Drop policy_config from the tokenizer step
Browse files`policy_config` duplicated fields that `config.json` already owns (`author_model_config`, `processor_metadata`, `camera_order`, `runtime_system`, ...). The copy could not be kept in sync, which is why the loader had to re-apply `runtime_system` onto the step after every load.
The live `G05Config` is now authoritative: `G05TokenizerStep.get_config()` no longer serializes `policy_config`, and `reconcile_g05_processors` re-injects it on load. Dropping it here shrinks this file by about 80%. No behaviour change; weights are untouched.
- policy_preprocessor.json +0 -769
policy_preprocessor.json
CHANGED
|
@@ -251,775 +251,6 @@
|
|
| 251 |
"registry_name": "g05_tokenizer",
|
| 252 |
"config": {
|
| 253 |
"checkpoint_path": "lerobot/g05_base",
|
| 254 |
-
"policy_config": {
|
| 255 |
-
"author_model_config": {
|
| 256 |
-
"pretrained_model_path": null,
|
| 257 |
-
"hf_processor_class": "g05.models.g05.qwen35.processing.Qwen35ProcessorWrapper",
|
| 258 |
-
"base_vocab_size": 248044,
|
| 259 |
-
"padded_vocab_size": 248320,
|
| 260 |
-
"add_loc_tokens": false,
|
| 261 |
-
"image_token_index": 248056,
|
| 262 |
-
"pad_token_id": 0,
|
| 263 |
-
"vocab_size": 248320,
|
| 264 |
-
"eos_token_id": 248044,
|
| 265 |
-
"action_dim": 27,
|
| 266 |
-
"proprio_dim": 27,
|
| 267 |
-
"proprio_encoder": "mlp",
|
| 268 |
-
"horizon_steps": 32,
|
| 269 |
-
"cond_steps": 1,
|
| 270 |
-
"num_obs_steps": 1,
|
| 271 |
-
"attn_implementation": "eager",
|
| 272 |
-
"position_ids_type": "pi0fast",
|
| 273 |
-
"action_position_offset": null,
|
| 274 |
-
"checkpoint_vision": true,
|
| 275 |
-
"checkpoint_vlm": false,
|
| 276 |
-
"checkpoint_action_expert": true,
|
| 277 |
-
"vlm": {
|
| 278 |
-
"name": "g05.models.g05.qwen35.mixture_qwen35.MixtureQwen35",
|
| 279 |
-
"hidden_size": 2048,
|
| 280 |
-
"intermediate_size": 6144,
|
| 281 |
-
"num_hidden_layers": 24,
|
| 282 |
-
"num_attention_heads": 8,
|
| 283 |
-
"num_key_value_heads": 2,
|
| 284 |
-
"head_dim": 256,
|
| 285 |
-
"rms_norm_eps": 1e-06,
|
| 286 |
-
"max_position_embeddings": 262144,
|
| 287 |
-
"attention_bias": false,
|
| 288 |
-
"input_type": "embedding",
|
| 289 |
-
"vocab_size": 248320,
|
| 290 |
-
"pad_token_id": 0,
|
| 291 |
-
"output_type": "lm_head",
|
| 292 |
-
"adaptive_mode": null,
|
| 293 |
-
"time_hidden_size": 0,
|
| 294 |
-
"use_final_norm": true,
|
| 295 |
-
"hidden_act": "silu",
|
| 296 |
-
"rope_parameters": {
|
| 297 |
-
"rope_theta": 10000000.0,
|
| 298 |
-
"rope_type": "default",
|
