embed_dim: 8
lossconfig:
  type: LPIPS
  params:
    kl_weight: 1.0e-06
    perceptual_weight: 1.0
ddconfig:
  double_z: true
  z_channels: 8
  t_frames: 17
  in_channels: 3
  out_ch: 3
  ch: 128
  ch_mult:
  - 1
  - 2
  - 4
  - 4
  spatial_stride:
  - 2
  - 2
  - 2
  - 1
  temporal_stride:
  - 2
  - 2
  - 1
  - 1
  num_res_blocks: 4
  attn_resolutions: []
  dropout: 0.0
  attn_type: causual_spatial_temporal
  revise_norm_bug: True

fps_ds: 8
resolution_video: 256
resolution_image: 256
model_name: CausualVAEVideo
precision: bf16
keep_aspect_ratio: 'False'


#scaling_factor_size_128_video: 0.21585
#scaling_factor_size_256_video: 0.22970
#scaling_factor_size_512_video: 0.21397
scaling_factor_video: 0.22970

#scaling_factor_size_128_image: 0.21585
#scaling_factor_size_256_image: 0.26691
#scaling_factor_size_512_image: 0.270743
scaling_factor_image: 0.26691