sound_tokenizer/config.json

{
  "_class_name": "Cosmos3AVAEAudioTokenizer",
  "_diffusers_version": "0.39.0",
  "activation": "snakebeta",
  "anti_aliasing": false,
  "bottleneck": {
    "type": "vae"
  },
  "bottleneck_type": "vae",
  "causal": false,
  "dec_anti_aliasing": false,
  "dec_c_mults": [
    1,
    2,
    4,
    8,
    16
  ],
  "dec_dim": 320,
  "dec_final_tanh": false,
  "dec_out_channels": 2,
  "dec_strides": [
    2,
    4,
    5,
    6,
    8
  ],
  "dec_type": "oobleck",
  "dec_use_nearest_upsample": false,
  "dec_use_snake": true,
  "dec_use_tanh_at_final": false,
  "enc_c_mults": [
    1,
    2,
    4
  ],
  "enc_dim": 192,
  "enc_hop_length": 16,
  "enc_identity_init": false,
  "enc_intermediate_dim": 768,
  "enc_latent_dim": 128,
  "enc_n_fft": 64,
  "enc_num_blocks": 2,
  "enc_num_layers": 12,
  "enc_strides": [
    4,
    5,
    6
  ],
  "enc_type": "spec_convnext",
  "enc_use_snake": true,
  "encoder_enabled": true,
  "hop_size": 1920,
  "input_channels": 1,
  "latent_mean": null,
  "latent_std": null,
  "model_type": "autoencoder_v2",
  "normalize_volume": true,
  "padding_mode": "zeros",
  "sampling_rate": 48000,
  "snake_logscale": true,
  "stereo": true,
  "use_cuda_kernel": false,
  "use_wav_as_input": true,
  "vocoder_input_dim": 64
}