{
  "_name_or_path": "subatomicseer/s2a-ss",
  "acoustic_codebook_size": 1024,
  "acoustic_downsample_factor": 320,
  "acoustic_n_codebooks": 12,
  "acoustic_sample_rate": 16000,
  "architectures": [
    "SoundStorm"
  ],
  "bos_token_id": 1,
  "encoder_config": {
    "attn_dropout": 0.0,
    "attn_flash": true,
    "conv_dropout": 0.0,
    "conv_expansion_factor": 2,
    "conv_kernel_size": 5,
    "depth": 16,
    "dim_head": 64,
    "ff_dropout": 0.0,
    "ff_mult": 4,
    "heads": 16
  },
  "eos_token_id": 2,
  "hidden_size": 1024,
  "label_smoothing": 0.0,
  "model_type": "soundstorm",
  "num_quantizers": 12,
  "num_semantic_token_ids": 1024,
  "pad_token_id": 0,
  "torch_dtype": "float32",
  "transformers_version": "4.41.1"
}