transformer/config.json

{
  "action_dim": 384,
  "adaln_lora_dim": 256,
  "base_fps": 24,
  "concat_padding_mask": true,
  "condition_model_type": "DreamDojoConditionModel",
  "crossattn_emb_channels": 1024,
  "crossattn_proj_in_channels": 100352,
  "enable_fps_modulation": true,
  "eps": 1e-06,
  "hidden_dim_in_action_embedder": null,
  "in_channels": 16,
  "max_frames": 128,
  "max_img_h": 240,
  "max_img_w": 240,
  "mlp_ratio": 4.0,
  "model_channels": 2048,
  "model_type": "dreamdojo",
  "motion_consistency_dim": 1,
  "motion_consistency_weight": 0.1,
  "num_action_per_latent_frame": 4,
  "num_blocks": 28,
  "num_heads": 16,
  "out_channels": 16,
  "patch_spatial": 2,
  "patch_temporal": 1,
  "rope_h_extrapolation_ratio": 1.0,
  "rope_t_extrapolation_ratio": 1.0,
  "rope_w_extrapolation_ratio": 1.0,
  "shift": 5.0,
  "tie_word_embeddings": false,
  "timestep_scale": 1.0,
  "transformers_version": "4.57.3",
  "use_adaln_lora": true,
  "use_crossattn_projection": true,
  "zero_init_action_embedder": false
}