.bin.compute_embeddings import compute_embeddings
from TTS.bin.resample import resample_files
from TTS.config.shared_configs import BaseDatasetConfig
from TTS.tts.configs.vits_config import VitsConfig
from TTS.tts.datasets import load_tts_samples
from TTS.tts.models.vits import CharactersConfig, Vits, VitsArgs, VitsAudioConfig
from TTS.utils.downloaders import download_libri_tts

torch.set_num_threads(24)

# pylint: disable=W0105
"""
    This recipe replicates the first experiment proposed in the CML-TTS paper (https://arxiv.org/abs/2306.10097). It uses the YourTTS model.
    YourTTS model is based on the VITS model however it uses external speaker embeddings extracted from a pre-trained speaker encoder and has small architecture changes.
"""
CURRENT_PATH = os.path.dirname(os.path.abspath(__file__))

# Name of the run for the Trainer
RUN_NAME = "YourTTS-CML-TTS"

# Path where you want to save the models outputs (configs, checkpoints and tensorboard logs)
OUT_PATH = os.path.dirname(os.path.abspath(__file__))  # "/raid/coqui/Checkpoints/original-YourTTS/"

# If you want to do transfer learning and speedup your training you can set here the path to the CML-TTS available checkpoint that cam be downloaded here:  https://drive.google.com/u/2/uc?id=1yDCSJ1pFZQTHhL09GMbOrdjcPULApa0p
RESTORE_PATH = "/raid/edresson/CML_YourTTS/checkpoints_yourtts_cml_tts_dataset/best_model.pth"  # Download the checkpoint here:  https://drive.google.com/u/2/uc?id=1yDCSJ1pFZQTHhL09GMbOrdjcPULApa0p

# This paramter is useful to debug, it skips the training epochs and just do the evaluation  and produce the test sentences
SKIP_TRAIN_EPOCH = False

# Set here the batch size to be used in training and evaluation
BATCH_SIZE = 32

# Training Sampling rate and the target sampling rate for resampling the downloaded dataset (Note: If you change this you might need to redownload the dataset !!)
# Note: If you add new datasets, please make sure that the dataset sampling rate and this parameter are matching, otherwise resample your audios
SAMPLE_RATE = 24000

# Max audio length in seconds to be used in training (every audio bigger than it will be ignored)
MAX_AUDIO_LEN_IN_SECONDS = float("inf")

### Download CML-TTS dataset
# You need to download the dataset for all languages manually and extract it to a path and then set the CML_DATASET_PATH to this path: https://github.com/freds0/CML-TTS-Dataset#download
CML_DATASET_PATH = "./datasets/CML-TTS-Dataset/"


### Download LibriTTS dataset
# it will automatic download the dataset, if you have problems you can comment it and manually donwload and extract it ! Download link: https://www.openslr.org/resources/60/train-clean-360.tar.gz
LIBRITTS_DOWNLOAD_PATH = "./datasets/LibriTTS/"
# Check if LibriTTS dataset is not already downloaded, if not download it
if not os.path.exists(LIBRITTS_DOWNLOAD_PATH):
    print(">>> Downloading LibriTTS dataset:")
    download_libri_tts(LIBRITTS_DOWNLOAD_PATH, subset="libri-tts-clean-360")

# init LibriTTS configs
libritts_config = BaseDatasetConfig(
    formatter="libri_tts",
    dataset_name="libri_tts",
    meta_file_train="",
    meta_file_val="",
    path=os.path.join(LIBRITTS_DOWNLOAD_PATH, "train-clean-360/"),
    language="en",
)

# init CML-TTS configs
pt_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_portuguese_v0.1/"),
    language="pt-br",
)

pl_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_polish_v0.1/"),
    language="pl",
)

it_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_italian_v0.1/"),
    language="it",
)

fr_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_french_v0.1/"),
    language="fr",
)

du_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_dutch_v0.1/"),
    language="du",
)

ge_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_german_v0.1/"),
    language="ge",
)

sp_config = BaseDatasetConfig(
    formatter="cml_tts",
    dataset_name="cml_tts",
    meta_file_train="train.csv",
    meta_file_val="",
    path=os.path.join(CML_DATASET_PATH, "cml_tts_dataset_spanish_v0.1/"),
    language="sp",
)

