{
  "transformer": {
    "in_channels": 128,
    "out_channels": 128,
    "audio_in_channels": 8,
    "audio_out_channels": 8,
    "num_layers": 4,
    "num_attention_heads": 2,
    "attention_head_dim": 32,
    "audio_num_attention_heads": 2,
    "audio_attention_head_dim": 32,
    "cross_attention_dim": 64,
    "audio_cross_attention_dim": 64,
    "caption_channels": 8,
    "use_prompt_embeddings": false,
    "cross_attn_mod": true,
    "audio_cross_attn_mod": true,
    "use_prompt_adaln_single": false,
    "use_keyframes_abs_pos_embedding": true
  },
  "connectors": {
    "caption_channels": 8,
    "text_proj_in_factor": 3,
    "video_connector_num_attention_heads": 2,
    "video_connector_attention_head_dim": 32,
    "video_connector_num_layers": 2,
    "video_connector_num_learnable_registers": 4,
    "audio_connector_num_attention_heads": 2,
    "audio_connector_attention_head_dim": 32,
    "audio_connector_num_layers": 2,
    "audio_connector_num_learnable_registers": 4,
    "video_hidden_dim": 64,
    "audio_hidden_dim": 64,
    "per_modality_projections": true
  },
  "vae": {
    "in_channels": 3,
    "out_channels": 3,
    "latent_channels": 128,
    "block_out_channels": [
      16,
      32
    ],
    "decoder_block_out_channels": [
      16
    ],
    "layers_per_block": [
      1,
      1,
      1
    ],
    "decoder_layers_per_block": [
      1,
      1
    ],
    "spatio_temporal_scaling": [
      true,
      true
    ],
    "decoder_spatio_temporal_scaling": [
      true
    ],
    "downsample_type": [
      "spatiotemporal",
      "spatiotemporal"
    ],
    "upsample_type": [
      "spatiotemporal"
    ],
    "upsample_residual": [
      true
    ],
    "upsample_factor": [
      2
    ],
    "patch_size": 4,
    "patch_size_t": 1
  },
  "audio_vae": {
    "base_channels": 8,
    "output_channels": 2,
    "ch_mult": [
      1,
      2
    ],
    "num_res_blocks": 1,
    "in_channels": 2,
    "resolution": 64,
    "latent_channels": 8,
    "mel_bins": 4
  },
  "vocoder": {
    "in_channels": 8,
    "hidden_channels": 32,
    "out_channels": 2,
    "upsample_kernel_sizes": [
      8,
      8
    ],
    "upsample_factors": [
      4,
      4
    ],
    "resnet_kernel_sizes": [
      3
    ],
    "resnet_dilations": [
      [
        1,
        1,
        1
      ]
    ]
  },
  "text_encoder": {
    "vocab_size": 262144,
    "hidden_size": 8,
    "intermediate_size": 16,
    "num_hidden_layers": 2,
    "num_attention_heads": 1,
    "num_key_value_heads": 1,
    "head_dim": 32,
    "max_position_embeddings": 4096,
    "tie_word_embeddings": true
  }
}