vae/config.json

{
  "_class_name": "VideoAutoencoderKLWrapper",
  "_diffusers_version": "0.35.2",
  "act_fn": "silu",
  "attention": true,
  "block_out_channels": [
    128,
    256,
    512,
    512
  ],
  "down_block_types": [
    "DownEncoderBlock3D",
    "DownEncoderBlock3D",
    "DownEncoderBlock3D",
    "DownEncoderBlock3D"
  ],
  "extra_cond_dim": null,
  "force_upcast": false,
  "freeze_encoder": false,
  "gradient_checkpoint": false,
  "grouping": false,
  "in_channels": 3,
  "inflation_mode": "pad",
  "latent_channels": 16,
  "layers_per_block": 2,
  "memory_limit": null,
  "norm_num_groups": 32,
  "out_channels": 3,
  "sample_size": 32,
  "scaling_factor": 0.9152,
  "slicing": null,
  "slicing_sample_min_size": 4,
  "slicing_up_num": 0,
  "spatial_downsample_factor": 8,
  "temporal_downsample_factor": 4,
  "temporal_scale_num": 2,
  "time_receptive_field": "full",
  "up_block_types": [
    "UpDecoderBlock3D",
    "UpDecoderBlock3D",
    "UpDecoderBlock3D",
    "UpDecoderBlock3D"
  ],
  "use_post_quant_conv": false,
  "use_quant_conv": false
}