# Add here all datasets configs Note: If you want to add new datasets, just add them here and it will automatically compute the speaker embeddings (d-vectors) for this new dataset :)
DATASETS_CONFIG_LIST = [libritts_config, pt_config, pl_config, it_config, fr_config, du_config, ge_config, sp_config]

### Extract speaker embeddings
SPEAKER_ENCODER_CHECKPOINT_PATH = (
    "https://github.com/coqui-ai/TTS/releases/download/speaker_encoder_model/model_se.pth.tar"
)
SPEAKER_ENCODER_CONFIG_PATH = "https://github.com/coqui-ai/TTS/releases/download/speaker_encoder_model/config_se.json"

D_VECTOR_FILES = []  # List of speaker embeddings/d-vectors to be used during the training

# Iterates all the dataset configs checking if the speakers embeddings are already computated, if not compute it
for dataset_conf in DATASETS_CONFIG_LIST:
    # Check if the embeddings weren't already computed, if not compute it
    embeddings_file = os.path.join(dataset_conf.path, "speakers.pth")
    if not os.path.isfile(embeddings_file):
        print(f">>> Computing the speaker embeddings for the {dataset_conf.dataset_name} dataset")
        compute_embeddings(
            SPEAKER_ENCODER_CHECKPOINT_PATH,
            SPEAKER_ENCODER_CONFIG_PATH,
            embeddings_file,
            old_speakers_file=None,
            config_dataset_path=None,
            formatter_name=dataset_conf.formatter,
            dataset_name=dataset_conf.dataset_name,
            dataset_path=dataset_conf.path,
            meta_file_train=dataset_conf.meta_file_train,
            meta_file_val=dataset_conf.meta_file_val,
            disable_cuda=False,
            no_eval=False,
        )
    D_VECTOR_FILES.append(embeddings_file)


# Audio config used in training.
audio_config = VitsAudioConfig(
    sample_rate=SAMPLE_RATE,
    hop_length=256,
    win_length=1024,
    fft_size=1024,
    mel_fmin=0.0,
    mel_fmax=None,
    num_mels=80,
)

# Init VITSArgs setting the arguments that are needed for the YourTTS model
model_args = VitsArgs(
    spec_segment_size=62,
    hidden_channels=192,
    hidden_channels_ffn_text_encoder=768,
    num_heads_text_encoder=2,
    num_layers_text_encoder=10,
    kernel_size_text_encoder=3,
    dropout_p_text_encoder=0.1,
    d_vector_file=D_VECTOR_FILES,
    use_d_vector_file=True,
    d_vector_dim=512,
    speaker_encoder_model_path=SPEAKER_ENCODER_CHECKPOINT_PATH,
    speaker_encoder_config_path=SPEAKER_ENCODER_CONFIG_PATH,
    resblock_type_decoder="2",  # In the paper, we accidentally trained the YourTTS using ResNet blocks type 2, if you like you can use the ResNet blocks type 1 like the VITS model
    # Useful parameters to enable the Speaker Consistency Loss (SCL) described in the paper
    use_speaker_encoder_as_loss=False,
    # Useful parameters to enable multilingual training
    use_language_embedding=True,
    embedded_language_dim=4,
)

