transformer/config.json

{
  "model_type": "dreamdojo",
  "model_channels": 5120,
  "num_heads": 40,
  "num_blocks": 36,
  "in_channels": 16,
  "out_channels": 16,
  "patch_spatial": 2,
  "patch_temporal": 1,
  "crossattn_emb_channels": 1024,
  "crossattn_proj_in_channels": 100352,
  "use_crossattn_projection": true,
  "use_adaln_lora": true,
  "adaln_lora_dim": 256,
  "action_dim": 384,
  "num_action_per_latent_frame": 4,
  "motion_consistency_dim": 1,
  "motion_consistency_weight": 0.1,
  "condition_model_type": "DreamDojoConditionModel",
  "tie_word_embeddings": false,
  "timestep_scale": 1.0
}