| 299 |
-
"mrope_section": [
|
| 300 |
-
11,
|
| 301 |
-
11,
|
| 302 |
-
10
|
| 303 |
-
],
|
| 304 |
-
"mrope_interleaved": true,
|
| 305 |
-
"partial_rotary_factor": 0.25
|
| 306 |
-
},
|
| 307 |
-
"linear_attn_backend": "fla",
|
| 308 |
-
"linear_conv_kernel_dim": 4,
|
| 309 |
-
"linear_key_head_dim": 128,
|
| 310 |
-
"linear_value_head_dim": 128,
|
| 311 |
-
"linear_num_key_heads": 16,
|
| 312 |
-
"linear_num_value_heads": 16,
|
| 313 |
-
"layer_types": [
|
| 314 |
-
"linear_attention",
|
| 315 |
-
"linear_attention",
|
| 316 |
-
"linear_attention",
|
| 317 |
-
"full_attention",
|
| 318 |
-
"linear_attention",
|
| 319 |
-
"linear_attention",
|
| 320 |
-
"linear_attention",
|
| 321 |
-
"full_attention",
|
| 322 |
-
"linear_attention",
|
| 323 |
-
"linear_attention",
|
| 324 |
-
"linear_attention",
|
| 325 |
-
"full_attention",
|
| 326 |
-
"linear_attention",
|
| 327 |
-
"linear_attention",
|
| 328 |
-
"linear_attention",
|
| 329 |
-
"full_attention",
|
| 330 |
-
"linear_attention",
|
| 331 |
-
"linear_attention",
|
| 332 |
-
"linear_attention",
|
| 333 |
-
"full_attention",
|
| 334 |
-
"linear_attention",
|
| 335 |
-
"linear_attention",
|
| 336 |
-
"linear_attention",
|
| 337 |
-
"full_attention"
|
| 338 |
-
]
|
| 339 |
-
},
|
| 340 |
-
"action_expert": {
|
| 341 |
-
"name": "g05.models.g05.qwen35.mixture_qwen35.MixtureQwen35",
|
| 342 |
-
"hidden_size": 1024,
|
| 343 |
-
"intermediate_size": 4096,
|
| 344 |
-
"num_hidden_layers": 24,
|
| 345 |
-
"num_attention_heads": 8,
|
| 346 |
-
"num_key_value_heads": 2,
|
| 347 |
-
"head_dim": 256,
|
| 348 |
-
"rms_norm_eps": 1e-06,
|
| 349 |
-
"max_position_embeddings": 262144,
|
| 350 |
-
"attention_bias": false,
|
| 351 |
-
"adaptive_mode": "adaLN",
|
| 352 |
-
"time_hidden_size": 1024,
|
| 353 |
-
"input_type": "linear",
|
| 354 |
-
"input_dim": 27,
|
| 355 |
-
"output_type": "linear",
|
| 356 |
-
"output_dim": 27,
|
| 357 |
-
"use_final_norm": true,
|
| 358 |
-
"hidden_act": "silu",
|
| 359 |
-
"rope_parameters": {
|
| 360 |
-
"rope_theta": 10000000.0,
|
| 361 |
-
"rope_type": "default",
|
| 362 |
-
"mrope_section": [
|
| 363 |
-
11,
|
| 364 |
-
11,
|
| 365 |
-
10
|
| 366 |
-
],
|
| 367 |
-
"mrope_interleaved": true,
|
| 368 |
-
"partial_rotary_factor": 0.25
|
| 369 |
-
},
|
| 370 |
-
"linear_attn_backend": "fla",
|
| 371 |
-
"linear_conv_kernel_dim": 4,
|
| 372 |
-
"linear_key_head_dim": 128,
|
| 373 |
-
"linear_value_head_dim": 128,
|
| 374 |
-
"linear_num_key_heads": 16,
|
| 375 |
-
"linear_num_value_heads": 16,
|
| 376 |
-
"layer_types": [
