dit/config.json

{
  "_class_name": "NaDiT",
  "_diffusers_version": "0.35.2",
  "ada": "single",
  "block_type": [
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr",
    "mmdit_sr"
  ],
  "dtype": "bfloat16",
  "emb_dim": 18432,
  "expand_ratio": 4,
  "head_dim": 128,
  "heads": 24,
  "mlp_type": "normal",
  "norm": "fusedrms",
  "norm_eps": 1e-05,
  "num_layers": 36,
  "patch_size": [
    1,
    2,
    2
  ],
  "qk_bias": false,
  "qk_norm": "fusedrms",
  "qk_rope": true,
  "shared_mlp": false,
  "shared_qkv": false,
  "temporal_shifted": false,
  "temporal_window_size": null,
  "txt_dim": 3072,
  "txt_in_dim": 5120,
  "vid_dim": 3072,
  "vid_in_channels": 33,
  "vid_out_channels": 16,
  "window": [
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ],
    [
      4,
      3,
      3
    ]
  ],
  "window_method": [
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize",
    "720pwin_by_size_bysize",
    "720pswin_by_size_bysize"
  ]
}