models/video_vae_v3/s8_c16_t4_inflation_sd3.yaml

__object__:
  path: models.video_vae_v3.modules.attn_video_vae
  name: VideoAutoencoderKLWrapper
  args: as_params

act_fn: silu
block_out_channels:
  - 128
  - 256
  - 512
  - 512
down_block_types:
  - DownEncoderBlock3D
  - DownEncoderBlock3D
  - DownEncoderBlock3D
  - DownEncoderBlock3D
in_channels: 3
latent_channels: 16
layers_per_block: 2
norm_num_groups: 32
out_channels: 3
slicing_sample_min_size: 4
temporal_scale_num: 2
inflation_mode: pad
up_block_types:
  - UpDecoderBlock3D
  - UpDecoderBlock3D
  - UpDecoderBlock3D
  - UpDecoderBlock3D
spatial_downsample_factor: 8
temporal_downsample_factor: 4
use_quant_conv: False
use_post_quant_conv: False