| { |
| "action_horizon": 16, |
| "action_model_lora_alpha": 32, |
| "action_model_lora_dropout": 0.0, |
| "action_model_lora_rank": 16, |
| "action_model_lora_target_modules": [ |
| "vl_qkv", |
| "vl_proj", |
| "sa_qkv", |
| "sa_proj", |
| "p_qkv", |
| "p_proj", |
| "linear1", |
| "linear2" |
| ], |
| "action_model_use_lora": false, |
| "actlat_embodiment_id": "openarm_prq", |
| "actlat_mode": false, |
| "actlat_prq_stats_path": "", |
| "actlat_standardize": false, |
| "actlat_tokenizer_path": "", |
| "actlat_z_stats_path": "", |
| "anchor_dim_norm": "", |
| "anchor_dim_stats_path": "", |
| "anchor_human_tips_weight": 1.0, |
| "anchor_readout_asym": false, |
| "anchor_readout_dim": 24, |
| "anchor_readout_shuffle": false, |
| "anchor_readout_taps": "cog,d2,s3,s7", |
| "anchor_readout_tweight": true, |
| "anchor_readout_weight": 0.0, |
| "anchor_tips_dims": 15, |
| "architectures": [ |
| "RLDX" |
| ], |
| "backbone_lora_alpha": 32, |
| "backbone_lora_dropout": 0.0, |
| "backbone_lora_num_layers": -1, |
| "backbone_lora_rank": 16, |
| "backbone_lora_target_modules": [ |
| "q_proj", |
| "k_proj", |
| "v_proj", |
| "o_proj", |
| "gate_proj", |
| "up_proj", |
| "down_proj" |
| ], |
| "backbone_use_lora": false, |
| "color_jitter_params": null, |
| "conversation_image_first": false, |
| "diffusion_model_cfg": { |
| "action_model_max_seq_len": 512, |
| "attention_head_dim": 64, |
| "depth_multi_stream": 4, |
| "depth_single_stream": 8, |
| "dropout": 0.2, |
| "final_dropout": true, |
| "num_attention_heads": 24, |
| "output_dim": 1024, |
| "physics_dim": 0, |
| "positional_embeddings": "rope_sa_only", |
| "pre_norm": "layer_norm", |
| "qk_norm": "rms_norm", |
| "rope_theta": 10000.0, |
| "sa_dim": 1536, |
| "temb_type": "input_token", |
| "use_physics": false, |
| "use_swiglu": true, |
| "vl_dim": 4096 |
| }, |
| "dtype": "bfloat16", |
| "egopi_split_embodiment": false, |
| "embod_reg_gather": false, |
| "embod_reg_hand_bins": 0, |
| "embod_reg_hand_edges": "", |
| "embod_reg_layer": 2, |
| "embod_reg_mode": "", |
| "embod_reg_tbins": 0, |
| "embod_reg_vic_cov": 0.04, |
| "embod_reg_vic_var": 1.0, |
| "embod_reg_weight": 0.05, |
| "freeze_cog_tokens": false, |
| "freeze_modules": "", |
| "general_embodiment_train_ratio": 0, |
| "gr1hp_align_dim": 24, |
| "gr1hp_align_only": false, |
| "gr1hp_align_shuffle": false, |
| "gr1hp_align_weight": 0.0, |
| "gr1hp_split_embodiment": false, |
| "image_max_area": 65536, |
| "image_resize_m": 32, |
| "keep_full_lm": false, |
| "memory_cfg": { |
| "hidden_size": 4096, |
| "intermediate_size": 16384, |
| "max_position_embeddings": 32, |
| "num_attention_heads": 16, |
| "num_hidden_layers": 2, |
| "num_key_value_heads": 16, |
| "rms_norm_eps": 1e-05, |
| "use_causal_attn": true, |
| "use_rope": true |
| }, |
| "model_name": "RLWRLD/RLDX-1-VLM", |
| "model_revision": null, |
| "model_type": "RLDX-1", |
| "n_cog_tokens": 64, |
| "physics_delta_indices": null, |
| "physics_dims": [], |
| "physics_keys": [], |
| "physics_use_flow_matching": true, |
| "random_crop_fraction": null, |
| "random_rotation_angle": null, |
| "rtc_inference_delay": 0, |
| "rtc_inference_exec_horizon": 0, |
| "rtc_inference_mode": "none", |
| "rtc_jacobian_steps_only": 3, |
| "rtc_training_max_delay": 0, |
| "stage_from_checkpoint": "", |
| "state_dropout_prob": 0.0, |
| "train_stage": "", |
| "transformers_version": "4.57.0", |
| "tune_diffusion_model": true, |
| "tune_llm": false, |
| "tune_projector": true, |
| "tune_top_llm_layers": 0, |
| "tune_visual": false, |
| "use_memory": false, |
| "use_motion": false, |
| "use_physics": false, |
| "use_video": true, |
| "video_length": 1, |
| "video_stride": 2, |
| "vlm_lang_loss": false, |
| "vlm_lang_weight": 0.5 |
| } |
|
|
