Ref2VA/audio_vae/metadata.json

{
  "metadata": {
    "kwargs": {
      "attn_proj": true,
      "decoder_dim": 1024,
      "decoder_rates": [
        5,
        5,
        2,
        2,
        2,
        2,
        2
      ],
      "decoder_type": "bigvgan",
      "encoder_dim": 64,
      "encoder_rates": [
        2,
        4,
        4,
        5,
        5
      ],
      "latent_dim": 2048,
      "sample_rate": 32000,
      "vae_latent_channels": 32
    }
  }
}