config.json

{
  "architectures": [
    "LightningSpeechV2Model"
  ],
  "bos_token_id": 1,
  "encoder_config": {
    "attn_dropout": 0.0,
    "attn_flash": true,
    "conv_dropout": 0.0,
    "conv_kernel_size": 5,
    "cross_attention": false,
    "depth": 16,
    "dim_head": 64,
    "ff_dropout": 0.0,
    "ff_mult": 4,
    "heads": 16
  },
  "eos_token_id": 2,
  "hidden_size": 1024,
  "injection_layers": [
    7,
    9,
    11,
    13
  ],
  "model_type": "lightning_speech_v2",
  "num_semantic_tokens": 1024,
  "pad_token_id": 0,
  "quantizer_drop_prob": 0.75,
  "torch_dtype": "bfloat16",
  "transformers_version": "4.42.3"
}