processor_config.json
| 1 | { |
| 2 | "processor_class": "Gr00tN1d6Processor", |
| 3 | "processor_kwargs": { |
| 4 | "modality_configs": { |
| 5 | "behavior_r1_pro": { |
| 6 | "video": { |
| 7 | "delta_indices": [ |
| 8 | 0 |
| 9 | ], |
| 10 | "modality_keys": [ |
| 11 | "observation.images.rgb.head_256_256", |
| 12 | "observation.images.rgb.left_wrist_256_256", |
| 13 | "observation.images.rgb.right_wrist_256_256" |
| 14 | ], |
| 15 | "sin_cos_embedding_keys": null, |
| 16 | "mean_std_embedding_keys": null, |
| 17 | "action_configs": null |
| 18 | }, |
| 19 | "state": { |
| 20 | "delta_indices": [ |
| 21 | 0 |
| 22 | ], |
| 23 | "modality_keys": [ |
| 24 | "robot_pos", |
| 25 | "robot_ori_cos", |
| 26 | "robot_ori_sin", |
| 27 | "robot_2d_ori", |
| 28 | "robot_2d_ori_cos", |
| 29 | "robot_2d_ori_sin", |
| 30 | "robot_lin_vel", |
| 31 | "robot_ang_vel", |
| 32 | "arm_left_qpos", |
| 33 | "arm_left_qpos_sin", |
| 34 | "arm_left_qpos_cos", |
| 35 | "eef_left_pos", |
| 36 | "eef_left_quat", |
| 37 | "gripper_left_qpos", |
| 38 | "arm_right_qpos", |
| 39 | "arm_right_qpos_sin", |
| 40 | "arm_right_qpos_cos", |
| 41 | "eef_right_pos", |
| 42 | "eef_right_quat", |
| 43 | "gripper_right_qpos", |
| 44 | "trunk_qpos" |
| 45 | ], |
| 46 | "sin_cos_embedding_keys": null, |
| 47 | "mean_std_embedding_keys": null, |
| 48 | "action_configs": null |
| 49 | }, |
| 50 | "action": { |
| 51 | "delta_indices": [ |
| 52 | 0, |
| 53 | 1, |
| 54 | 2, |
| 55 | 3, |
| 56 | 4, |
| 57 | 5, |
| 58 | 6, |
| 59 | 7, |
| 60 | 8, |
| 61 | 9, |
| 62 | 10, |
| 63 | 11, |
| 64 | 12, |
| 65 | 13, |
| 66 | 14, |
| 67 | 15, |
| 68 | 16, |
| 69 | 17, |
| 70 | 18, |
| 71 | 19, |
| 72 | 20, |
| 73 | 21, |
| 74 | 22, |
| 75 | 23, |
| 76 | 24, |
| 77 | 25, |
| 78 | 26, |
| 79 | 27, |
| 80 | 28, |
| 81 | 29, |
| 82 | 30, |
| 83 | 31 |
| 84 | ], |
| 85 | "modality_keys": [ |
| 86 | "base", |
| 87 | "torso", |
| 88 | "left_arm", |
| 89 | "left_gripper", |
| 90 | "right_arm", |
| 91 | "right_gripper" |
| 92 | ], |
| 93 | "sin_cos_embedding_keys": null, |
| 94 | "mean_std_embedding_keys": null, |
| 95 | "action_configs": [ |
| 96 | { |
| 97 | "rep": "ABSOLUTE", |
| 98 | "type": "NON_EEF", |
| 99 | "format": "DEFAULT", |
| 100 | "state_key": null |
| 101 | }, |
| 102 | { |
| 103 | "rep": "RELATIVE", |
| 104 | "type": "NON_EEF", |
| 105 | "format": "DEFAULT", |
| 106 | "state_key": "trunk_qpos" |
| 107 | }, |
| 108 | { |
| 109 | "rep": "RELATIVE", |
| 110 | "type": "NON_EEF", |
| 111 | "format": "DEFAULT", |
| 112 | "state_key": "arm_left_qpos" |
| 113 | }, |
| 114 | { |
| 115 | "rep": "ABSOLUTE", |
| 116 | "type": "NON_EEF", |
| 117 | "format": "DEFAULT", |
| 118 | "state_key": null |
| 119 | }, |
| 120 | { |
| 121 | "rep": "RELATIVE", |
| 122 | "type": "NON_EEF", |
| 123 | "format": "DEFAULT", |
| 124 | "state_key": "arm_right_qpos" |
| 125 | }, |
| 126 | { |
| 127 | "rep": "ABSOLUTE", |
| 128 | "type": "NON_EEF", |
| 129 | "format": "DEFAULT", |