# General training config, here you can change the batch size and others useful parameters
config = VitsConfig(
    output_path=OUT_PATH,
    model_args=model_args,
    run_name=RUN_NAME,
    project_name="YourTTS",
    run_description="""
            - YourTTS trained using CML-TTS and LibriTTS datasets
        """,
    dashboard_logger="tensorboard",
    logger_uri=None,
    audio=audio_config,
    batch_size=BATCH_SIZE,
    batch_group_size=48,
    eval_batch_size=BATCH_SIZE,
    num_loader_workers=8,
    eval_split_max_size=256,
    print_step=50,
    plot_step=100,
    log_model_step=1000,
    save_step=5000,
    save_n_checkpoints=2,
    save_checkpoints=True,
    target_loss="loss_1",
    print_eval=False,
    use_phonemes=False,
    phonemizer="espeak",
    phoneme_language="en",
    compute_input_seq_cache=True,
    add_blank=True,
    text_cleaner="multilingual_cleaners",
    characters=CharactersConfig(
        characters_class="TTS.tts.models.vits.VitsCharacters",
        pad="_",
        eos="&",
        bos="*",
        blank=None,
        characters="ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz\u00a1\u00a3\u00b7\u00b8\u00c0\u00c1\u00c2\u00c3\u00c4\u00c5\u00c7\u00c8\u00c9\u00ca\u00cb\u00cc\u00cd\u00ce\u00cf\u00d1\u00d2\u00d3\u00d4\u00d5\u00d6\u00d9\u00da\u00db\u00dc\u00df\u00e0\u00e1\u00e2\u00e3\u00e4\u00e5\u00e7\u00e8\u00e9\u00ea\u00eb\u00ec\u00ed\u00ee\u00ef\u00f1\u00f2\u00f3\u00f4\u00f5\u00f6\u00f9\u00fa\u00fb\u00fc\u0101\u0104\u0105\u0106\u0107\u010b\u0119\u0141\u0142\u0143\u0144\u0152\u0153\u015a\u015b\u0161\u0178\u0179\u017a\u017b\u017c\u020e\u04e7\u05c2\u1b20",
        punctuations="\u2014!'(),-.:;?\u00bf ",
        phonemes="iy\u0268\u0289\u026fu\u026a\u028f\u028ae\u00f8\u0258\u0259\u0275\u0264o\u025b\u0153\u025c\u025e\u028c\u0254\u00e6\u0250a\u0276\u0251\u0252\u1d7b\u0298\u0253\u01c0\u0257\u01c3\u0284\u01c2\u0260\u01c1\u029bpbtd\u0288\u0256c\u025fk\u0261q\u0262\u0294\u0274\u014b\u0272\u0273n\u0271m\u0299r\u0280\u2c71\u027e\u027d\u0278\u03b2fv\u03b8\u00f0sz\u0283\u0292\u0282\u0290\u00e7\u029dx\u0263\u03c7\u0281\u0127\u0295h\u0266\u026c\u026e\u028b\u0279\u027bj\u0270l\u026d\u028e\u029f\u02c8\u02cc\u02d0\u02d1\u028dw\u0265\u029c\u02a2\u02a1\u0255\u0291\u027a\u0267\u025a\u02de\u026b'\u0303' ",
        is_unique=True,
        is_sorted=True,
    ),
    phoneme_cache_path=None,
    precompute_num_workers=12,
    start_by_longest=True,
    datasets=DATASETS_CONFIG_LIST,
    cudnn_benchmark=False,
    max_audio_len=SAMPLE_RATE * MAX_AUDIO_LEN_IN_SECONDS,
    mixed_precision=False,
    test_sentences=[
        ["Voc\u00ea ter\u00e1 a vista do topo da montanha que voc\u00ea escalar.", "9351", None, "pt-br"],
        ["Quando voc\u00ea n\u00e3o corre nenhum risco, voc\u00ea arrisca tudo.", "12249", None, "pt-br"],
        [
            "S\u00e3o necess\u00e1rios muitos anos de trabalho para ter sucesso da noite para o dia.",
            "2961",
            None,
            "pt-br",
        ],
        ["You'll have the view of the top of the mountain that you climb.", "LTTS_6574", None, "en"],
        ["When you don\u2019t take any risks, you risk everything.", "LTTS_6206", None, "en"],
        ["Are necessary too many years of work to succeed overnight.", "LTTS_5717", None, "en"],
        ["Je hebt uitzicht op de top van de berg die je beklimt.", "960", None, "du"],
        ["Als je geen risico neemt, riskeer je alles.", "2450", None, "du"],
