Instructions to use maximellerbach/omx_multicubes_lingbot_vla_v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use maximellerbach/omx_multicubes_lingbot_vla_v2 with LeRobot:
- Notebooks
- Google Colab
- Kaggle
| { | |
| "type": "lingbot_vla_v2", | |
| "n_obs_steps": 1, | |
| "input_features": { | |
| "observation.state": { | |
| "type": "STATE", | |
| "shape": [ | |
| 6 | |
| ] | |
| }, | |
| "observation.images.wrist": { | |
| "type": "VISUAL", | |
| "shape": [ | |
| 3, | |
| 480, | |
| 640 | |
| ] | |
| }, | |
| "observation.images.top": { | |
| "type": "VISUAL", | |
| "shape": [ | |
| 3, | |
| 480, | |
| 640 | |
| ] | |
| } | |
| }, | |
| "output_features": { | |
| "action": { | |
| "type": "ACTION", | |
| "shape": [ | |
| 6 | |
| ] | |
| } | |
| }, | |
| "device": "cuda", | |
| "use_amp": false, | |
| "use_peft": false, | |
| "push_to_hub": true, | |
| "repo_id": "maximellerbach/omx_multicubes_lingbot_vla_v2", | |
| "private": null, | |
| "tags": null, | |
| "license": null, | |
| "pretrained_path": "robbyant/lingbot-vla-v2-6b", | |
| "pretrained_revision": null, | |
| "chunk_size": 50, | |
| "n_action_steps": 50, | |
| "max_action_dim": 55, | |
| "max_state_dim": 55, | |
| "normalization_mapping": { | |
| "VISUAL": "IDENTITY", | |
| "STATE": "MEAN_STD", | |
| "ACTION": "MEAN_STD" | |
| }, | |
| "tokenizer_path": "Qwen/Qwen3-VL-4B-Instruct", | |
| "vlm_family": "qwen3_vl", | |
| "tokenizer_max_length": 72, | |
| "resize_imgs_with_padding": [ | |
| 224, | |
| 224 | |
| ], | |
| "image_max_pixels": 262144, | |
| "image_min_pixels": 131072, | |
| "num_steps": 10, | |
| "robot_config_path": "/fsx/maxime_ellerbach/robots/runs/lingbot_vla_v2/omx_multicubes_robot_config.yaml", | |
| "norm_stats_path": "/fsx/maxime_ellerbach/robots/runs/lingbot_vla_v2/omx_multicubes_norm_stats.json", | |
| "processor_path": "Qwen/Qwen3-VL-4B-Instruct", | |
| "dtype": "bfloat16", | |
| "canonical_joints": { | |
| "arm.position": 14, | |
| "end.position": 14, | |
| "effector.position": 2, | |
| "hand.position": 12, | |
| "waist.position": 4, | |
| "head.position": 2, | |
| "base.velocity": 3, | |
| "reserved.slots": 4 | |
| }, | |
| "canonical_norm_type": { | |
| "arm.position": "meanstd", | |
| "end.position": "meanstd", | |
| "effector.position": "meanstd", | |
| "hand.position": "meanstd", | |
| "waist.position": "meanstd", | |
| "head.position": "meanstd", | |
| "base.velocity": "meanstd", | |
| "reserved.slots": "meanstd" | |
| }, | |
| "canonical_cameras": [ | |
| "camera_top", | |
| "camera_wrist_left", | |
| "camera_wrist_right" | |
| ], | |
| "use_qwen3_chat_template": true, | |
| "return_image_grid_thw": true, | |
| "qwen3vl_use_vision_boundaries": true, | |
| "precompute_grid_thw": false, | |
| "use_qwen3_fixed_grid_cache": true, | |
| "expert_hidden_size": 768, | |
| "expert_intermediate_size": 2752, | |
| "action_num_attention_heads": 32, | |
| "action_num_key_value_heads": 8, | |
| "action_head_dim": 128, | |
| "action_fp32": false, | |
| "use_moe": true, | |
| "token_moe_layers": [ | |
| 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 | |
| ], | |
| "token_num_experts": 32, | |
| "token_top_k": 4, | |
| "token_moe_intermediate_size": 512, | |
| "token_shared_intermediate_size": 704, | |
| "bias_update_speed": 0.0, | |
| "bias_centering": false, | |
| "bias_update_interval": 1, | |
| "sequence_wise_loss_coeff": 0.001, | |
| "sequence_wise_mode": "per_sequence", | |
| "router_z_loss_coeff": 0.0001, | |
| "router_activation": "sigmoid", | |
| "routed_scaling_factor": 4.0, | |
| "use_shared_expert_gate": false, | |
| "moe_implementation": "fused", | |
| "use_depth": false, | |
| "num_task_tokens": 8, | |
| "align_params": {}, | |
| "enable_expert_vision": false, | |
| "expert_vision_type": null, | |
| "attention_implementation": "eager", | |
| "vit_attn_implementation": "eager", | |
| "use_cache": true, | |
| "post_training": true, | |
| "freeze_vision_encoder": true, | |
| "train_expert_only": false, | |
| "train_state_proj": true, | |
| "vlm_causal": false, | |
| "vocab_size": 0, | |
| "use_lm_head": false, | |
| "loss_type": "fm", | |
| "adanorm_time": true, | |
| "split_gate_liner": false, | |
| "nosplit_gate_liner": false, | |
| "separate_time_proj": false, | |
| "final_norm_adanorm": false, | |
| "norm_qkv": false, | |
| "optimizer_lr": 1e-05, | |
| "optimizer_betas": [ | |
| 0.9, | |
| 0.95 | |
| ], | |
| "optimizer_eps": 1e-08, | |
| "optimizer_weight_decay": 0.0, | |
| "optimizer_grad_clip_norm": 1.0, | |
| "scheduler_warmup_steps": 1000, | |
| "scheduler_decay_steps": 30000, | |
| "scheduler_decay_lr": 1e-05 | |
| } |