pretrain_video_S_768d12l.json

{
  "decoder_config": {
    "image_dim": 4,
    "patch_size": 2,
    "decoder_style": "dit",
    "num_layers": 12,
    "hidden_size": 768,
    "intermediate_size": 3072,
    "num_heads": 12,
    "decode_mode": "diffusion"
  },
  "encoder_hidden_size": 768,
  "latent_h": 32,
  "latent_w": 32,
  "latent_mean": [
    -0.69,
    -0.48,
    -0.6,
    0.28
  ],
  "latent_std": [
    12.38,
    11.22,
    7.93,
    21.22
  ],
  "pretrain_mode": "video",
  "source_ckpt": "ttvidt-dit-pretrain/2u6ev4md/checkpoints/epoch=14-step=100000.ckpt"
}