jstm/molmoact2_single_arm
017
1{2 "type": "molmoact2",3 "n_obs_steps": 1,4 "input_features": {5 "observation.state": {6 "type": "STATE",7 "shape": [8 99 ]10 },11 "observation.images.external1_camera": {12 "type": "VISUAL",13 "shape": [14 3,15 480,16 64017 ]18 },19 "observation.images.external2_camera": {20 "type": "VISUAL",21 "shape": [22 3,23 480,24 64025 ]26 },27 "observation.images.hand_camera": {28 "type": "VISUAL",29 "shape": [30 3,31 480,32 64033 ]34 }35 },36 "output_features": {37 "action": {38 "type": "ACTION",39 "shape": [40 741 ]42 }43 },44 "device": "cuda",45 "use_amp": false,46 "use_peft": false,47 "push_to_hub": true,48 "repo_id": "jstm/molmoact2_single_arm",49 "private": null,50 "tags": null,51 "license": null,52 "pretrained_path": null,53 "pretrained_revision": null,54 "checkpoint_path": "allenai/MolmoAct2-LIBERO",55 "checkpoint_revision": null,56 "checkpoint_force_download": false,57 "chunk_size": 10,58 "n_action_steps": 10,59 "action_mode": "continuous",60 "inference_action_mode": null,61 "discrete_action_tokenizer": "allenai/MolmoAct2-FAST-Tokenizer",62 "discrete_generation_max_steps": null,63 "norm_tag": null,64 "setup_type": "single franka robotic arm in maniskill",65 "control_mode": "delta end-effector pose",66 "image_keys": [67 "observation.images.external1_camera",68 "observation.images.external2_camera",69 "observation.images.hand_camera"70 ],71 "normalize_language": true,72 "add_setup_tokens": true,73 "add_control_tokens": true,74 "normalize_gripper": false,75 "num_state_tokens": 256,76 "max_sequence_length": null,77 "expected_max_action_dim": 32,78 "num_flow_timesteps": 8,79 "flow_matching_cutoff": 1.0,80 "flow_matching_time_offset": 0.001,81 "flow_matching_time_scale": 0.999,82 "flow_matching_beta_alpha": 1.0,83 "flow_matching_beta_beta": 1.5,84 "num_inference_steps": null,85 "mask_action_dim_padding": true,86 "enable_inference_cuda_graph": true,87 "per_episode_seed": false,88 "eval_seed": null,89 "rtc_config": null,90 "joint_signs": null,91 "joint_offsets": null,92 "enable_lora_vlm": false,93 "lora_rank": 64,94 "lora_alpha": 16,95 "lora_dropout": 0.05,96 "lora_bias": "none",97 "enable_lora_action_expert": false,98 "enable_knowledge_insulation": false,99 "freeze_embedding": true,100 "train_action_expert_only": true,101 "gradient_checkpointing": true,102 "model_dtype": "bfloat16",103 "softmax_auxiliary_loss": true,104 "softmax_auxiliary_loss_scale": 0.0001,105 "discrete_loss_token_weighting": "root_subsegments_root_tokens",106 "optimizer_lr": 1e-05,107 "optimizer_vit_lr": 5e-06,108 "optimizer_connector_lr": 5e-06,109 "optimizer_action_expert_lr": 5e-05,110 "optimizer_betas": [111 0.9,112 0.95113 ],114 "optimizer_eps": 1e-06,115 "optimizer_weight_decay": 0.0,116 "optimizer_grad_clip_norm": 1.0,117 "scheduler_warmup_steps": 200,118 "scheduler_decay_steps": 100000,119 "scheduler_decay_lr": 1e-06,120 "normalization_mapping": {121 "VISUAL": "IDENTITY",122 "STATE": "QUANTILES",123 "ACTION": "QUANTILES"124 },125 "dataset_feature_names": {126 "action": [127 "action_0",128 "action_1",129 "action_2",130 "action_3",131 "action_4",132 "action_5",133 "action_6"134 ],135 "observation.state": [136 "joint_0",137 "joint_1",138 "joint_2",139 "joint_3",140 "joint_4",141 "joint_5",142 "joint_6",143 "joint_7",144 "joint_8"145 ]146 }147}