config.json

{
  "_class_name": "VCtrlModel",
  "activation_fn": "gelu-approximate",
  "attention_bias": true,
  "attention_head_dim": 64,
  "dropout": 0.0,
  "extra_conditioning_channels": 1,
  "flip_sin_to_cos": true,
  "freq_shift": 0,
  "in_channels": 16,
  "norm_elementwise_affine": true,
  "norm_eps": 1e-05,
  "num_attention_heads": 48,
  "num_layers": 7,
  "patch_size": 2,
  "time_embed_dim": 512,
  "timestep_activation_fn": "silu",
  "use_rotary_positional_embeddings": true,
  "image_to_video": true
}