Instructions to use maximellerbach/omx_multicubes_lawam with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- LeRobot
How to use maximellerbach/omx_multicubes_lawam with LeRobot:
- Notebooks
- Google Colab
- Kaggle
| { | |
| "type": "lawam", | |
| "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_lawam", | |
| "private": null, | |
| "tags": null, | |
| "license": null, | |
| "pretrained_path": null, | |
| "pretrained_revision": null, | |
| "chunk_size": 30, | |
| "n_action_steps": 30, | |
| "num_video_frames": 2, | |
| "normalization_mapping": { | |
| "VISUAL": "IDENTITY", | |
| "STATE": "MIN_MAX", | |
| "ACTION": "MIN_MAX" | |
| }, | |
| "base_vlm": "Qwen/Qwen3-VL-2B-Instruct", | |
| "base_vlm_path": null, | |
| "lam_ckpt_path": null, | |
| "hf_cache_dir": null, | |
| "lam_dim": 1024, | |
| "lam_num_heads": 16, | |
| "lam_ffn_expansion_factor": 4, | |
| "lam_enc_layers": 24, | |
| "lam_codebook_size": 32, | |
| "lam_code_dim": 32, | |
| "lam_max_state_dim": 14, | |
| "lam_num_queries": 1, | |
| "lam_dec_layers": 12, | |
| "lam_dropout": 0.0, | |
| "lam_vq_type": "vae", | |
| "lam_vq_layer_norm": true, | |
| "lam_norm_latents": true, | |
| "lam_norm_latents_type": "ln", | |
| "lam_enc_add_state": false, | |
| "lam_enc_modal_mask": true, | |
| "lam_latent_layer_to_use": -2, | |
| "lam_multi_input": false, | |
| "lam_num_embodiments": 32, | |
| "lam_image_hw": [ | |
| 256, | |
| 256 | |
| ], | |
| "lam_patch_size": 16, | |
| "lam_decoder_last_ln": true, | |
| "dinov3_hidden_size": 768, | |
| "dinov3_intermediate_size": 3072, | |
| "dinov3_num_hidden_layers": 12, | |
| "dinov3_num_attention_heads": 12, | |
| "dinov3_num_register_tokens": 4, | |
| "lawam_checkpoint_path": null, | |
| "lawam_dataset_stats_path": null, | |
| "lawam_unnorm_key": null, | |
| "primary_image_features": [ | |
| "observation.images.top" | |
| ], | |
| "wrist_image_features": [ | |
| "observation.images.wrist" | |
| ], | |
| "default_task": "Execute the robot action.", | |
| "action_hz": 30.0, | |
| "embodiment_id": 25, | |
| "enable_primary_video_aug": false, | |
| "enable_primary_random_resized_crop": false, | |
| "guidance_scale": null, | |
| "num_inference_steps": null, | |
| "latent_action_placeholder_token": "<ACT_PH>", | |
| "num_action_queries": 8, | |
| "flow_action_num_queries": 8, | |
| "perceptual_weight": 0.1, | |
| "lam_encoder_distill_weight": 0.1, | |
| "enable_loss_distill": true, | |
| "future_prediction": true, | |
| "detach_future_feature": true, | |
| "repeated_diffusion_steps": 2, | |
| "flow_action_dim": 32, | |
| "flow_state_dim": 32, | |
| "flow_hidden_dim": 1024, | |
| "flow_num_layers": 16, | |
| "flow_attention_heads": 16, | |
| "flow_vlm_dim": 2048, | |
| "flow_vision_dim": 768, | |
| "flow_num_vision_tokens": 256, | |
| "flow_num_target_vision_tokens": -1, | |
| "flow_use_state": false, | |
| "flow_num_embodiments": 32, | |
| "flow_horizon_sec": 1.0, | |
| "flow_cfg_drop_prob": 0.0, | |
| "flow_cfg_guidance_scale": 1.0, | |
| "flow_num_inference_steps": 10, | |
| "flow_num_timestep_buckets": 1000, | |
| "flow_interleave_self_attention": true, | |
| "flow_use_alternate_vldit": true, | |
| "flow_attend_text_every_n_blocks": 2, | |
| "flow_noise_beta_alpha": 1.5, | |
| "flow_noise_beta_beta": 1.0, | |
| "flow_noise_s": 0.999, | |
| "flow_token_independent_noise": false, | |
| "flow_use_action_positional_embeddings": true, | |
| "clip_normalized_actions": false, | |
| "pre_snap_gripper_action": false, | |
| "binarize_gripper_action": false, | |
| "gripper_dim": 6, | |
| "gripper_threshold": 0.5, | |
| "freeze_vision_backbone": false, | |
| "freeze_llm_backbone": false, | |
| "freeze_embedding": true, | |
| "unfreeze_vision_merger": true, | |
| "unfreeze_lam_decoder": true, | |
| "keep_llm_first_n_layers": 16, | |
| "unfreeze_llm_last_n_layers": -1, | |
| "optimizer_lr": 0.0001, | |
| "optimizer_betas": [ | |
| 0.9, | |
| 0.95 | |
| ], | |
| "optimizer_eps": 1e-08, | |
| "optimizer_weight_decay": 1e-08, | |
| "optimizer_grad_clip_norm": 1.0, | |
| "scheduler_warmup_steps": 1500, | |
| "scheduler_decay_steps": 30000, | |
| "scheduler_decay_lr": 5e-07 | |
| } |