config.json

{
  "model": "xtts",
  "run_name": "xttsv2_finetune_es",
  "output_path": "/tmp/output_model",
  "datasets": [
  {
    "formatter": "json",
    "meta_file_train": "/tmp/voxpopuli_es_500/metadata.json",
    "path": "/tmp/voxpopuli_es_500/wav_data",  
    "language": "es"
  }
  "batch_size": 4,
  "eval_batch_size": 4,
  "num_loader_workers": 2,
  "num_eval_loader_workers": 1,
  "run_eval": true,
  "test_delay_epochs": 1,
  "epochs": 5,
  "save_step": 100,
  "print_step": 25,
  "eval_step": 100,
  "mixed_precision": true,
  "cudnn_benchmark": true,
  "lr": 1e-4,
  "grad_clip": 5.0,
  "loss_masking": true,
  "use_speaker_embedding": true,
  "compute_input_seq_cache": true,
  "eval_split_size": 0.05,
  "text_cleaner": "multilingual_cleaners",
  "use_phonemes": false,
  "phoneme_language": "es",
  "min_text_len": 1,
  "max_text_len": 200,
  "min_audio_len": 1000,
  "max_audio_len": 160000,
  "output_step": 25,
  "logger": "tensorboard",
  "seed": 1234
}