config.json

{
  "boundary": 0.875,
  "cross_attn_norm": true,
  "dim": 5120,
  "dual_model": true,
  "eps": 1e-6,
  "ffn_dim": 13824,
  "frame_num": 81,
  "freq_dim": 256,
  "in_dim": 16,
  "max_area": 0,
  "model_type": "t2v",
  "model_version": "2.2",
  "num_heads": 40,
  "num_layers": 40,
  "num_train_timesteps": 1000,
  "out_dim": 16,
  "patch_size": [
    1,
    2,
    2
  ],
  "qk_norm": true,
  "sample_fps": 16,
  "sample_guide_scale": [
    3.0,
    4.0
  ],
  "sample_neg_prompt": "色调艳丽，过曝，静态，细节模糊不清，字幕，风格，作品，画作，画面，静止，整体发灰，最差质量，低质量，JPEG压缩残留，丑陋的，残缺的，多余的手指，画得不好的手部，画得不好的脸部，畸形的，毁容的，形态畸形的肢体，手指融合，静止不动的画面，杂乱的背景，三条腿，背景人很多，倒着走",
  "sample_shift": 12.0,
  "sample_steps": 40,
  "t5_dim": 4096,
  "t5_dim_attn": 4096,
  "t5_dim_ffn": 10240,
  "t5_num_buckets": 32,
  "t5_num_heads": 64,
  "t5_num_layers": 24,
  "t5_vocab_size": 256384,
  "text_dim": 4096,
  "text_len": 512,
  "vae_stride": [
    4,
    8,
    8
  ],
  "vae_z_dim": 16,
  "window_size": [
    -1,
    -1
  ]
}