vae/config.json

{
  "_class_name": "AutoencoderKLJoyVideoEdit",
  "_diffusers_version": "0.40.0.dev0",
  "block_in_channels": [
    128,
    256,
    512,
    1024
  ],
  "chunk_size": 48,
  "in_channels": 3,
  "latent_channels": 64,
  "latents_mean": [
    0.003708,
    0.018799,
    -0.049072,
    -1.171875,
    0.064453,
    0.648438,
    -0.507812,
    0.030273,
    -0.090332,
    0.10498,
    -0.18457,
    0.667969,
    -0.863281,
    -0.12793,
    0.000511,
    0.472656,
    -0.636719,
    0.761719,
    0.170898,
    -0.482422,
    0.267578,
    0.092285,
    -0.066406,
    -0.002029,
    0.201172,
    0.026489,
    -0.073242,
    0.016479,
    -0.449219,
    0.070312,
    -0.423828,
    0.804688,
    -1.773438,
    -0.117676,
    0.010986,
    -0.092285,
    -0.003448,
    -0.133789,
    -0.230469,
    -0.410156,
    -0.292969,
    0.414062,
    -0.150391,
    -0.045654,
    -0.213867,
    -0.126953,
    -0.062012,
    -1.039062,
    0.058838,
    -0.015442,
    -0.054932,
    0.100098,
    -0.112793,
    0.0177,
    0.213867,
    -0.003906,
    0.172852,
    0.003281,
    -0.257812,
    0.010071,
    0.008362,
    -0.163086,
    0.126953,
    -1.34375
  ],
  "latents_std": [
    0.5625,
    1.710938,
    0.695312,
    2.453125,
    0.769531,
    3.265625,
    3.140625,
    0.835938,
    0.570312,
    0.757812,
    0.925781,
    2.046875,
    2.171875,
    0.503906,
    1.53125,
    1.03125,
    1.90625,
    2.375,
    0.5625,
    0.964844,
    0.699219,
    0.648438,
    3.890625,
    0.707031,
    2.265625,
    0.878906,
    0.550781,
    0.451172,
    2.46875,
    0.53125,
    1.914062,
    3.234375,
    4.65625,
    1.1875,
    0.65625,
    0.738281,
    0.851562,
    0.71875,
    0.796875,
    2.78125,
    1.445312,
    0.589844,
    0.535156,
    0.628906,
    0.734375,
    0.597656,
    0.921875,
    3.09375,
    0.585938,
    0.527344,
    0.570312,
    1.84375,
    0.574219,
    0.617188,
    0.65625,
    0.75,
    0.601562,
    0.539062,
    1.664062,
    0.777344,
    0.507812,
    0.652344,
    0.699219,
    2.8125
  ],
  "layers_per_block": 2,
  "out_channels": 3,
  "patch_size": 2,
  "temporal_downsample": [
    true,
    true,
    true,
    false
  ]
}