| { |
| "dtype": "bfloat16", |
| "architectures": [ |
| "RLDX" |
| ], |
| "transformers_version": "4.57.0", |
| "attn_implementation": null, |
| "model_type": "RLDX-1", |
| "model_dtype": "bfloat16", |
| "model_name": "RLWRLD/RLDX-1-VLM", |
| "backbone_model_type": "vtc_qwen3_vl", |
| "model_revision": null, |
| "tune_top_llm_layers": 4, |
| "backbone_embedding_dim": 4096, |
| "tune_llm": false, |
| "tune_visual": false, |
| "select_layer": 18, |
| "reproject_vision": false, |
| "use_flash_attention": true, |
| "load_bf16": true, |
| "freeze_cog_tokens": false, |
| "image": { |
| "max_area": 65536, |
| "resize_m": 32, |
| "crop_fraction": null |
| }, |
| "augmentation": { |
| "rotation_angle": null, |
| "color_jitter": { |
| "brightness": 0.3, |
| "contrast": 0.4, |
| "hue": 0.08, |
| "saturation": 0.5 |
| } |
| }, |
| "formalize_language": true, |
| "apply_sincos_state_encoding": false, |
| "use_relative_action": true, |
| "use_percentiles": true, |
| "conversation_image_first": false, |
| "max_state_dim": 64, |
| "max_action_dim": 64, |
| "action_horizon": 16, |
| "hidden_size": 1024, |
| "input_embedding_dim": 1536, |
| "general_embodiment_train_ratio": 0.0, |
| "mask_padded_action_dims": true, |
| "add_pos_embed": true, |
| "attn_dropout": 0.0, |
| "use_vlln": false, |
| "max_seq_len": 1024, |
| "n_cog_tokens": 64, |
| "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, |
| "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_swiglu": true, |
| "vl_dim": 4096 |
| }, |
| "num_inference_timesteps": 4, |
| "noise_beta_alpha": 1.5, |
| "noise_beta_beta": 1.0, |
| "noise_s": 0.999, |
| "num_timestep_buckets": 1000, |
| "tune_projector": true, |
| "tune_diffusion_model": true, |
| "tune_vlln": true, |
| "action_model_lora": null, |
| "backbone_lora": null, |
| "state_dropout_prob": 0.0, |
| "state_additive_noise_scale": 0.0, |
| "rtc_training_max_delay": 0, |
| "memory": null, |
| "motion": { |
| "insert_layer": 9, |
| "d_hid": 512, |
| "window": [ |
| 5, |
| 9, |
| 9 |
| ], |
| "ext_chnls": [ |
| 256 |
| ], |
| "int_chnls": [ |
| 256, |
| 256, |
| 512 |
| ], |
| "corr_func": "cosine", |
| "n_encoders": 1, |
| "use_layerscale": false, |
| "layerscale_init": 1e-05, |
| "use_layernorm": true, |
| "use_syncbn": false, |
| "injection_point": "vision_encoder", |
| "pool_type": "avg", |
| "drop": true, |
| "gradient_check": false |
| }, |
| "use_video": true, |
| "video_length": 4, |
| "video_stride": 2, |
| "physics": null, |
| "max_num_embodiments": 8, |
| "format_version": 2 |
| } |
|
|
