egs/tts/VALLE_V2/exp_ar_libritts.json

{
    "model_type": "VALLE_V2_AR",
    "log_dir": "./ckpt/VALLE_V2",
    "use_speechtokenizer": true,
    "train": {
        "gradient_accumulation_step": 1,
        "find_unused_parameters": false,
        "tracker": ["tensorboard"],
        "max_epoch": 1000,
        "save_checkpoint_stride": [500],
        "keep_last": [1],
        "run_eval": [true],
        "dataloader": {
          "num_worker": 4,
          "pin_memory": true,
          "persistent_workers": true
        },
        "dataset": {
          "use_dynamic_batchsize": false,
          "name": "libritts"
        },
        "optimizer": "adamW",
        "adamw": {
          "lr": 1e-4
        },
        "scheduler": {
          "warmup_steps": 25000,
          "total_steps": 800000,
          "min_lr": 1e-5
        },
        "exponentiallr": {
          "gamma": 0.999999
        },
        "batch_size": 5,
        "max_tokens": 5000,
        "max_sentences": 64,
        "random_seed": 0
    }, 
    "dataset": {
      "dataset_list":["train-clean-360"],
      "data_dir": "/path/to/your/libritts"  // You can also change to other splits like "dev-clean"
    },
    "model": {
      "phone_vocab_size": 300,
      "target_vocab_size": 1024,
      "pad_token_id": 1324,
      "bos_target_id": 1325,
      "eos_target_id": 1326,
      "bos_phone_id": 1327,
      "eos_phone_id": 1328,
      "bos_prompt_id": 1329,
      "eos_prompt_id": 1330,
      "num_hidden_layers": 16
    }
    }