|
| 377 |
-
"full_attention",
|
| 378 |
-
"full_attention",
|
| 379 |
-
"full_attention",
|
| 380 |
-
"full_attention",
|
| 381 |
-
"full_attention",
|
| 382 |
-
"full_attention",
|
| 383 |
-
"full_attention",
|
| 384 |
-
"full_attention",
|
| 385 |
-
"full_attention",
|
| 386 |
-
"full_attention",
|
| 387 |
-
"full_attention",
|
| 388 |
-
"full_attention",
|
| 389 |
-
"full_attention",
|
| 390 |
-
"full_attention",
|
| 391 |
-
"full_attention",
|
| 392 |
-
"full_attention",
|
| 393 |
-
"full_attention",
|
| 394 |
-
"full_attention",
|
| 395 |
-
"full_attention",
|
| 396 |
-
"full_attention",
|
| 397 |
-
"full_attention",
|
| 398 |
-
"full_attention",
|
| 399 |
-
"full_attention",
|
| 400 |
-
"full_attention"
|
| 401 |
-
]
|
| 402 |
-
},
|
| 403 |
-
"fm": {
|
| 404 |
-
"time_convention": "pi_convention",
|
| 405 |
-
"flow_sampling": "beta",
|
| 406 |
-
"flow_sig_min": 0.001,
|
| 407 |
-
"num_inference_steps": 10,
|
| 408 |
-
"num_flow_samples": 4,
|
| 409 |
-
"fm_weight": 1.0,
|
| 410 |
-
"joint_training": false,
|
| 411 |
-
"padding_action_weight": 0.0,
|
| 412 |
-
"zero_pad_action_target": false,
|
| 413 |
-
"final_action_clip_value": null,
|
| 414 |
-
"action_causal": false,
|
| 415 |
-
"horizon_steps": 32,
|
| 416 |
-
"action_dim": 27,
|
| 417 |
-
"use_correlated_noise": false,
|
| 418 |
-
"correlation_beta": 0.5,
|
| 419 |
-
"action_correlation_cholesky_path": "data/stats/action_correlation_cholesky.npy",
|
| 420 |
-
"correlated_noise_embodiments": []
|
| 421 |
-
},
|
| 422 |
-
"ar": {
|
| 423 |
-
"ce_weight": 1.0,
|
| 424 |
-
"vocab_size": 248320,
|
| 425 |
-
"text_token_end": 248044,
|
| 426 |
-
"loc_token_end": 248077,
|
| 427 |
-
"use_fused_ce": true,
|
| 428 |
-
"ce_z_loss_scale": 0.0,
|
| 429 |
-
"block_wise_autoregressive": false,
|
| 430 |
-
"bos_blk_id": null,
|
| 431 |
-
"eos_blk_id": null,
|
| 432 |
-
"block_size": null,
|
| 433 |
-
"do_sample": false,
|
| 434 |
-
"max_new_tokens": 300,
|
| 435 |
-
"temperature": 0.7,
|
| 436 |
-
"top_k": 128,
|
| 437 |
-
"top_p": 0.95,
|
| 438 |
-
"repetition_penalty": 1.2,
|
| 439 |
-
"no_repeat_ngram_size": 3
|
| 440 |
-
},
|
| 441 |
-
"vision": {
|
| 442 |
-
"name": "g05.models.g05.qwen35.vision.Qwen3_5VisionModel",
|
| 443 |
-
"depth": 24,
|
| 444 |
-
"hidden_size": 1024,
|
| 445 |
-
"num_heads": 16,
|
| 446 |
-
"patch_size": 16,
|
| 447 |
-
"temporal_patch_size": 2,
|
| 448 |
-
"spatial_merge_size": 2,
|
| 449 |
-
"in_channels": 3,
|
| 450 |
-
"intermediate_size": 4096,
|
| 451 |
-
"out_hidden_size": 2048,
|
| 452 |
-
