Robotics
Transformers
Safetensors
RLDX-1
vla
vision-language-action
manipulation
flow-matching
rldx
allex
memory
physics
Instructions to use RLWRLD/RLDX-1-MT-ALLEX with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use RLWRLD/RLDX-1-MT-ALLEX with Transformers:
# Load model directly from transformers import RLDX model = RLDX.from_pretrained("RLWRLD/RLDX-1-MT-ALLEX", dtype="auto") - Notebooks
- Google Colab
- Kaggle
| { | |
| "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, | |
| "action_horizon": 40, | |
| "allow_missing_physics": true, | |
| "architectures": [ | |
| "RLDX" | |
| ], | |
| "backbone_model_type": "vtc_qwen3_vl", | |
| "backbone_trainable_params_fp32": true, | |
| "color_jitter_params": { | |
| "brightness": 0.3, | |
| "contrast": 0.4, | |
| "hue": 0.08, | |
| "saturation": 0.5 | |
| }, | |
| "concat_memory": true, | |
| "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": 48, | |
| "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": true, | |
| "use_swiglu": true, | |
| "vl_dim": 4096 | |
| }, | |
| "dtype": "bfloat16", | |
| "freeze_cog_tokens": false, | |
| "load_bf16": true, | |
| "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 | |
| }, | |
| "memory_dropout_prob": 0.3, | |
| "memory_length": 4, | |
| "memory_n_cog_tokens": 16, | |
| "memory_video_delta_indices": [ | |
| -48, | |
| -32, | |
| -16, | |
| 0 | |
| ], | |
| "model_name": "RLWRLD/RLDX-1-VLM", | |
| "model_type": "RLDX-1", | |
| "motion_drop": true, | |
| "use_motion": true, | |
| "motion_gradient_check": false, | |
| "motion_injection_point": "vision_encoder", | |
| "motion_insert_layer": 9, | |
| "motion_pool_type": "avg", | |
| "n_cog_tokens": 64, | |
| "general_embodiment_train_ratio": 0, | |
| "physics_delta_indices": [ | |
| 0, | |
| 1, | |
| 2, | |
| 3, | |
| 4, | |
| 5, | |
| 6, | |
| 7, | |
| 8, | |
| 9, | |
| 10, | |
| 11, | |
| 12, | |
| 13, | |
| 14, | |
| 15, | |
| 16, | |
| 17, | |
| 18, | |
| 19, | |
| 20, | |
| 21, | |
| 22, | |
| 23, | |
| 24, | |
| 25, | |
| 26, | |
| 27, | |
| 28, | |
| 29, | |
| 30, | |
| 31, | |
| 32, | |
| 33, | |
| 34, | |
| 35, | |
| 36, | |
| 37, | |
| 38, | |
| 39, | |
| 40 | |
| ], | |
| "physics_dims": [ | |
| 48 | |
| ], | |
| "physics_dropout_prob": 0.3, | |
| "physics_keys": [ | |
| "torque" | |
| ], | |
| "physics_loss_weight": 0.1, | |
| "physics_use_flow_matching": true, | |
| "qwen3_collator": true, | |
| "random_rotation_angle": null, | |
| "reproject_vision": false, | |
| "state_dropout_prob": 0.3, | |
| "transformers_version": "4.57.0", | |
| "tune_diffusion_model": true, | |
| "tune_llm": false, | |
| "tune_projector": true, | |
| "tune_top_llm_layers": 4, | |
| "tune_visual": false, | |
| "use_memory": true, | |
| "use_physics": true, | |
| "use_relative_action": true, | |
| "use_video": true, | |
| "video_length": 4 | |
| } |