| 130 | "state_key": null |
| 131 | } |
| 132 | ] |
| 133 | }, |
| 134 | "language": { |
| 135 | "delta_indices": [ |
| 136 | 0 |
| 137 | ], |
| 138 | "modality_keys": [ |
| 139 | "annotation.human.coarse_action" |
| 140 | ], |
| 141 | "sin_cos_embedding_keys": null, |
| 142 | "mean_std_embedding_keys": null, |
| 143 | "action_configs": null |
| 144 | } |
| 145 | }, |
| 146 | "gr1": { |
| 147 | "video": { |
| 148 | "delta_indices": [ |
| 149 | 0 |
| 150 | ], |
| 151 | "modality_keys": [ |
| 152 | "ego_view_bg_crop_pad_res256_freq20" |
| 153 | ], |
| 154 | "sin_cos_embedding_keys": null, |
| 155 | "mean_std_embedding_keys": null, |
| 156 | "action_configs": null |
| 157 | }, |
| 158 | "state": { |
| 159 | "delta_indices": [ |
| 160 | 0 |
| 161 | ], |
| 162 | "modality_keys": [ |
| 163 | "left_arm", |
| 164 | "right_arm", |
| 165 | "left_hand", |
| 166 | "right_hand", |
| 167 | "waist" |
| 168 | ], |
| 169 | "sin_cos_embedding_keys": [ |
| 170 | "left_arm", |
| 171 | "right_arm", |
| 172 | "left_hand", |
| 173 | "right_hand", |
| 174 | "waist" |
| 175 | ], |
| 176 | "mean_std_embedding_keys": null, |
| 177 | "action_configs": null |
| 178 | }, |
| 179 | "action": { |
| 180 | "delta_indices": [ |
| 181 | 0, |
| 182 | 1, |
| 183 | 2, |
| 184 | 3, |
| 185 | 4, |
| 186 | 5, |
| 187 | 6, |
| 188 | 7, |
| 189 | 8, |
| 190 | 9, |
| 191 | 10, |
| 192 | 11, |
| 193 | 12, |
| 194 | 13, |
| 195 | 14, |
| 196 | 15 |
| 197 | ], |
| 198 | "modality_keys": [ |
| 199 | "left_arm", |
| 200 | "right_arm", |
| 201 | "left_hand", |
| 202 | "right_hand", |
| 203 | "waist" |
| 204 | ], |
| 205 | "sin_cos_embedding_keys": null, |
| 206 | "mean_std_embedding_keys": null, |
| 207 | "action_configs": [ |
| 208 | { |
| 209 | "rep": "RELATIVE", |
| 210 | "type": "NON_EEF", |
| 211 | "format": "DEFAULT", |
| 212 | "state_key": null |
| 213 | }, |
| 214 | { |
| 215 | "rep": "RELATIVE", |
| 216 | "type": "NON_EEF", |
| 217 | "format": "DEFAULT", |
| 218 | "state_key": null |
| 219 | }, |
| 220 | { |
| 221 | "rep": "RELATIVE", |
| 222 | "type": "NON_EEF", |
| 223 | "format": "DEFAULT", |
| 224 | "state_key": null |
| 225 | }, |
| 226 | { |
| 227 | "rep": "RELATIVE", |
| 228 | "type": "NON_EEF", |
| 229 | "format": "DEFAULT", |
| 230 | "state_key": null |
| 231 | }, |
| 232 | { |
| 233 | "rep": "ABSOLUTE", |
| 234 | "type": "NON_EEF", |
| 235 | "format": "DEFAULT", |
| 236 | "state_key": null |
| 237 | } |
| 238 | ] |
| 239 | }, |
| 240 | "language": { |
| 241 | "delta_indices": [ |
| 242 | 0 |
| 243 | ], |
| 244 | "modality_keys": [ |
| 245 | "task" |
| 246 | ], |
| 247 | "sin_cos_embedding_keys": null, |
| 248 | "mean_std_embedding_keys": null, |
| 249 | "action_configs": null |
| 250 | } |
| 251 | }, |
| 252 | "robocasa_panda_omron": { |
| 253 | "video": { |
| 254 | "delta_indices": [ |
| 255 | 0 |
| 256 | ], |
| 257 | "modality_keys": [ |
| 258 | "res256_image_side_0", |
| 259 | "res256_image_side_1", |
| 260 | "res256_image_wrist_0" |
| 261 | ], |
