Initial XTTS Urdu model upload

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 sample_wav.wav filter=lfs diff=lfs merge=lfs -text
+vocals_chunk_0181_1306890ms_wip8_24k.wav filter=lfs diff=lfs merge=lfs -text

checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/checkpoint_24.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7ef89c6696a7dbc7ff24fdfcdd76399802d4618f8375976e66227f1a6dea6d4
+size 2886723130

checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 1,
+    "batch_size": 16,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 5e-06,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/events.out.tfevents.1743585073.394654dd939a.14791.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ecb0422ccc42fc4dff87e335c878a5ce0c781474eabe80053e263030e3ac4e06
+size 15576

checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/train_gpt_xtts.py

@@ -0,0 +1,237 @@
+import os
+import gc
+
+from trainer import Trainer, TrainerArgs
+
+from TTS.config.shared_configs import BaseDatasetConfig
+from TTS.tts.datasets import load_tts_samples
+from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrainerConfig, XttsAudioConfig
+from TTS.utils.manage import ModelManager
+
+from dataclasses import dataclass, field
+from typing import Optional
+from transformers import HfArgumentParser
+
+import argparse
+
+def create_xtts_trainer_parser():
+    parser = argparse.ArgumentParser(description="Arguments for XTTS Trainer")
+
+    parser.add_argument("--output_path", type=str, required=True,
+                        help="Path to pretrained + checkpoint model")
+    parser.add_argument("--metadatas", nargs='+', type=str, required=True,
+                        help="train_csv_path,eval_csv_path,language")
+    parser.add_argument("--num_epochs", type=int, default=1,
+                        help="Number of epochs")
+    parser.add_argument("--batch_size", type=int, default=1,
+                        help="Mini batch size")
+    parser.add_argument("--grad_acumm", type=int, default=1,
+                        help="Grad accumulation steps")
+    parser.add_argument("--max_audio_length", type=int, default=255995,
+                        help="Max audio length")
+    parser.add_argument("--max_text_length", type=int, default=200,
+                        help="Max text length")
+    parser.add_argument("--weight_decay", type=float, default=1e-2,
+                        help="Weight decay")
+    parser.add_argument("--lr", type=float, default=5e-6,
+                        help="Learning rate")
+    parser.add_argument("--save_step", type=int, default=5000,
+                        help="Save step")
+
+    return parser
+
+
+
+def train_gpt(metadatas, num_epochs, batch_size, grad_acumm, output_path, max_audio_length, max_text_length, lr, weight_decay, save_step):
+    #  Logging parameters
+    RUN_NAME = "GPT_XTTS_FT"
+    PROJECT_NAME = "XTTS_trainer"
+    DASHBOARD_LOGGER = "tensorboard"
+    LOGGER_URI = None
+
+    # Set here the path that the checkpoints will be saved. Default: ./run/training/
+    # OUT_PATH = os.path.join(output_path, "run", "training")
+    OUT_PATH = output_path
+
+    # Training Parameters
+    OPTIMIZER_WD_ONLY_ON_WEIGHTS = True  # for multi-gpu training please make it False
+    START_WITH_EVAL = False  # if True it will star with evaluation
+    BATCH_SIZE = batch_size  # set here the batch size
+    GRAD_ACUMM_STEPS = grad_acumm  # set here the grad accumulation steps
+
+
+    # Define here the dataset that you want to use for the fine-tuning on.
+    DATASETS_CONFIG_LIST = []
+    for metadata in metadatas:
+        train_csv, eval_csv, language = metadata.split(",")
+        print(train_csv, eval_csv, language)
+
+        config_dataset = BaseDatasetConfig(
+            formatter="coqui",
+            dataset_name="ft_dataset",
+            path=os.path.dirname(train_csv),
+            meta_file_train=os.path.basename(train_csv),
+            meta_file_val=os.path.basename(eval_csv),
+            language=language,
+        )
+
+        DATASETS_CONFIG_LIST.append(config_dataset)
+
+    # Define the path where XTTS v2.0.1 files will be downloaded
+    CHECKPOINTS_OUT_PATH = os.path.join(OUT_PATH, "XTTS_v2.0_original_model_files/")
+    os.makedirs(CHECKPOINTS_OUT_PATH, exist_ok=True)
+
+
+    # DVAE files
+    DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/dvae.pth"
+    MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/mel_stats.pth"
+
+    # Set the path to the downloaded files
+    DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(DVAE_CHECKPOINT_LINK))
+    MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(MEL_NORM_LINK))
+
+    # download DVAE files if needed
+    if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
+        print(" > Downloading DVAE files!")
+        ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)
+
+
+    # Download XTTS v2.0 checkpoint if needed
+    TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json"
+    XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth"
+    XTTS_CONFIG_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json"
+
+    # XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
+    TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(TOKENIZER_FILE_LINK))  # vocab.json file
+    XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CHECKPOINT_LINK))  # model.pth file
+    XTTS_CONFIG_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CONFIG_LINK))  # config.json file
+
+    # download XTTS v2.0 files if needed
+    if not os.path.isfile(TOKENIZER_FILE):
+        print(" > Downloading XTTS v2.0 tokenizer!")
+        ModelManager._download_model_files(
+            [TOKENIZER_FILE_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CHECKPOINT):
+        print(" > Downloading XTTS v2.0 checkpoint!")
+        ModelManager._download_model_files(
+            [XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CONFIG_FILE):
+        print(" > Downloading XTTS v2.0 config!")
+        ModelManager._download_model_files(
+            [XTTS_CONFIG_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+
+    # init args and config
+    model_args = GPTArgs(
+        max_conditioning_length=132300,  # 6 secs
+        min_conditioning_length=11025,  # 0.5 secs
+        debug_loading_failures=False,
+        max_wav_length=max_audio_length,  # ~11.6 seconds
+        max_text_length=max_text_length,
+        mel_norm_file=MEL_NORM_FILE,
+        dvae_checkpoint=DVAE_CHECKPOINT,
+        xtts_checkpoint=XTTS_CHECKPOINT,  # checkpoint path of the model that you want to fine-tune
+        tokenizer_file=TOKENIZER_FILE,
+        gpt_num_audio_tokens=1026,
+        gpt_start_audio_token=1024,
+        gpt_stop_audio_token=1025,
+        gpt_use_masking_gt_prompt_approach=True,
+        gpt_use_perceiver_resampler=True,
+    )
+    # define audio config
+    audio_config = XttsAudioConfig(sample_rate=22050, dvae_sample_rate=22050, output_sample_rate=24000)
+    # training parameters config
+
+    config = GPTTrainerConfig()
+
+    config.load_json(XTTS_CONFIG_FILE)
+
+    config.epochs = num_epochs
+    config.output_path = OUT_PATH
+    config.model_args = model_args
+    config.run_name = RUN_NAME
+    config.project_name = PROJECT_NAME
+    config.run_description = """
+        GPT XTTS training
+        """,
+    config.dashboard_logger = DASHBOARD_LOGGER
+    config.logger_uri = LOGGER_URI
+    config.audio = audio_config
+    config.batch_size = BATCH_SIZE
+    config.num_loader_workers = 8
+    config.eval_split_max_size = 256
+    config.print_step = 50
+    config.plot_step = 100
+    config.log_model_step = 100
+    config.save_step = save_step
+    config.save_n_checkpoints = 1
+    config.save_checkpoints = True
+    config.print_eval = False
+    config.optimizer = "AdamW"
+    config.optimizer_wd_only_on_weights = OPTIMIZER_WD_ONLY_ON_WEIGHTS
+    config.optimizer_params = {"betas": [0.9, 0.96], "eps": 1e-8, "weight_decay": weight_decay}
+    config.lr = lr
+    config.lr_scheduler = "MultiStepLR"
+    config.lr_scheduler_params = {"milestones": [50000 * 18, 150000 * 18, 300000 * 18], "gamma": 0.5, "last_epoch": -1}
+    config.test_sentences = []
+
+    # init the model from config
+    model = GPTTrainer.init_from_config(config)
+
+    # load training samples
+    train_samples, eval_samples = load_tts_samples(
+        DATASETS_CONFIG_LIST,
+        eval_split=True,
+        eval_split_max_size=config.eval_split_max_size,
+        eval_split_size=config.eval_split_size,
+    )
+
+    # init the trainer and 🚀
+    trainer = Trainer(
+        TrainerArgs(
+            restore_path=None,  # xtts checkpoint is restored via xtts_checkpoint key so no need of restore it using Trainer restore_path parameter
+            skip_train_epoch=False,
+            start_with_eval=START_WITH_EVAL,
+            grad_accum_steps=GRAD_ACUMM_STEPS
+        ),
+        config,
+        output_path=os.path.join(output_path, "run", "training"),
+        model=model,
+        train_samples=train_samples,
+        eval_samples=eval_samples,
+    )
+    trainer.fit()
+
+    # get the longest text audio file to use as speaker reference
+    samples_len = [len(item["text"].split(" ")) for item in train_samples]
+    longest_text_idx =  samples_len.index(max(samples_len))
+    speaker_ref = train_samples[longest_text_idx]["audio_file"]
+
+    trainer_out_path = trainer.output_path
+
+    # deallocate VRAM and RAM
+    del model, trainer, train_samples, eval_samples
+    gc.collect()
+
+    return trainer_out_path
+
+if __name__ == "__main__":
+    parser = create_xtts_trainer_parser()
+    args = parser.parse_args()
+
+    trainer_out_path = train_gpt(
+        metadatas=args.metadatas,
+        output_path=args.output_path,
+        num_epochs=args.num_epochs,
+        batch_size=args.batch_size,
+        grad_acumm=args.grad_acumm,
+        weight_decay=args.weight_decay,
+        lr=args.lr,
+        max_text_length=args.max_text_length,
+        max_audio_length=args.max_audio_length,
+        save_step=args.save_step
+    )
+
+    print(f"Checkpoint saved in dir: {trainer_out_path}")

checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/trainer_0_log.txt

@@ -0,0 +1,34 @@
+ > Training Environment:
+ | > Backend: Torch
+ | > Mixed precision: False
+ | > Precision: float32
+ | > Current device: 0
+ | > Num. of GPUs: 1
+ | > Num. of CPUs: 192
+ | > Num. of Torch Threads: 1
+ | > Torch seed: 54321
+ | > Torch CUDNN: True
+ | > Torch CUDNN deterministic: False
+ | > Torch CUDNN benchmark: False
+ | > Torch TF32 MatMul: False
+ > Start Tensorboard: tensorboard --logdir=checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3
+
+ > Model has 526224149 parameters
+
+ > EPOCH: 0/1
+ --> checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3
+
+ > TRAINING (2025-04-02 09:11:14) 
+
+   --> TIME: 2025-04-02 09:11:16 -- STEP: 0/6206 -- GLOBAL_STEP: 0
+     | > loss_text_ce: 0.20948722958564758  (0.20948722958564758)
+     | > loss_mel_ce: 5.6751861572265625  (5.6751861572265625)
+     | > loss: 1.4711683988571167  (1.4711683988571167)
+     | > current_lr: 5e-06 
+     | > step_time: 0.6851  (0.6850714683532715)
+     | > loader_time: 1.3635  (1.3634731769561768)
+
+ > Keyboard interrupt detected.
+ > Saving model before exiting...
+
+ > CHECKPOINT : checkpoints/GPT_XTTS_FT-April-02-2025_09+11AM-8e59ec3/checkpoint_24.pth

checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/checkpoint_90.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6035c0b71da5fa095e35021c4970365a26e4c504abddde1f89f02af9844936b
+size 2924473458

checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 1,
+    "batch_size": 16,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 5e-06,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/events.out.tfevents.1743585158.394654dd939a.15416.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91b60bac6e5d8dbfc005ce7b6e248ceb2a78d10e4e64bd27e2ea4641003b8e42
+size 15576

checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/train_gpt_xtts.py

@@ -0,0 +1,237 @@
+import os
+import gc
+
+from trainer import Trainer, TrainerArgs
+
+from TTS.config.shared_configs import BaseDatasetConfig
+from TTS.tts.datasets import load_tts_samples
+from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrainerConfig, XttsAudioConfig
+from TTS.utils.manage import ModelManager
+
+from dataclasses import dataclass, field
+from typing import Optional
+from transformers import HfArgumentParser
+
+import argparse
+
+def create_xtts_trainer_parser():
+    parser = argparse.ArgumentParser(description="Arguments for XTTS Trainer")
+
+    parser.add_argument("--output_path", type=str, required=True,
+                        help="Path to pretrained + checkpoint model")
+    parser.add_argument("--metadatas", nargs='+', type=str, required=True,
+                        help="train_csv_path,eval_csv_path,language")
+    parser.add_argument("--num_epochs", type=int, default=1,
+                        help="Number of epochs")
+    parser.add_argument("--batch_size", type=int, default=1,
+                        help="Mini batch size")
+    parser.add_argument("--grad_acumm", type=int, default=1,
+                        help="Grad accumulation steps")
+    parser.add_argument("--max_audio_length", type=int, default=255995,
+                        help="Max audio length")
+    parser.add_argument("--max_text_length", type=int, default=200,
+                        help="Max text length")
+    parser.add_argument("--weight_decay", type=float, default=1e-2,
+                        help="Weight decay")
+    parser.add_argument("--lr", type=float, default=5e-6,
+                        help="Learning rate")
+    parser.add_argument("--save_step", type=int, default=5000,
+                        help="Save step")
+
+    return parser
+
+
+
+def train_gpt(metadatas, num_epochs, batch_size, grad_acumm, output_path, max_audio_length, max_text_length, lr, weight_decay, save_step):
+    #  Logging parameters
+    RUN_NAME = "GPT_XTTS_FT"
+    PROJECT_NAME = "XTTS_trainer"
+    DASHBOARD_LOGGER = "tensorboard"
+    LOGGER_URI = None
+
+    # Set here the path that the checkpoints will be saved. Default: ./run/training/
+    # OUT_PATH = os.path.join(output_path, "run", "training")
+    OUT_PATH = output_path
+
+    # Training Parameters
+    OPTIMIZER_WD_ONLY_ON_WEIGHTS = True  # for multi-gpu training please make it False
+    START_WITH_EVAL = False  # if True it will star with evaluation
+    BATCH_SIZE = batch_size  # set here the batch size
+    GRAD_ACUMM_STEPS = grad_acumm  # set here the grad accumulation steps
+
+
+    # Define here the dataset that you want to use for the fine-tuning on.
+    DATASETS_CONFIG_LIST = []
+    for metadata in metadatas:
+        train_csv, eval_csv, language = metadata.split(",")
+        print(train_csv, eval_csv, language)
+
+        config_dataset = BaseDatasetConfig(
+            formatter="coqui",
+            dataset_name="ft_dataset",
+            path=os.path.dirname(train_csv),
+            meta_file_train=os.path.basename(train_csv),
+            meta_file_val=os.path.basename(eval_csv),
+            language=language,
+        )
+
+        DATASETS_CONFIG_LIST.append(config_dataset)
+
+    # Define the path where XTTS v2.0.1 files will be downloaded
+    CHECKPOINTS_OUT_PATH = os.path.join(OUT_PATH, "XTTS_v2.0_original_model_files/")
+    os.makedirs(CHECKPOINTS_OUT_PATH, exist_ok=True)
+
+
+    # DVAE files
+    DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/dvae.pth"
+    MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/mel_stats.pth"
+
+    # Set the path to the downloaded files
+    DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(DVAE_CHECKPOINT_LINK))
+    MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(MEL_NORM_LINK))
+
+    # download DVAE files if needed
+    if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
+        print(" > Downloading DVAE files!")
+        ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)
+
+
+    # Download XTTS v2.0 checkpoint if needed
+    TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json"
+    XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth"
+    XTTS_CONFIG_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json"
+
+    # XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
+    TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(TOKENIZER_FILE_LINK))  # vocab.json file
+    XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CHECKPOINT_LINK))  # model.pth file
+    XTTS_CONFIG_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CONFIG_LINK))  # config.json file
+
+    # download XTTS v2.0 files if needed
+    if not os.path.isfile(TOKENIZER_FILE):
+        print(" > Downloading XTTS v2.0 tokenizer!")
+        ModelManager._download_model_files(
+            [TOKENIZER_FILE_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CHECKPOINT):
+        print(" > Downloading XTTS v2.0 checkpoint!")
+        ModelManager._download_model_files(
+            [XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CONFIG_FILE):
+        print(" > Downloading XTTS v2.0 config!")
+        ModelManager._download_model_files(
+            [XTTS_CONFIG_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+
+    # init args and config
+    model_args = GPTArgs(
+        max_conditioning_length=132300,  # 6 secs
+        min_conditioning_length=11025,  # 0.5 secs
+        debug_loading_failures=False,
+        max_wav_length=max_audio_length,  # ~11.6 seconds
+        max_text_length=max_text_length,
+        mel_norm_file=MEL_NORM_FILE,
+        dvae_checkpoint=DVAE_CHECKPOINT,
+        xtts_checkpoint=XTTS_CHECKPOINT,  # checkpoint path of the model that you want to fine-tune
+        tokenizer_file=TOKENIZER_FILE,
+        gpt_num_audio_tokens=1026,
+        gpt_start_audio_token=1024,
+        gpt_stop_audio_token=1025,
+        gpt_use_masking_gt_prompt_approach=True,
+        gpt_use_perceiver_resampler=True,
+    )
+    # define audio config
+    audio_config = XttsAudioConfig(sample_rate=22050, dvae_sample_rate=22050, output_sample_rate=24000)
+    # training parameters config
+
+    config = GPTTrainerConfig()
+
+    config.load_json(XTTS_CONFIG_FILE)
+
+    config.epochs = num_epochs
+    config.output_path = OUT_PATH
+    config.model_args = model_args
+    config.run_name = RUN_NAME
+    config.project_name = PROJECT_NAME
+    config.run_description = """
+        GPT XTTS training
+        """,
+    config.dashboard_logger = DASHBOARD_LOGGER
+    config.logger_uri = LOGGER_URI
+    config.audio = audio_config
+    config.batch_size = BATCH_SIZE
+    config.num_loader_workers = 8
+    config.eval_split_max_size = 256
+    config.print_step = 50
+    config.plot_step = 100
+    config.log_model_step = 100
+    config.save_step = save_step
+    config.save_n_checkpoints = 1
+    config.save_checkpoints = True
+    config.print_eval = False
+    config.optimizer = "AdamW"
+    config.optimizer_wd_only_on_weights = OPTIMIZER_WD_ONLY_ON_WEIGHTS
+    config.optimizer_params = {"betas": [0.9, 0.96], "eps": 1e-8, "weight_decay": weight_decay}
+    config.lr = lr
+    config.lr_scheduler = "MultiStepLR"
+    config.lr_scheduler_params = {"milestones": [50000 * 18, 150000 * 18, 300000 * 18], "gamma": 0.5, "last_epoch": -1}
+    config.test_sentences = []
+
+    # init the model from config
+    model = GPTTrainer.init_from_config(config)
+
+    # load training samples
+    train_samples, eval_samples = load_tts_samples(
+        DATASETS_CONFIG_LIST,
+        eval_split=True,
+        eval_split_max_size=config.eval_split_max_size,
+        eval_split_size=config.eval_split_size,
+    )
+
+    # init the trainer and 🚀
+    trainer = Trainer(
+        TrainerArgs(
+            restore_path=None,  # xtts checkpoint is restored via xtts_checkpoint key so no need of restore it using Trainer restore_path parameter
+            skip_train_epoch=False,
+            start_with_eval=START_WITH_EVAL,
+            grad_accum_steps=GRAD_ACUMM_STEPS
+        ),
+        config,
+        output_path=os.path.join(output_path, "run", "training"),
+        model=model,
+        train_samples=train_samples,
+        eval_samples=eval_samples,
+    )
+    trainer.fit()
+
+    # get the longest text audio file to use as speaker reference
+    samples_len = [len(item["text"].split(" ")) for item in train_samples]
+    longest_text_idx =  samples_len.index(max(samples_len))
+    speaker_ref = train_samples[longest_text_idx]["audio_file"]
+
+    trainer_out_path = trainer.output_path
+
+    # deallocate VRAM and RAM
+    del model, trainer, train_samples, eval_samples
+    gc.collect()
+
+    return trainer_out_path
+
+if __name__ == "__main__":
+    parser = create_xtts_trainer_parser()
+    args = parser.parse_args()
+
+    trainer_out_path = train_gpt(
+        metadatas=args.metadatas,
+        output_path=args.output_path,
+        num_epochs=args.num_epochs,
+        batch_size=args.batch_size,
+        grad_acumm=args.grad_acumm,
+        weight_decay=args.weight_decay,
+        lr=args.lr,
+        max_text_length=args.max_text_length,
+        max_audio_length=args.max_audio_length,
+        save_step=args.save_step
+    )
+
+    print(f"Checkpoint saved in dir: {trainer_out_path}")

checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/trainer_0_log.txt

@@ -0,0 +1,43 @@
+ > Training Environment:
+ | > Backend: Torch
+ | > Mixed precision: False
+ | > Precision: float32
+ | > Current device: 0
+ | > Num. of GPUs: 1
+ | > Num. of CPUs: 192
+ | > Num. of Torch Threads: 1
+ | > Torch seed: 54321
+ | > Torch CUDNN: True
+ | > Torch CUDNN deterministic: False
+ | > Torch CUDNN benchmark: False
+ | > Torch TF32 MatMul: False
+ > Start Tensorboard: tensorboard --logdir=checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3
+
+ > Model has 526224149 parameters
+
+ > EPOCH: 0/1
+ --> checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3
+
+ > TRAINING (2025-04-02 09:12:39) 
+
+   --> TIME: 2025-04-02 09:12:42 -- STEP: 0/6206 -- GLOBAL_STEP: 0
+     | > loss_text_ce: 0.2279348075389862  (0.2279348075389862)
+     | > loss_mel_ce: 5.657868385314941  (5.657868385314941)
+     | > loss: 1.4714508056640625  (1.4714508056640625)
+     | > current_lr: 5e-06 
+     | > step_time: 1.4127  (1.4127211570739746)
+     | > loader_time: 1.1588  (1.158794641494751)
+
+
+   --> TIME: 2025-04-02 09:13:03 -- STEP: 50/6206 -- GLOBAL_STEP: 50
+     | > loss_text_ce: 0.1972172111272812  (0.217392498254776)
+     | > loss_mel_ce: 5.164442539215088  (5.290682277679443)
+     | > loss: 1.3404148817062378  (1.3770187044143674)
+     | > current_lr: 5e-06 
+     | > step_time: 0.3686  (0.30237638473510753)
+     | > loader_time: 0.0186  (0.017733364105224608)
+
+ > Keyboard interrupt detected.
+ > Saving model before exiting...
+
+ > CHECKPOINT : checkpoints/GPT_XTTS_FT-April-02-2025_09+12AM-8e59ec3/checkpoint_90.pth

checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/checkpoint_96.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29a36ff22e5344a2f0350d7189ca45fcc1554be370fd83bdc7a67280dc337206
+size 4282236096

checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 20,
+    "batch_size": 32,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 2e-05,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/events.out.tfevents.1743585338.394654dd939a.16244.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a1a9a4e324333dd0e530eecb9c46f7f3f8842f874bd1432dc32af4714b9777f
+size 15577

checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/train_gpt_xtts.py

@@ -0,0 +1,237 @@
+import os
+import gc
+
+from trainer import Trainer, TrainerArgs
+
+from TTS.config.shared_configs import BaseDatasetConfig
+from TTS.tts.datasets import load_tts_samples
+from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrainerConfig, XttsAudioConfig
+from TTS.utils.manage import ModelManager
+
+from dataclasses import dataclass, field
+from typing import Optional
+from transformers import HfArgumentParser
+
+import argparse
+
+def create_xtts_trainer_parser():
+    parser = argparse.ArgumentParser(description="Arguments for XTTS Trainer")
+
+    parser.add_argument("--output_path", type=str, required=True,
+                        help="Path to pretrained + checkpoint model")
+    parser.add_argument("--metadatas", nargs='+', type=str, required=True,
+                        help="train_csv_path,eval_csv_path,language")
+    parser.add_argument("--num_epochs", type=int, default=1,
+                        help="Number of epochs")
+    parser.add_argument("--batch_size", type=int, default=1,
+                        help="Mini batch size")
+    parser.add_argument("--grad_acumm", type=int, default=1,
+                        help="Grad accumulation steps")
+    parser.add_argument("--max_audio_length", type=int, default=255995,
+                        help="Max audio length")
+    parser.add_argument("--max_text_length", type=int, default=200,
+                        help="Max text length")
+    parser.add_argument("--weight_decay", type=float, default=1e-2,
+                        help="Weight decay")
+    parser.add_argument("--lr", type=float, default=5e-6,
+                        help="Learning rate")
+    parser.add_argument("--save_step", type=int, default=5000,
+                        help="Save step")
+
+    return parser
+
+
+
+def train_gpt(metadatas, num_epochs, batch_size, grad_acumm, output_path, max_audio_length, max_text_length, lr, weight_decay, save_step):
+    #  Logging parameters
+    RUN_NAME = "GPT_XTTS_FT"
+    PROJECT_NAME = "XTTS_trainer"
+    DASHBOARD_LOGGER = "tensorboard"
+    LOGGER_URI = None
+
+    # Set here the path that the checkpoints will be saved. Default: ./run/training/
+    # OUT_PATH = os.path.join(output_path, "run", "training")
+    OUT_PATH = output_path
+
+    # Training Parameters
+    OPTIMIZER_WD_ONLY_ON_WEIGHTS = True  # for multi-gpu training please make it False
+    START_WITH_EVAL = False  # if True it will star with evaluation
+    BATCH_SIZE = batch_size  # set here the batch size
+    GRAD_ACUMM_STEPS = grad_acumm  # set here the grad accumulation steps
+
+
+    # Define here the dataset that you want to use for the fine-tuning on.
+    DATASETS_CONFIG_LIST = []
+    for metadata in metadatas:
+        train_csv, eval_csv, language = metadata.split(",")
+        print(train_csv, eval_csv, language)
+
+        config_dataset = BaseDatasetConfig(
+            formatter="coqui",
+            dataset_name="ft_dataset",
+            path=os.path.dirname(train_csv),
+            meta_file_train=os.path.basename(train_csv),
+            meta_file_val=os.path.basename(eval_csv),
+            language=language,
+        )
+
+        DATASETS_CONFIG_LIST.append(config_dataset)
+
+    # Define the path where XTTS v2.0.1 files will be downloaded
+    CHECKPOINTS_OUT_PATH = os.path.join(OUT_PATH, "XTTS_v2.0_original_model_files/")
+    os.makedirs(CHECKPOINTS_OUT_PATH, exist_ok=True)
+
+
+    # DVAE files
+    DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/dvae.pth"
+    MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/mel_stats.pth"
+
+    # Set the path to the downloaded files
+    DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(DVAE_CHECKPOINT_LINK))
+    MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(MEL_NORM_LINK))
+
+    # download DVAE files if needed
+    if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
+        print(" > Downloading DVAE files!")
+        ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)
+
+
+    # Download XTTS v2.0 checkpoint if needed
+    TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json"
+    XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth"
+    XTTS_CONFIG_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json"
+
+    # XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
+    TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(TOKENIZER_FILE_LINK))  # vocab.json file
+    XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CHECKPOINT_LINK))  # model.pth file
+    XTTS_CONFIG_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CONFIG_LINK))  # config.json file
+
+    # download XTTS v2.0 files if needed
+    if not os.path.isfile(TOKENIZER_FILE):
+        print(" > Downloading XTTS v2.0 tokenizer!")
+        ModelManager._download_model_files(
+            [TOKENIZER_FILE_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CHECKPOINT):
+        print(" > Downloading XTTS v2.0 checkpoint!")
+        ModelManager._download_model_files(
+            [XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CONFIG_FILE):
+        print(" > Downloading XTTS v2.0 config!")
+        ModelManager._download_model_files(
+            [XTTS_CONFIG_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+
+    # init args and config
+    model_args = GPTArgs(
+        max_conditioning_length=132300,  # 6 secs
+        min_conditioning_length=11025,  # 0.5 secs
+        debug_loading_failures=False,
+        max_wav_length=max_audio_length,  # ~11.6 seconds
+        max_text_length=max_text_length,
+        mel_norm_file=MEL_NORM_FILE,
+        dvae_checkpoint=DVAE_CHECKPOINT,
+        xtts_checkpoint=XTTS_CHECKPOINT,  # checkpoint path of the model that you want to fine-tune
+        tokenizer_file=TOKENIZER_FILE,
+        gpt_num_audio_tokens=1026,
+        gpt_start_audio_token=1024,
+        gpt_stop_audio_token=1025,
+        gpt_use_masking_gt_prompt_approach=True,
+        gpt_use_perceiver_resampler=True,
+    )
+    # define audio config
+    audio_config = XttsAudioConfig(sample_rate=22050, dvae_sample_rate=22050, output_sample_rate=24000)
+    # training parameters config
+
+    config = GPTTrainerConfig()
+
+    config.load_json(XTTS_CONFIG_FILE)
+
+    config.epochs = num_epochs
+    config.output_path = OUT_PATH
+    config.model_args = model_args
+    config.run_name = RUN_NAME
+    config.project_name = PROJECT_NAME
+    config.run_description = """
+        GPT XTTS training
+        """,
+    config.dashboard_logger = DASHBOARD_LOGGER
+    config.logger_uri = LOGGER_URI
+    config.audio = audio_config
+    config.batch_size = BATCH_SIZE
+    config.num_loader_workers = 8
+    config.eval_split_max_size = 256
+    config.print_step = 50
+    config.plot_step = 100
+    config.log_model_step = 100
+    config.save_step = save_step
+    config.save_n_checkpoints = 1
+    config.save_checkpoints = True
+    config.print_eval = False
+    config.optimizer = "AdamW"
+    config.optimizer_wd_only_on_weights = OPTIMIZER_WD_ONLY_ON_WEIGHTS
+    config.optimizer_params = {"betas": [0.9, 0.96], "eps": 1e-8, "weight_decay": weight_decay}
+    config.lr = lr
+    config.lr_scheduler = "MultiStepLR"
+    config.lr_scheduler_params = {"milestones": [50000 * 18, 150000 * 18, 300000 * 18], "gamma": 0.5, "last_epoch": -1}
+    config.test_sentences = []
+
+    # init the model from config
+    model = GPTTrainer.init_from_config(config)
+
+    # load training samples
+    train_samples, eval_samples = load_tts_samples(
+        DATASETS_CONFIG_LIST,
+        eval_split=True,
+        eval_split_max_size=config.eval_split_max_size,
+        eval_split_size=config.eval_split_size,
+    )
+
+    # init the trainer and 🚀
+    trainer = Trainer(
+        TrainerArgs(
+            restore_path=None,  # xtts checkpoint is restored via xtts_checkpoint key so no need of restore it using Trainer restore_path parameter
+            skip_train_epoch=False,
+            start_with_eval=START_WITH_EVAL,
+            grad_accum_steps=GRAD_ACUMM_STEPS
+        ),
+        config,
+        output_path=os.path.join(output_path, "run", "training"),
+        model=model,
+        train_samples=train_samples,
+        eval_samples=eval_samples,
+    )
+    trainer.fit()
+
+    # get the longest text audio file to use as speaker reference
+    samples_len = [len(item["text"].split(" ")) for item in train_samples]
+    longest_text_idx =  samples_len.index(max(samples_len))
+    speaker_ref = train_samples[longest_text_idx]["audio_file"]
+
+    trainer_out_path = trainer.output_path
+
+    # deallocate VRAM and RAM
+    del model, trainer, train_samples, eval_samples
+    gc.collect()
+
+    return trainer_out_path
+
+if __name__ == "__main__":
+    parser = create_xtts_trainer_parser()
+    args = parser.parse_args()
+
+    trainer_out_path = train_gpt(
+        metadatas=args.metadatas,
+        output_path=args.output_path,
+        num_epochs=args.num_epochs,
+        batch_size=args.batch_size,
+        grad_acumm=args.grad_acumm,
+        weight_decay=args.weight_decay,
+        lr=args.lr,
+        max_text_length=args.max_text_length,
+        max_audio_length=args.max_audio_length,
+        save_step=args.save_step
+    )
+
+    print(f"Checkpoint saved in dir: {trainer_out_path}")

checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/trainer_0_log.txt

@@ -0,0 +1,43 @@
+ > Training Environment:
+ | > Backend: Torch
+ | > Mixed precision: False
+ | > Precision: float32
+ | > Current device: 0
+ | > Num. of GPUs: 1
+ | > Num. of CPUs: 192
+ | > Num. of Torch Threads: 1
+ | > Torch seed: 54321
+ | > Torch CUDNN: True
+ | > Torch CUDNN deterministic: False
+ | > Torch CUDNN benchmark: False
+ | > Torch TF32 MatMul: False
+ > Start Tensorboard: tensorboard --logdir=checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3
+
+ > Model has 526224149 parameters
+
+ > EPOCH: 0/20
+ --> checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3
+
+ > TRAINING (2025-04-02 09:15:40) 
+
+   --> TIME: 2025-04-02 09:15:42 -- STEP: 0/3103 -- GLOBAL_STEP: 0
+     | > loss_text_ce: 0.24508634209632874  (0.24508634209632874)
+     | > loss_mel_ce: 5.678146839141846  (5.678146839141846)
+     | > loss: 1.4808082580566406  (1.4808082580566406)
+     | > current_lr: 2e-05 
+     | > step_time: 1.2583  (1.2583298683166504)
+     | > loader_time: 1.2637  (1.2637250423431396)
+
+
+   --> TIME: 2025-04-02 09:16:27 -- STEP: 50/3103 -- GLOBAL_STEP: 50
+     | > loss_text_ce: 0.175664022564888  (0.2083575162291527)
+     | > loss_mel_ce: 4.82735538482666  (5.0298005485534665)
+     | > loss: 1.2507548332214355  (1.30953950881958)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5013  (0.6319035625457766)
+     | > loader_time: 0.0204  (0.022382225990295407)
+
+ > Keyboard interrupt detected.
+ > Saving model before exiting...
+
+ > CHECKPOINT : checkpoints/GPT_XTTS_FT-April-02-2025_09+15AM-8e59ec3/checkpoint_96.pth

checkpoints/GPT_XTTS_FT-April-02-2025_09+17AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 20,
+    "batch_size": 32,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 2e-05,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+17AM-8e59ec3/events.out.tfevents.1743585455.394654dd939a.16991.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c49d4e0da42ff2e9406a9b7dc35457d9a0d7034c80d6d46b03cb943cdd5e143f
+size 88

checkpoints/GPT_XTTS_FT-April-02-2025_09+17AM-8e59ec3/trainer_0_log.txt

@@ -0,0 +1,14 @@
+ > Training Environment:
+ | > Backend: Torch
+ | > Mixed precision: False
+ | > Precision: float32
+ | > Current device: 0
+ | > Num. of GPUs: 1
+ | > Num. of CPUs: 192
+ | > Num. of Torch Threads: 1
+ | > Torch seed: 54321
+ | > Torch CUDNN: True
+ | > Torch CUDNN deterministic: False
+ | > Torch CUDNN benchmark: False
+ | > Torch TF32 MatMul: False
+ > Start Tensorboard: tensorboard --logdir=checkpoints/GPT_XTTS_FT-April-02-2025_09+17AM-8e59ec3

checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/checkpoint_1048.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f95156da01f00f7211275fc3ca997c96539524dc1dcb3251a4be067facfc705a
+size 3807986387

checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 50,
+    "batch_size": 32,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 2e-05,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/events.out.tfevents.1743585491.394654dd939a.17232.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdf2aa7905439cbfb42184c7f92a7750513b8757ecfae7db016688d991c2b6bb
+size 19381

checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/train_gpt_xtts.py

@@ -0,0 +1,237 @@
+import os
+import gc
+
+from trainer import Trainer, TrainerArgs
+
+from TTS.config.shared_configs import BaseDatasetConfig
+from TTS.tts.datasets import load_tts_samples
+from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrainerConfig, XttsAudioConfig
+from TTS.utils.manage import ModelManager
+
+from dataclasses import dataclass, field
+from typing import Optional
+from transformers import HfArgumentParser
+
+import argparse
+
+def create_xtts_trainer_parser():
+    parser = argparse.ArgumentParser(description="Arguments for XTTS Trainer")
+
+    parser.add_argument("--output_path", type=str, required=True,
+                        help="Path to pretrained + checkpoint model")
+    parser.add_argument("--metadatas", nargs='+', type=str, required=True,
+                        help="train_csv_path,eval_csv_path,language")
+    parser.add_argument("--num_epochs", type=int, default=1,
+                        help="Number of epochs")
+    parser.add_argument("--batch_size", type=int, default=1,
+                        help="Mini batch size")
+    parser.add_argument("--grad_acumm", type=int, default=1,
+                        help="Grad accumulation steps")
+    parser.add_argument("--max_audio_length", type=int, default=255995,
+                        help="Max audio length")
+    parser.add_argument("--max_text_length", type=int, default=200,
+                        help="Max text length")
+    parser.add_argument("--weight_decay", type=float, default=1e-2,
+                        help="Weight decay")
+    parser.add_argument("--lr", type=float, default=5e-6,
+                        help="Learning rate")
+    parser.add_argument("--save_step", type=int, default=5000,
+                        help="Save step")
+
+    return parser
+
+
+
+def train_gpt(metadatas, num_epochs, batch_size, grad_acumm, output_path, max_audio_length, max_text_length, lr, weight_decay, save_step):
+    #  Logging parameters
+    RUN_NAME = "GPT_XTTS_FT"
+    PROJECT_NAME = "XTTS_trainer"
+    DASHBOARD_LOGGER = "tensorboard"
+    LOGGER_URI = None
+
+    # Set here the path that the checkpoints will be saved. Default: ./run/training/
+    # OUT_PATH = os.path.join(output_path, "run", "training")
+    OUT_PATH = output_path
+
+    # Training Parameters
+    OPTIMIZER_WD_ONLY_ON_WEIGHTS = True  # for multi-gpu training please make it False
+    START_WITH_EVAL = False  # if True it will star with evaluation
+    BATCH_SIZE = batch_size  # set here the batch size
+    GRAD_ACUMM_STEPS = grad_acumm  # set here the grad accumulation steps
+
+
+    # Define here the dataset that you want to use for the fine-tuning on.
+    DATASETS_CONFIG_LIST = []
+    for metadata in metadatas:
+        train_csv, eval_csv, language = metadata.split(",")
+        print(train_csv, eval_csv, language)
+
+        config_dataset = BaseDatasetConfig(
+            formatter="coqui",
+            dataset_name="ft_dataset",
+            path=os.path.dirname(train_csv),
+            meta_file_train=os.path.basename(train_csv),
+            meta_file_val=os.path.basename(eval_csv),
+            language=language,
+        )
+
+        DATASETS_CONFIG_LIST.append(config_dataset)
+
+    # Define the path where XTTS v2.0.1 files will be downloaded
+    CHECKPOINTS_OUT_PATH = os.path.join(OUT_PATH, "XTTS_v2.0_original_model_files/")
+    os.makedirs(CHECKPOINTS_OUT_PATH, exist_ok=True)
+
+
+    # DVAE files
+    DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/dvae.pth"
+    MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/mel_stats.pth"
+
+    # Set the path to the downloaded files
+    DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(DVAE_CHECKPOINT_LINK))
+    MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(MEL_NORM_LINK))
+
+    # download DVAE files if needed
+    if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
+        print(" > Downloading DVAE files!")
+        ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)
+
+
+    # Download XTTS v2.0 checkpoint if needed
+    TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json"
+    XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth"
+    XTTS_CONFIG_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json"
+
+    # XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
+    TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(TOKENIZER_FILE_LINK))  # vocab.json file
+    XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CHECKPOINT_LINK))  # model.pth file
+    XTTS_CONFIG_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CONFIG_LINK))  # config.json file
+
+    # download XTTS v2.0 files if needed
+    if not os.path.isfile(TOKENIZER_FILE):
+        print(" > Downloading XTTS v2.0 tokenizer!")
+        ModelManager._download_model_files(
+            [TOKENIZER_FILE_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CHECKPOINT):
+        print(" > Downloading XTTS v2.0 checkpoint!")
+        ModelManager._download_model_files(
+            [XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CONFIG_FILE):
+        print(" > Downloading XTTS v2.0 config!")
+        ModelManager._download_model_files(
+            [XTTS_CONFIG_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+
+    # init args and config
+    model_args = GPTArgs(
+        max_conditioning_length=132300,  # 6 secs
+        min_conditioning_length=11025,  # 0.5 secs
+        debug_loading_failures=False,
+        max_wav_length=max_audio_length,  # ~11.6 seconds
+        max_text_length=max_text_length,
+        mel_norm_file=MEL_NORM_FILE,
+        dvae_checkpoint=DVAE_CHECKPOINT,
+        xtts_checkpoint=XTTS_CHECKPOINT,  # checkpoint path of the model that you want to fine-tune
+        tokenizer_file=TOKENIZER_FILE,
+        gpt_num_audio_tokens=1026,
+        gpt_start_audio_token=1024,
+        gpt_stop_audio_token=1025,
+        gpt_use_masking_gt_prompt_approach=True,
+        gpt_use_perceiver_resampler=True,
+    )
+    # define audio config
+    audio_config = XttsAudioConfig(sample_rate=22050, dvae_sample_rate=22050, output_sample_rate=24000)
+    # training parameters config
+
+    config = GPTTrainerConfig()
+
+    config.load_json(XTTS_CONFIG_FILE)
+
+    config.epochs = num_epochs
+    config.output_path = OUT_PATH
+    config.model_args = model_args
+    config.run_name = RUN_NAME
+    config.project_name = PROJECT_NAME
+    config.run_description = """
+        GPT XTTS training
+        """,
+    config.dashboard_logger = DASHBOARD_LOGGER
+    config.logger_uri = LOGGER_URI
+    config.audio = audio_config
+    config.batch_size = BATCH_SIZE
+    config.num_loader_workers = 8
+    config.eval_split_max_size = 256
+    config.print_step = 50
+    config.plot_step = 100
+    config.log_model_step = 100
+    config.save_step = save_step
+    config.save_n_checkpoints = 1
+    config.save_checkpoints = True
+    config.print_eval = False
+    config.optimizer = "AdamW"
+    config.optimizer_wd_only_on_weights = OPTIMIZER_WD_ONLY_ON_WEIGHTS
+    config.optimizer_params = {"betas": [0.9, 0.96], "eps": 1e-8, "weight_decay": weight_decay}
+    config.lr = lr
+    config.lr_scheduler = "MultiStepLR"
+    config.lr_scheduler_params = {"milestones": [50000 * 18, 150000 * 18, 300000 * 18], "gamma": 0.5, "last_epoch": -1}
+    config.test_sentences = []
+
+    # init the model from config
+    model = GPTTrainer.init_from_config(config)
+
+    # load training samples
+    train_samples, eval_samples = load_tts_samples(
+        DATASETS_CONFIG_LIST,
+        eval_split=True,
+        eval_split_max_size=config.eval_split_max_size,
+        eval_split_size=config.eval_split_size,
+    )
+
+    # init the trainer and 🚀
+    trainer = Trainer(
+        TrainerArgs(
+            restore_path=None,  # xtts checkpoint is restored via xtts_checkpoint key so no need of restore it using Trainer restore_path parameter
+            skip_train_epoch=False,
+            start_with_eval=START_WITH_EVAL,
+            grad_accum_steps=GRAD_ACUMM_STEPS
+        ),
+        config,
+        output_path=os.path.join(output_path, "run", "training"),
+        model=model,
+        train_samples=train_samples,
+        eval_samples=eval_samples,
+    )
+    trainer.fit()
+
+    # get the longest text audio file to use as speaker reference
+    samples_len = [len(item["text"].split(" ")) for item in train_samples]
+    longest_text_idx =  samples_len.index(max(samples_len))
+    speaker_ref = train_samples[longest_text_idx]["audio_file"]
+
+    trainer_out_path = trainer.output_path
+
+    # deallocate VRAM and RAM
+    del model, trainer, train_samples, eval_samples
+    gc.collect()
+
+    return trainer_out_path
+
+if __name__ == "__main__":
+    parser = create_xtts_trainer_parser()
+    args = parser.parse_args()
+
+    trainer_out_path = train_gpt(
+        metadatas=args.metadatas,
+        output_path=args.output_path,
+        num_epochs=args.num_epochs,
+        batch_size=args.batch_size,
+        grad_acumm=args.grad_acumm,
+        weight_decay=args.weight_decay,
+        lr=args.lr,
+        max_text_length=args.max_text_length,
+        max_audio_length=args.max_audio_length,
+        save_step=args.save_step
+    )
+
+    print(f"Checkpoint saved in dir: {trainer_out_path}")

checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/trainer_0_log.txt

@@ -0,0 +1,214 @@
+ > Training Environment:
+ | > Backend: Torch
+ | > Mixed precision: False
+ | > Precision: float32
+ | > Current device: 0
+ | > Num. of GPUs: 1
+ | > Num. of CPUs: 192
+ | > Num. of Torch Threads: 1
+ | > Torch seed: 54321
+ | > Torch CUDNN: True
+ | > Torch CUDNN deterministic: False
+ | > Torch CUDNN benchmark: False
+ | > Torch TF32 MatMul: False
+ > Start Tensorboard: tensorboard --logdir=checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3
+
+ > Model has 526224149 parameters
+
+ > EPOCH: 0/50
+ --> checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3
+
+ > TRAINING (2025-04-02 09:18:12) 
+
+   --> TIME: 2025-04-02 09:18:15 -- STEP: 0/3103 -- GLOBAL_STEP: 0
+     | > loss_text_ce: 0.22596414387226105  (0.22596414387226105)
+     | > loss_mel_ce: 5.678620338439941  (5.678620338439941)
+     | > loss: 1.4761461019515991  (1.4761461019515991)
+     | > current_lr: 2e-05 
+     | > step_time: 1.1725  (1.1725168228149414)
+     | > loader_time: 1.4278  (1.4277825355529785)
+
+
+   --> TIME: 2025-04-02 09:19:00 -- STEP: 50/3103 -- GLOBAL_STEP: 50
+     | > loss_text_ce: 0.1635490506887436  (0.1931218272447586)
+     | > loss_mel_ce: 4.82426118850708  (5.039961290359496)
+     | > loss: 1.2469525337219238  (1.3082707810401921)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5022  (0.6329782819747927)
+     | > loader_time: 0.0224  (0.023704633712768555)
+
+
+   --> TIME: 2025-04-02 09:19:44 -- STEP: 100/3103 -- GLOBAL_STEP: 100
+     | > loss_text_ce: 0.11297561228275299  (0.16677283406257634)
+     | > loss_mel_ce: 4.75046443939209  (4.901695904731754)
+     | > loss: 1.2158600091934204  (1.2671171844005593)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6213  (0.6289737367630006)
+     | > loader_time: 0.0187  (0.023205628395080568)
+
+
+   --> TIME: 2025-04-02 09:20:29 -- STEP: 150/3103 -- GLOBAL_STEP: 150
+     | > loss_text_ce: 0.0975332111120224  (0.14623819912473357)
+     | > loss_mel_ce: 4.747553825378418  (4.825411011377972)
+     | > loss: 1.2112717628479004  (1.242912301222484)
+     | > current_lr: 2e-05 
+     | > step_time: 0.753  (0.6259132862091067)
+     | > loader_time: 0.0208  (0.022733476956685386)
+
+
+   --> TIME: 2025-04-02 09:21:14 -- STEP: 200/3103 -- GLOBAL_STEP: 200
+     | > loss_text_ce: 0.08699331432580948  (0.13224070142954575)
+     | > loss_mel_ce: 4.600005626678467  (4.778431077003483)
+     | > loss: 1.1717497110366821  (1.2276679438352596)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5184  (0.6275241088867191)
+     | > loader_time: 0.018  (0.022192956209182744)
+
+
+   --> TIME: 2025-04-02 09:21:58 -- STEP: 250/3103 -- GLOBAL_STEP: 250
+     | > loss_text_ce: 0.0748153030872345  (0.12184768083691583)
+     | > loss_mel_ce: 4.631954193115234  (4.739824426651005)
+     | > loss: 1.1766923666000366  (1.2154180264472971)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6806  (0.628249227523804)
+     | > loader_time: 0.0189  (0.022299116134643557)
+
+
+   --> TIME: 2025-04-02 09:22:43 -- STEP: 300/3103 -- GLOBAL_STEP: 300
+     | > loss_text_ce: 0.06787222623825073  (0.11369345329701891)
+     | > loss_mel_ce: 4.513082027435303  (4.70801657676697)
+     | > loss: 1.1452385187149048  (1.205427505970002)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6019  (0.6278318238258366)
+     | > loader_time: 0.0178  (0.021908087730407713)
+
+
+   --> TIME: 2025-04-02 09:23:27 -- STEP: 350/3103 -- GLOBAL_STEP: 350
+     | > loss_text_ce: 0.0629011020064354  (0.10694028889494275)
+     | > loss_mel_ce: 4.57236909866333  (4.678290811266221)
+     | > loss: 1.1588175296783447  (1.1963077746118822)
+     | > current_lr: 2e-05 
+     | > step_time: 0.7106  (0.627729578018189)
+     | > loader_time: 0.0206  (0.02201144695281982)
+
+
+   --> TIME: 2025-04-02 09:24:13 -- STEP: 400/3103 -- GLOBAL_STEP: 400
+     | > loss_text_ce: 0.06310830265283585  (0.10152350591495626)
+     | > loss_mel_ce: 4.429666996002197  (4.649858270883559)
+     | > loss: 1.1231938600540161  (1.1878454446792603)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6256  (0.6287157744169236)
+     | > loader_time: 0.0206  (0.021938490271568304)
+
+
+   --> TIME: 2025-04-02 09:24:57 -- STEP: 450/3103 -- GLOBAL_STEP: 450
+     | > loss_text_ce: 0.06332644075155258  (0.09703049760725757)
+     | > loss_mel_ce: 4.434024333953857  (4.627943340937296)
+     | > loss: 1.1243376731872559  (1.1812434607081939)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5987  (0.628130240970188)
+     | > loader_time: 0.0194  (0.021769250763787166)
+
+
+   --> TIME: 2025-04-02 09:25:41 -- STEP: 500/3103 -- GLOBAL_STEP: 500
+     | > loss_text_ce: 0.057986967265605927  (0.09330656228959555)
+     | > loss_mel_ce: 4.323246479034424  (4.599339053153992)
+     | > loss: 1.0953083038330078  (1.1731614048480983)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5302  (0.6275670042037966)
+     | > loader_time: 0.0156  (0.021510847568511966)
+
+
+   --> TIME: 2025-04-02 09:26:25 -- STEP: 550/3103 -- GLOBAL_STEP: 550
+     | > loss_text_ce: 0.059549812227487564  (0.09016217888078902)
+     | > loss_mel_ce: 4.448150634765625  (4.574896022189747)
+     | > loss: 1.1269251108169556  (1.1662645506858824)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6817  (0.6266948769309306)
+     | > loader_time: 0.0261  (0.021388005343350495)
+
+
+   --> TIME: 2025-04-02 09:27:10 -- STEP: 600/3103 -- GLOBAL_STEP: 600
+     | > loss_text_ce: 0.053500913083553314  (0.08742072052011887)
+     | > loss_mel_ce: 4.334932804107666  (4.550314415295918)
+     | > loss: 1.0971084833145142  (1.1594337842861808)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6099  (0.6271455967426303)
+     | > loader_time: 0.0204  (0.021457889477411903)
+
+
+   --> TIME: 2025-04-02 09:27:54 -- STEP: 650/3103 -- GLOBAL_STEP: 650
+     | > loss_text_ce: 0.05324619635939598  (0.08500067840401943)
+     | > loss_mel_ce: 4.238860607147217  (4.527919439902667)
+     | > loss: 1.0730266571044922  (1.153230029436258)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6918  (0.6268535947799684)
+     | > loader_time: 0.0198  (0.021263816540057833)
+
+
+   --> TIME: 2025-04-02 09:28:39 -- STEP: 700/3103 -- GLOBAL_STEP: 700
+     | > loss_text_ce: 0.05458847060799599  (0.08288035475781985)
+     | > loss_mel_ce: 4.294866561889648  (4.506881997925891)
+     | > loss: 1.0873637199401855  (1.147440587963376)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5274  (0.6274660440853662)
+     | > loader_time: 0.0185  (0.021218415328434534)
+
+
+   --> TIME: 2025-04-02 09:29:24 -- STEP: 750/3103 -- GLOBAL_STEP: 750
+     | > loss_text_ce: 0.05593346431851387  (0.08095550758143269)
+     | > loss_mel_ce: 4.254919528961182  (4.48457054233551)
+     | > loss: 1.0777132511138916  (1.1413815121650688)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5765  (0.6276734336217242)
+     | > loader_time: 0.0189  (0.021116169293721513)
+
+
+   --> TIME: 2025-04-02 09:30:09 -- STEP: 800/3103 -- GLOBAL_STEP: 800
+     | > loss_text_ce: 0.05430208891630173  (0.07929737158119683)
+     | > loss_mel_ce: 4.200708389282227  (4.4646048179268805)
+     | > loss: 1.0637526512145996  (1.1359755472093815)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5275  (0.6276474064588544)
+     | > loader_time: 0.0191  (0.021008467972278594)
+
+
+   --> TIME: 2025-04-02 09:30:54 -- STEP: 850/3103 -- GLOBAL_STEP: 850
+     | > loss_text_ce: 0.05214960500597954  (0.07773133277454801)
+     | > loss_mel_ce: 4.283355236053467  (4.444886454694409)
+     | > loss: 1.0838762521743774  (1.1306544467982107)
+     | > current_lr: 2e-05 
+     | > step_time: 0.7254  (0.6280438235226797)
+     | > loader_time: 0.021  (0.02097405882442699)
+
+
+   --> TIME: 2025-04-02 09:31:38 -- STEP: 900/3103 -- GLOBAL_STEP: 900
+     | > loss_text_ce: 0.054857004433870316  (0.07632090492794914)
+     | > loss_mel_ce: 4.0890212059021  (4.425407789548235)
+     | > loss: 1.0359694957733154  (1.125432173477277)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5269  (0.627747194766998)
+     | > loader_time: 0.0183  (0.020858750873141818)
+
+
+   --> TIME: 2025-04-02 09:32:22 -- STEP: 950/3103 -- GLOBAL_STEP: 950
+     | > loss_text_ce: 0.049733009189367294  (0.0750451090147621)
+     | > loss_mel_ce: 4.017485618591309  (4.4084909547002615)
+     | > loss: 1.0168046951293945  (1.120884016124824)
+     | > current_lr: 2e-05 
+     | > step_time: 0.5768  (0.627245255771436)
+     | > loader_time: 0.0185  (0.020782094754670796)
+
+
+   --> TIME: 2025-04-02 09:33:05 -- STEP: 1000/3103 -- GLOBAL_STEP: 1000
+     | > loss_text_ce: 0.05076472461223602  (0.0738591817505658)
+     | > loss_mel_ce: 4.133179664611816  (4.3908965120315475)
+     | > loss: 1.0459860563278198  (1.1161889236569387)
+     | > current_lr: 2e-05 
+     | > step_time: 0.6261  (0.6264848773479459)
+     | > loader_time: 0.0211  (0.020686251640319824)
+
+ > Keyboard interrupt detected.
+ > Saving model before exiting...
+
+ > CHECKPOINT : checkpoints/GPT_XTTS_FT-April-02-2025_09+18AM-8e59ec3/checkpoint_1048.pth

