From 6700bb1bcfc972a193a3921538152120ef7c35b1 Mon Sep 17 00:00:00 2001
From: WeberJulian <julian.weber@hotmail.fr>
Date: Wed, 8 Dec 2021 19:42:45 +0100
Subject: [PATCH] Add recipe for multi-lingual VITS

---
 .../multilingual/vits_tts/train_vits_tts.py   | 117 ++++++++++++++++++
 1 file changed, 117 insertions(+)
 create mode 100644 recipes/multilingual/vits_tts/train_vits_tts.py

diff --git a/recipes/multilingual/vits_tts/train_vits_tts.py b/recipes/multilingual/vits_tts/train_vits_tts.py
new file mode 100644
index 00000000..6beaef38
--- /dev/null
+++ b/recipes/multilingual/vits_tts/train_vits_tts.py
@@ -0,0 +1,117 @@
+import os
+from glob import glob
+
+from TTS.config.shared_configs import BaseAudioConfig
+from TTS.trainer import Trainer, TrainingArgs
+from TTS.tts.configs.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 Vits, VitsArgs
+from TTS.tts.utils.speakers import SpeakerManager
+from TTS.tts.utils.languages import LanguageManager
+from TTS.utils.audio import AudioProcessor
+
+output_path = os.path.dirname(os.path.abspath(__file__))
+
+mailabs_path = '/home/julian/workspace/mailabs/**'
+dataset_paths = glob(mailabs_path)
+dataset_config = [BaseDatasetConfig(name="mailabs", meta_file_train=None, path=path, language=path.split('/')[-1]) for path in dataset_paths]
+
+audio_config = BaseAudioConfig(
+    sample_rate=16000,
+    win_length=1024,
+    hop_length=256,
+    num_mels=80,
+    preemphasis=0.0,
+    ref_level_db=20,
+    log_func="np.log",
+    do_trim_silence=False,
+    trim_db=23.0,
+    mel_fmin=0,
+    mel_fmax=None,
+    spec_gain=1.0,
+    signal_norm=True,
+    do_amp_to_db_linear=False,
+    resample=False,
+)
+
+vitsArgs = VitsArgs(
+    use_language_embedding=True,
+    embedded_language_dim=4,
+    use_speaker_embedding=True,
+    use_sdp=False,
+)
+
+config = VitsConfig(
+    model_args=vitsArgs,
+    audio=audio_config,
+    run_name="vits_vctk",
+    use_speaker_embedding=True,
+    batch_size=32,
+    eval_batch_size=16,
+    batch_group_size=0,
+    num_loader_workers=4,
+    num_eval_loader_workers=4,
+    run_eval=True,
+    test_delay_epochs=-1,
+    epochs=1000,
+    text_cleaner="multilingual_cleaners",
+    use_phonemes=False,
+    phoneme_language="en-us",
+    phoneme_cache_path=os.path.join(output_path, "phoneme_cache"),
+    compute_input_seq_cache=True,
+    print_step=25,
+    use_language_weighted_sampler= True,
+    print_eval=False,
+    mixed_precision=False,
+    sort_by_audio_len=True,
+    min_seq_len=32 * 256 * 4,
+    max_seq_len=160000,
+    output_path=output_path,
+    datasets=dataset_config,
+    characters= {
+        "pad": "_",
+        "eos": "&",
+        "bos": "*",
+        "characters": "!¡'(),-.:;¿?abcdefghijklmnopqrstuvwxyzµßàáâäåæçèéêëìíîïñòóôöùúûüąćęłńœśşźżƒабвгдежзийклмнопрстуфхцчшщъыьэюяёєіїґӧ «°±µ»$%&‘’‚“`”„",
+        "punctuations": "!¡'(),-.:;¿? ",
+        "phonemes": None,
+        "unique": True
+    },
+    test_sentences=[
+        ["It took me quite a long time to develop a voice, and now that I have it I'm not going to be silent.", 'mary_ann', None, 'en_US'],
+        ["Il m'a fallu beaucoup de temps pour d\u00e9velopper une voix, et maintenant que je l'ai, je ne vais pas me taire.", "ezwa", None, 'fr_FR'],
+        ["Ich finde, dieses Startup ist wirklich unglaublich.", "eva_k", None, 'de_DE'],
+        ["Я думаю, что этот стартап действительно удивительный.", "oblomov", None, 'ru_RU'],
+    ]
+)
+
+# init audio processor
+ap = AudioProcessor(**config.audio.to_dict())
+
+# load training samples
+train_samples, eval_samples = load_tts_samples(dataset_config, eval_split=True)
+
+# init speaker manager for multi-speaker training
+# it maps speaker-id to speaker-name in the model and data-loader
+speaker_manager = SpeakerManager()
+speaker_manager.set_speaker_ids_from_data(train_samples + eval_samples)
+config.model_args.num_speakers = speaker_manager.num_speakers
+
+language_manager = LanguageManager(config=config)
+config.model_args.num_languages = language_manager.num_languages
+
+# init model
+model = Vits(config, speaker_manager, language_manager)
+
+# init the trainer and 🚀
+trainer = Trainer(
+    TrainingArgs(),
+    config,
+    output_path,
+    model=model,
+    train_samples=train_samples,
+    eval_samples=eval_samples,
+    training_assets={"audio_processor": ap},
+)
+trainer.fit()