{
    "_class_name": "AutoencoderKLConv3D",
    "block_out_channels": [
      128,
      256,
      512,
      1024,
      1024
    ],
    "in_channels": 3,
    "out_channels": 3,
    "latent_channels": 32,
    "layers_per_block": 2,
    "ffactor_spatial": 16,
    "ffactor_temporal": 4,
    "sample_size": 256,
    "sample_tsize": 64,
    "downsample_match_channel": true,
    "spatial_compression_ratio": 16,
    "time_compression_ratio": 4,
    "upsample_match_channel": true,
    "scaling_factor": 1.03682
  }