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/best_model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dc21f21c76867234785eee1994e6a734a501a9fa0e4f3f6b8051dcab38d0ad7
+size 5701312341

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/best_model_65163.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dc21f21c76867234785eee1994e6a734a501a9fa0e4f3f6b8051dcab38d0ad7
+size 5701312341

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/checkpoint_90000.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46369e174d4f2b792662d81b38336dbb9c4b333878bc251c3a071df467878074
+size 5701312277

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/config.json

@@ -0,0 +1,193 @@
+{
+    "output_path": "checkpoints/",
+    "logger_uri": null,
+    "run_name": "GPT_XTTS_FT",
+    "project_name": "XTTS_trainer",
+    "run_description": [
+        "\n        GPT XTTS training\n        "
+    ],
+    "print_step": 50,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": 100,
+    "save_step": 10000,
+    "save_n_checkpoints": 1,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 30,
+    "batch_size": 32,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 2e-05,
+    "optimizer": "AdamW",
+    "optimizer_params": {
+        "betas": [
+            0.9,
+            0.96
+        ],
+        "eps": 1e-08,
+        "weight_decay": 0.01
+    },
+    "lr_scheduler": "MultiStepLR",
+    "lr_scheduler_params": {
+        "milestones": [
+            900000,
+            2700000,
+            5400000
+        ],
+        "gamma": 0.5,
+        "last_epoch": -1
+    },
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 8,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "dvae_sample_rate": 22050
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": 256,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": "",
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "checkpoints/XTTS_v2.0_original_model_files/vocab.json",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 10479,
+        "gpt_start_text_token": 261,
+        "gpt_stop_text_token": 0,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400,
+        "min_conditioning_length": 11025,
+        "max_conditioning_length": 132300,
+        "gpt_loss_text_ce_weight": 0.01,
+        "gpt_loss_mel_ce_weight": 1.0,
+        "debug_loading_failures": false,
+        "max_wav_length": 330750,
+        "max_text_length": 500,
+        "mel_norm_file": "checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth",
+        "dvae_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/dvae.pth",
+        "xtts_checkpoint": "checkpoints/XTTS_v2.0_original_model_files/model.pth",
+        "vocoder": ""
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false,
+    "optimizer_wd_only_on_weights": true,
+    "weighted_loss_attrs": null,
+    "weighted_loss_multipliers": null,
+    "github_branch": "* main"
+}

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/events.out.tfevents.1743586468.394654dd939a.19042.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77cec419e754362f65014b942711942a79e1de5f2baec6ae3103a95eb5473b04
+size 410385

checkpoints/GPT_XTTS_FT-April-02-2025_09+34AM-8e59ec3/train_gpt_xtts.py

@@ -0,0 +1,237 @@
+import os
+import gc
+
+from trainer import Trainer, TrainerArgs
+
+from TTS.config.shared_configs import BaseDatasetConfig
+from TTS.tts.datasets import load_tts_samples
+from TTS.tts.layers.xtts.trainer.gpt_trainer import GPTArgs, GPTTrainer, GPTTrainerConfig, XttsAudioConfig
+from TTS.utils.manage import ModelManager
+
+from dataclasses import dataclass, field
+from typing import Optional
+from transformers import HfArgumentParser
+
+import argparse
+
+def create_xtts_trainer_parser():
+    parser = argparse.ArgumentParser(description="Arguments for XTTS Trainer")
+
+    parser.add_argument("--output_path", type=str, required=True,
+                        help="Path to pretrained + checkpoint model")
+    parser.add_argument("--metadatas", nargs='+', type=str, required=True,
+                        help="train_csv_path,eval_csv_path,language")
+    parser.add_argument("--num_epochs", type=int, default=1,
+                        help="Number of epochs")
+    parser.add_argument("--batch_size", type=int, default=1,
+                        help="Mini batch size")
+    parser.add_argument("--grad_acumm", type=int, default=1,
+                        help="Grad accumulation steps")
+    parser.add_argument("--max_audio_length", type=int, default=255995,
+                        help="Max audio length")
+    parser.add_argument("--max_text_length", type=int, default=200,
+                        help="Max text length")
+    parser.add_argument("--weight_decay", type=float, default=1e-2,
+                        help="Weight decay")
+    parser.add_argument("--lr", type=float, default=5e-6,
+                        help="Learning rate")
+    parser.add_argument("--save_step", type=int, default=5000,
+                        help="Save step")
+
+    return parser
+
+
+
+def train_gpt(metadatas, num_epochs, batch_size, grad_acumm, output_path, max_audio_length, max_text_length, lr, weight_decay, save_step):
+    #  Logging parameters
+    RUN_NAME = "GPT_XTTS_FT"
+    PROJECT_NAME = "XTTS_trainer"
+    DASHBOARD_LOGGER = "tensorboard"
+    LOGGER_URI = None
+
+    # Set here the path that the checkpoints will be saved. Default: ./run/training/
+    # OUT_PATH = os.path.join(output_path, "run", "training")
+    OUT_PATH = output_path
+
+    # Training Parameters
+    OPTIMIZER_WD_ONLY_ON_WEIGHTS = True  # for multi-gpu training please make it False
+    START_WITH_EVAL = False  # if True it will star with evaluation
+    BATCH_SIZE = batch_size  # set here the batch size
+    GRAD_ACUMM_STEPS = grad_acumm  # set here the grad accumulation steps
+
+
+    # Define here the dataset that you want to use for the fine-tuning on.
+    DATASETS_CONFIG_LIST = []
+    for metadata in metadatas:
+        train_csv, eval_csv, language = metadata.split(",")
+        print(train_csv, eval_csv, language)
+
+        config_dataset = BaseDatasetConfig(
+            formatter="coqui",
+            dataset_name="ft_dataset",
+            path=os.path.dirname(train_csv),
+            meta_file_train=os.path.basename(train_csv),
+            meta_file_val=os.path.basename(eval_csv),
+            language=language,
+        )
+
+        DATASETS_CONFIG_LIST.append(config_dataset)
+
+    # Define the path where XTTS v2.0.1 files will be downloaded
+    CHECKPOINTS_OUT_PATH = os.path.join(OUT_PATH, "XTTS_v2.0_original_model_files/")
+    os.makedirs(CHECKPOINTS_OUT_PATH, exist_ok=True)
+
+
+    # DVAE files
+    DVAE_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/dvae.pth"
+    MEL_NORM_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/mel_stats.pth"
+
+    # Set the path to the downloaded files
+    DVAE_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(DVAE_CHECKPOINT_LINK))
+    MEL_NORM_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(MEL_NORM_LINK))
+
+    # download DVAE files if needed
+    if not os.path.isfile(DVAE_CHECKPOINT) or not os.path.isfile(MEL_NORM_FILE):
+        print(" > Downloading DVAE files!")
+        ModelManager._download_model_files([MEL_NORM_LINK, DVAE_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True)
+
+
+    # Download XTTS v2.0 checkpoint if needed
+    TOKENIZER_FILE_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/vocab.json"
+    XTTS_CHECKPOINT_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/model.pth"
+    XTTS_CONFIG_LINK = "https://coqui.gateway.scarf.sh/hf-coqui/XTTS-v2/main/config.json"
+
+    # XTTS transfer learning parameters: You we need to provide the paths of XTTS model checkpoint that you want to do the fine tuning.
+    TOKENIZER_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(TOKENIZER_FILE_LINK))  # vocab.json file
+    XTTS_CHECKPOINT = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CHECKPOINT_LINK))  # model.pth file
+    XTTS_CONFIG_FILE = os.path.join(CHECKPOINTS_OUT_PATH, os.path.basename(XTTS_CONFIG_LINK))  # config.json file
+
+    # download XTTS v2.0 files if needed
+    if not os.path.isfile(TOKENIZER_FILE):
+        print(" > Downloading XTTS v2.0 tokenizer!")
+        ModelManager._download_model_files(
+            [TOKENIZER_FILE_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CHECKPOINT):
+        print(" > Downloading XTTS v2.0 checkpoint!")
+        ModelManager._download_model_files(
+            [XTTS_CHECKPOINT_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+    if not os.path.isfile(XTTS_CONFIG_FILE):
+        print(" > Downloading XTTS v2.0 config!")
+        ModelManager._download_model_files(
+            [XTTS_CONFIG_LINK], CHECKPOINTS_OUT_PATH, progress_bar=True
+        )
+
+    # init args and config
+    model_args = GPTArgs(
+        max_conditioning_length=132300,  # 6 secs
+        min_conditioning_length=11025,  # 0.5 secs
+        debug_loading_failures=False,
+        max_wav_length=max_audio_length,  # ~11.6 seconds
+        max_text_length=max_text_length,
+        mel_norm_file=MEL_NORM_FILE,
+        dvae_checkpoint=DVAE_CHECKPOINT,
+        xtts_checkpoint=XTTS_CHECKPOINT,  # checkpoint path of the model that you want to fine-tune
+        tokenizer_file=TOKENIZER_FILE,
+        gpt_num_audio_tokens=1026,
+        gpt_start_audio_token=1024,
+        gpt_stop_audio_token=1025,
+        gpt_use_masking_gt_prompt_approach=True,
+        gpt_use_perceiver_resampler=True,
+    )
+    # define audio config
+    audio_config = XttsAudioConfig(sample_rate=22050, dvae_sample_rate=22050, output_sample_rate=24000)
+    # training parameters config
+
+    config = GPTTrainerConfig()
+
+    config.load_json(XTTS_CONFIG_FILE)
+
+    config.epochs = num_epochs
+    config.output_path = OUT_PATH
+    config.model_args = model_args
+    config.run_name = RUN_NAME
+    config.project_name = PROJECT_NAME
+    config.run_description = """
+        GPT XTTS training
+        """,
+    config.dashboard_logger = DASHBOARD_LOGGER
+    config.logger_uri = LOGGER_URI
+    config.audio = audio_config
+    config.batch_size = BATCH_SIZE
+    config.num_loader_workers = 8
+    config.eval_split_max_size = 256
+    config.print_step = 50
+    config.plot_step = 100
+    config.log_model_step = 100
+    config.save_step = save_step
+    config.save_n_checkpoints = 1
+    config.save_checkpoints = True
+    config.print_eval = False
+    config.optimizer = "AdamW"
+    config.optimizer_wd_only_on_weights = OPTIMIZER_WD_ONLY_ON_WEIGHTS
+    config.optimizer_params = {"betas": [0.9, 0.96], "eps": 1e-8, "weight_decay": weight_decay}
+    config.lr = lr
+    config.lr_scheduler = "MultiStepLR"
+    config.lr_scheduler_params = {"milestones": [50000 * 18, 150000 * 18, 300000 * 18], "gamma": 0.5, "last_epoch": -1}
+    config.test_sentences = []
+
+    # init the model from config
+    model = GPTTrainer.init_from_config(config)
+
+    # load training samples
+    train_samples, eval_samples = load_tts_samples(
+        DATASETS_CONFIG_LIST,
+        eval_split=True,
+        eval_split_max_size=config.eval_split_max_size,
+        eval_split_size=config.eval_split_size,
+    )
+
+    # init the trainer and 🚀
+    trainer = Trainer(
+        TrainerArgs(
+            restore_path=None,  # xtts checkpoint is restored via xtts_checkpoint key so no need of restore it using Trainer restore_path parameter
+            skip_train_epoch=False,
+            start_with_eval=START_WITH_EVAL,
+            grad_accum_steps=GRAD_ACUMM_STEPS
+        ),
+        config,
+        output_path=os.path.join(output_path, "run", "training"),
+        model=model,
+        train_samples=train_samples,
+        eval_samples=eval_samples,
+    )
+    trainer.fit()
+
+    # get the longest text audio file to use as speaker reference
+    samples_len = [len(item["text"].split(" ")) for item in train_samples]
+    longest_text_idx =  samples_len.index(max(samples_len))
+    speaker_ref = train_samples[longest_text_idx]["audio_file"]
+
+    trainer_out_path = trainer.output_path
+
+    # deallocate VRAM and RAM
+    del model, trainer, train_samples, eval_samples
+    gc.collect()
+
+    return trainer_out_path
+
+if __name__ == "__main__":
+    parser = create_xtts_trainer_parser()
+    args = parser.parse_args()
+
+    trainer_out_path = train_gpt(
+        metadatas=args.metadatas,
+        output_path=args.output_path,
+        num_epochs=args.num_epochs,
+        batch_size=args.batch_size,
+        grad_acumm=args.grad_acumm,
+        weight_decay=args.weight_decay,
+        lr=args.lr,
+        max_text_length=args.max_text_length,
+        max_audio_length=args.max_audio_length,
+        save_step=args.save_step
+    )
+
+    print(f"Checkpoint saved in dir: {trainer_out_path}")