| 262 | "sin_cos_embedding_keys": null, |
| 263 | "mean_std_embedding_keys": null, |
| 264 | "action_configs": null |
| 265 | }, |
| 266 | "state": { |
| 267 | "delta_indices": [ |
| 268 | 0 |
| 269 | ], |
| 270 | "modality_keys": [ |
| 271 | "end_effector_position_relative", |
| 272 | "end_effector_rotation_relative", |
| 273 | "gripper_qpos", |
| 274 | "base_position", |
| 275 | "base_rotation" |
| 276 | ], |
| 277 | "sin_cos_embedding_keys": null, |
| 278 | "mean_std_embedding_keys": null, |
| 279 | "action_configs": null |
| 280 | }, |
| 281 | "action": { |
| 282 | "delta_indices": [ |
| 283 | 0, |
| 284 | 1, |
| 285 | 2, |
| 286 | 3, |
| 287 | 4, |
| 288 | 5, |
| 289 | 6, |
| 290 | 7, |
| 291 | 8, |
| 292 | 9, |
| 293 | 10, |
| 294 | 11, |
| 295 | 12, |
| 296 | 13, |
| 297 | 14, |
| 298 | 15 |
| 299 | ], |
| 300 | "modality_keys": [ |
| 301 | "end_effector_position", |
| 302 | "end_effector_rotation", |
| 303 | "gripper_close", |
| 304 | "base_motion", |
| 305 | "control_mode" |
| 306 | ], |
| 307 | "sin_cos_embedding_keys": null, |
| 308 | "mean_std_embedding_keys": null, |
| 309 | "action_configs": [ |
| 310 | { |
| 311 | "rep": "ABSOLUTE", |
| 312 | "type": "NON_EEF", |
| 313 | "format": "DEFAULT", |
| 314 | "state_key": null |
| 315 | }, |
| 316 | { |
| 317 | "rep": "ABSOLUTE", |
| 318 | "type": "NON_EEF", |
| 319 | "format": "DEFAULT", |
| 320 | "state_key": null |
| 321 | }, |
| 322 | { |
| 323 | "rep": "ABSOLUTE", |
| 324 | "type": "NON_EEF", |
| 325 | "format": "DEFAULT", |
| 326 | "state_key": null |
| 327 | }, |
| 328 | { |
| 329 | "rep": "ABSOLUTE", |
| 330 | "type": "NON_EEF", |
| 331 | "format": "DEFAULT", |
| 332 | "state_key": null |
| 333 | }, |
| 334 | { |
| 335 | "rep": "ABSOLUTE", |
| 336 | "type": "NON_EEF", |
| 337 | "format": "DEFAULT", |
| 338 | "state_key": null |
| 339 | } |
| 340 | ] |
| 341 | }, |
| 342 | "language": { |
| 343 | "delta_indices": [ |
| 344 | 0 |
| 345 | ], |
| 346 | "modality_keys": [ |
| 347 | "annotation.human.action.task_description" |
| 348 | ], |
| 349 | "sin_cos_embedding_keys": null, |
| 350 | "mean_std_embedding_keys": null, |
| 351 | "action_configs": null |
| 352 | } |
| 353 | } |
| 354 | }, |
| 355 | "image_crop_size": null, |
| 356 | "image_target_size": null, |
| 357 | "use_albumentations": true, |
| 358 | "random_rotation_angle": null, |
| 359 | "color_jitter_params": { |
| 360 | "brightness": 0.3, |
| 361 | "contrast": 0.4, |
| 362 | "saturation": 0.5, |
| 363 | "hue": 0.08 |
| 364 | }, |
| 365 | "shortest_image_edge": 256, |
| 366 | "crop_fraction": 0.95, |
| 367 | "model_name": "nvidia/Eagle-Block2A-2B-v2", |
| 368 | "model_type": "eagle", |
| 369 | "formalize_language": true, |
| 370 | "max_state_dim": 128, |
| 371 | "max_action_dim": 128, |
| 372 | "max_action_horizon": 50, |
| 373 | "use_percentiles": false, |
| 374 | "clip_outliers": true, |
| 375 | "apply_sincos_state_encoding": true, |
| 376 | "use_relative_action": true |
| 377 | } |
| 378 | } |