{ "use_identity_mlp": true, "use_vlm_directvlm_condition": true, "use_learnable_token_condition": true, "diffusion_c_input_dim": 2560, "text_encoder_norm": false, "connector_norm": false, "img_gen_scales": [ 16 ], "selected_hidden_states_layers": [ 5, 12, 20 ], "diffusion_inner_dim": 3840 }