checkpoints/XTTS_v2.0_original_model_files/config.json

@@ -0,0 +1,160 @@
+{
+    "output_path": "output",
+    "logger_uri": null,
+    "run_name": "run",
+    "project_name": null,
+    "run_description": "\ud83d\udc38Coqui trainer run.",
+    "print_step": 25,
+    "plot_step": 100,
+    "model_param_stats": false,
+    "wandb_entity": null,
+    "dashboard_logger": "tensorboard",
+    "save_on_interrupt": true,
+    "log_model_step": null,
+    "save_step": 10000,
+    "save_n_checkpoints": 5,
+    "save_checkpoints": true,
+    "save_all_best": false,
+    "save_best_after": 10000,
+    "target_loss": null,
+    "print_eval": false,
+    "test_delay_epochs": 0,
+    "run_eval": true,
+    "run_eval_steps": null,
+    "distributed_backend": "nccl",
+    "distributed_url": "tcp://localhost:54321",
+    "mixed_precision": false,
+    "precision": "fp16",
+    "epochs": 1000,
+    "batch_size": 32,
+    "eval_batch_size": 16,
+    "grad_clip": 0.0,
+    "scheduler_after_epoch": true,
+    "lr": 0.001,
+    "optimizer": "radam",
+    "optimizer_params": null,
+    "lr_scheduler": null,
+    "lr_scheduler_params": {},
+    "use_grad_scaler": false,
+    "allow_tf32": false,
+    "cudnn_enable": true,
+    "cudnn_deterministic": false,
+    "cudnn_benchmark": false,
+    "training_seed": 54321,
+    "model": "xtts",
+    "num_loader_workers": 0,
+    "num_eval_loader_workers": 0,
+    "use_noise_augment": false,
+    "audio": {
+        "sample_rate": 22050,
+        "output_sample_rate": 24000
+    },
+    "use_phonemes": false,
+    "phonemizer": null,
+    "phoneme_language": null,
+    "compute_input_seq_cache": false,
+    "text_cleaner": null,
+    "enable_eos_bos_chars": false,
+    "test_sentences_file": "",
+    "phoneme_cache_path": null,
+    "characters": null,
+    "add_blank": false,
+    "batch_group_size": 0,
+    "loss_masking": null,
+    "min_audio_len": 1,
+    "max_audio_len": Infinity,
+    "min_text_len": 1,
+    "max_text_len": Infinity,
+    "compute_f0": false,
+    "compute_energy": false,
+    "compute_linear_spec": false,
+    "precompute_num_workers": 0,
+    "start_by_longest": false,
+    "shuffle": false,
+    "drop_last": false,
+    "datasets": [
+        {
+            "formatter": "",
+            "dataset_name": "",
+            "path": "",
+            "meta_file_train": "",
+            "ignored_speakers": null,
+            "language": "",
+            "phonemizer": "",
+            "meta_file_val": "",
+            "meta_file_attn_mask": ""
+        }
+    ],
+    "test_sentences": [],
+    "eval_split_max_size": null,
+    "eval_split_size": 0.01,
+    "use_speaker_weighted_sampler": false,
+    "speaker_weighted_sampler_alpha": 1.0,
+    "use_language_weighted_sampler": false,
+    "language_weighted_sampler_alpha": 1.0,
+    "use_length_weighted_sampler": false,
+    "length_weighted_sampler_alpha": 1.0,
+    "model_args": {
+        "gpt_batch_size": 1,
+        "enable_redaction": false,
+        "kv_cache": true,
+        "gpt_checkpoint": null,
+        "clvp_checkpoint": null,
+        "decoder_checkpoint": null,
+        "num_chars": 255,
+        "tokenizer_file": "",
+        "gpt_max_audio_tokens": 605,
+        "gpt_max_text_tokens": 402,
+        "gpt_max_prompt_tokens": 70,
+        "gpt_layers": 30,
+        "gpt_n_model_channels": 1024,
+        "gpt_n_heads": 16,
+        "gpt_number_text_tokens": 6681,
+        "gpt_start_text_token": null,
+        "gpt_stop_text_token": null,
+        "gpt_num_audio_tokens": 1026,
+        "gpt_start_audio_token": 1024,
+        "gpt_stop_audio_token": 1025,
+        "gpt_code_stride_len": 1024,
+        "gpt_use_masking_gt_prompt_approach": true,
+        "gpt_use_perceiver_resampler": true,
+        "input_sample_rate": 22050,
+        "output_sample_rate": 24000,
+        "output_hop_length": 256,
+        "decoder_input_dim": 1024,
+        "d_vector_dim": 512,
+        "cond_d_vector_in_each_upsampling_layer": true,
+        "duration_const": 102400
+    },
+    "model_dir": null,
+    "languages": [
+        "en",
+        "es",
+        "fr",
+        "de",
+        "it",
+        "pt",
+        "pl",
+        "tr",
+        "ru",
+        "nl",
+        "cs",
+        "ar",
+        "zh-cn",
+        "hu",
+        "ko",
+        "ja",
+        "hi",
+        "ur"
+    ],
+    "temperature": 0.75,
+    "length_penalty": 1.0,
+    "repetition_penalty": 5.0,
+    "top_k": 50,
+    "top_p": 0.85,
+    "num_gpt_outputs": 1,
+    "gpt_cond_len": 30,
+    "gpt_cond_chunk_len": 4,
+    "max_ref_len": 30,
+    "sound_norm_refs": false
+}

checkpoints/XTTS_v2.0_original_model_files/dvae.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b29bc227d410d4991e0a8c09b858f77415013eeb9fba9650258e96095557d97a
+size 210514388

checkpoints/XTTS_v2.0_original_model_files/mel_stats.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f69422a8a8f344c4fca2f0c6b8d41d2151d6615b7321e48e6bb15ae949b119c
+size 1067

checkpoints/XTTS_v2.0_original_model_files/model.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c7ea20001c6a0a841c77e252d8409f6a74fb423e79b3206a0771ba5989776187
+size 1867929118

vocals_chunk_0181_1306890ms_wip8_24k.wav

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b497633a96893aa4cdefeac1766dbd729123c09893fae211f50df4765ba09996
+size 272682