{ "_gradient_checkpointing": false, "action_condition_type": "no_action_condition", "add_pos_embed": true, "architectures": [ "WLA" ], "auxiliary_drop_thresh": 0.0, "chunk_size": 32, "connector_num_hidden_layers": 1, "diffusion_model_cfg": { "attention_head_dim": 32, "cross_attention_dim": 2048, "dropout": 0.0, "final_dropout": false, "input_embedding_dim": 1024, "interleave_self_attention": true, "norm_type": "norm", "num_attention_heads": 32, "num_layers": 28, "output_dim": 1024, "positional_embeddings": null }, "diffusion_model_id": "/inspire/qb-ilm/project/qproject-fundationmodel/public/zzt/models/Sana_600M", "dtype": "bfloat16", "history_obs_step": 8, "in_channels": 64, "input_size": 16, "instruction_folder_path": "configs/robotwin_seen_instruction.json", "max_action_dim": 16, "max_input_text_tokens": 256, "max_seq_len": 1024, "max_state_dim": 16, "mllm_id": "/inspire/qb-ilm/project/qproject-fundationmodel/public/zzt/models/RynnBrain-2B", "model_type": "wla", "modules_to_freeze": [ "vae" ], "modules_to_unfreeze": [ "model.mllm_backbone.model.visual", "model.mllm_backbone.model.language_model", "model.mllm_backbone.model.language_model.embed_tokens", "model.world_expert", "model.connector", "model.action_expert" ], "noise_beta_alpha": 1.5, "noise_beta_beta": 1.0, "noise_s": 0.999, "noise_scheduler_id": "/inspire/qb-ilm/project/qproject-fundationmodel/public/zzt/models/Sana_600M", "num_inference_timesteps": 32, "num_metaqueries": 64, "num_target_vision_tokens": 0, "num_timestep_buckets": 1000, "primary_loss_mode": "action", "repeated_diffusion_steps": 8, "sample_num": 1, "scheduler_id": "/inspire/qb-ilm/project/qproject-fundationmodel/public/zzt/models/Sana_600M", "secondary_loss_mode": "image", "system_prompt": "", "training_mode": "image_action", "transformers_version": "4.57.1", "use_begin_frame_context": false, "use_history_obs": false, "vae_downsample_f": 32, "vae_id": "/inspire/qb-ilm/project/qproject-fundationmodel/public/zzt/models/Sana_600M" }