transformer/config.json

{
  "_class_name": "MotifVideoTransformer3DModel",
  "_diffusers_version": "0.36.0",
  "_library": "diffusers",
  "attention_head_dim": 128,
  "image_embed_dim": 1152,
  "in_channels": 33,
  "mlp_ratio": 4.0,
  "norm_type": "layer_norm",
  "num_attention_heads": 12,
  "num_decoder_layers": 8,
  "num_layers": 12,
  "num_single_layers": 24,
  "out_channels": 16,
  "patch_size": 2,
  "patch_size_t": 1,
  "qk_norm": "rms_norm",
  "rope_axes_dim": [
    16,
    56,
    56
  ],
  "rope_theta": 10000.0,
  "text_embed_dim": 2560,
  "enable_text_cross_attention_dual": false,
  "enable_text_cross_attention_single": true
}