{
  "_class_name": "AutoencoderKLTurboVAED",
  "_diffusers_version": "0.32.0.dev0",
  "decoder_block_out_channels": [
    64,
    128,
    256,
    512
  ],
  "decoder_causal": false,
  "in_channels": 3,
  "latent_channels": 48,
  "decoder_layers_per_block": [
    2,
    2,
    2,
    3,
    3
  ],
  "out_channels": 3,
  "patch_size": 2,
  "patch_size_t": 1,
  "resnet_norm_eps": 1e-06,
  "scaling_factor": 1.0,
  "decoder_spatio_temporal_scaling": [
    false,
    true,
    true,
    true
  ],
  "decoder_spatio_only": [
    false,
    true,
    false,
    false
  ],
  "decoder_is_dw_conv": [
    false,
    false,
    false,
    false,
    false
  ],
  "decoder_dw_kernel_size": 5,
  "aligned_feature_projection_mode": "conv-2layer",
  "aligned_feature_projection_dim": [
    [512, 1024],
    [512, 1024]
  ],
  "aligned_blks_indices": [0, 1],
  "spatial_compression_ratio": 16,
  "temporal_compression_ratio": 4,
  "first_chunk_size": 7,
  "step_size": 7,
  "use_unpatchify": true
}