        ["Zijn te veel jaren werk nodig om van de ene op de andere dag te slagen.", "10984", None, "du"],
        ["Vous aurez la vue sur le sommet de la montagne que vous gravirez.", "6381", None, "fr"],
        ["Quand tu ne prends aucun risque, tu risques tout.", "2825", None, "fr"],
        [
            "Sont n\u00e9cessaires trop d'ann\u00e9es de travail pour r\u00e9ussir du jour au lendemain.",
            "1844",
            None,
            "fr",
        ],
        ["Sie haben die Aussicht auf die Spitze des Berges, den Sie erklimmen.", "2314", None, "ge"],
        ["Wer nichts riskiert, riskiert alles.", "7483", None, "ge"],
        ["Es sind zu viele Jahre Arbeit notwendig, um \u00fcber Nacht erfolgreich zu sein.", "12461", None, "ge"],
        ["Avrai la vista della cima della montagna che sali.", "4998", None, "it"],
        ["Quando non corri alcun rischio, rischi tutto.", "6744", None, "it"],
        ["Are necessary too many years of work to succeed overnight.", "1157", None, "it"],
        [
            "B\u0119dziesz mie\u0107 widok na szczyt g\u00f3ry, na kt\u00f3r\u0105 si\u0119 wspinasz.",
            "7014",
            None,
            "pl",
        ],
        ["Kiedy nie podejmujesz \u017cadnego ryzyka, ryzykujesz wszystko.", "3492", None, "pl"],
        [
            "Potrzebne s\u0105 zbyt wiele lat pracy, aby odnie\u015b\u0107 sukces z dnia na dzie\u0144.",
            "1890",
            None,
            "pl",
        ],
        ["Tendr\u00e1s la vista de la cima de la monta\u00f1a que subes", "101", None, "sp"],
        ["Cuando no te arriesgas, lo arriesgas todo.", "5922", None, "sp"],
        [
            "Son necesarios demasiados a\u00f1os de trabajo para triunfar de la noche a la ma\u00f1ana.",
            "10246",
            None,
            "sp",
        ],
    ],
    # Enable the weighted sampler
    use_weighted_sampler=True,
    # Ensures that all speakers are seen in the training batch equally no matter how many samples each speaker has
    # weighted_sampler_attrs={"language": 1.0, "speaker_name": 1.0},
    weighted_sampler_attrs={"language": 1.0},
    weighted_sampler_multipliers={
        # "speaker_name": {
        # you can force the batching scheme to give a higher weight to a certain speaker and then this speaker will appears more frequently on the batch.
        # It will speedup the speaker adaptation process. Considering the CML train dataset and "new_speaker" as the speaker name of the speaker that you want to adapt.
        # The line above will make the balancer consider the "new_speaker" as 106 speakers so 1/4 of the number of speakers present on CML dataset.
        # 'new_speaker': 106, # (CML tot. train speaker)/4 = (424/4) = 106
        # }
    },
    # It defines the Speaker Consistency Loss (SCL) α to 9 like the YourTTS paper
    speaker_encoder_loss_alpha=9.0,
)

# Load all the datasets samples and split traning and evaluation sets
train_samples, eval_samples = load_tts_samples(
    config.datasets,
    eval_split=True,
    eval_split_max_size=config.eval_split_max_size,
    eval_split_size=config.eval_split_size,
)

# Init the model
model = Vits.init_from_config(config)

# Init the trainer and 🚀
trainer = Trainer(
    TrainerArgs(restore_path=RESTORE_PATH, skip_train_epoch=SKIP_TRAIN_EPOCH),
    config,
    output_path=OUT_PATH,
    model=model,
    train_samples=train_samples,
    eval_samples=eval_samples,
)
trainer.fit()