"num_position_embeddings": 2304,
|
| 453 |
-
"hidden_act": "gelu_pytorch_tanh",
|
| 454 |
-
"num_image_tokens": null,
|
| 455 |
-
"num_channels": 3,
|
| 456 |
-
"temporal_freq": 4,
|
| 457 |
-
"spacetime_mode": "factorized",
|
| 458 |
-
"token_drop_layer": null,
|
| 459 |
-
"temporal_pe_pretrain_frames": null,
|
| 460 |
-
"use_temporal_conv3d": false,
|
| 461 |
-
"batch_all_cameras": false
|
| 462 |
-
},
|
| 463 |
-
"vision_projector": {
|
| 464 |
-
"name": null
|
| 465 |
-
},
|
| 466 |
-
"action_tokenizer": "g05.tokenizer.interface.vq_base.VQActionTokenizer",
|
| 467 |
-
"AT_CONFIG": {
|
| 468 |
-
"vqvae_type": "g05.tokenizer.models.actioncodec2_v2.wrapper.ActionCodecV2Wrapper",
|
| 469 |
-
"block_wise_autoregressive": false,
|
| 470 |
-
"use_extra_tokens": true,
|
| 471 |
-
"block_size": 8,
|
| 472 |
-
"parts_meta": {
|
| 473 |
-
"left_control": 9,
|
| 474 |
-
"left_gripper": 1,
|
| 475 |
-
"right_control": 9,
|
| 476 |
-
"right_gripper": 1,
|
| 477 |
-
"lower_body": 7
|
| 478 |
-
},
|
| 479 |
-
"rule_based_key_patterns": [
|
| 480 |
-
"gripper"
|
| 481 |
-
],
|
| 482 |
-
"rule_based_min_block_len": 1,
|
| 483 |
-
"rule_based_binarize_threshold": 0.0,
|
| 484 |
-
"ckpt_dir": "checkpoints/action_tokenizer.pt",
|
| 485 |
-
"model_arch": {
|
| 486 |
-
"max_component_dim": 9,
|
| 487 |
-
"horizon": 32,
|
| 488 |
-
"horizon_patch_size": 8,
|
| 489 |
-
"conv_in_action_kernel": 2,
|
| 490 |
-
"encoder_channels": 256,
|
| 491 |
-
"latent_dim": 128,
|
| 492 |
-
"c_mults": [
|
| 493 |
-
1,
|
| 494 |
-
2,
|
| 495 |
-
2
|
| 496 |
-
],
|
| 497 |
-
"strides": [
|
| 498 |
-
[
|
| 499 |
-
1,
|
| 500 |
-
1
|
| 501 |
-
],
|
| 502 |
-
[
|
| 503 |
-
2,
|
| 504 |
-
1
|
| 505 |
-
],
|
| 506 |
-
[
|
| 507 |
-
2,
|
| 508 |
-
1
|
| 509 |
-
]
|
| 510 |
-
],
|
| 511 |
-
"transformer_depths": [
|
| 512 |
-
2,
|
| 513 |
-
2,
|
| 514 |
-
2
|
| 515 |
-
],
|
| 516 |
-
"num_heads": 8,
|
| 517 |
-
"dim_heads": 64,
|
| 518 |
-
"ffn_mult": 4.0,
|
| 519 |
-
"use_layer_scale": true,
|
| 520 |
-
"layer_scale_init": 0.01,
|
| 521 |
-
"use_qk_norm": true,
|
| 522 |
-
"rope_base": 10000,
|
| 523 |
-
"use_block_dct": true,
|
| 524 |
-
"block_dct_block_size": 8,
|
| 525 |
-
"n_codebooks": 4,
|
| 526 |
-
"codebook_size": 4096,
|
| 527 |
-
"codebook_dim": 8,
|
| 528 |
-
"quantizer_dropout": 0.5,
|
| 529 |
-
"ema_decay": 0.95,
|
| 530 |
-
"threshold_ema_dead": 2,
|
| 531 |
-
"use_rotation_trick": false,
|
| 532 |
-
"commitment_loss_weight": 0.25,
|
| 533 |
-
"reconstruction_loss_weight": 1.0,
|
| 534 |
-
"consistency_loss_weight": 0.0,
|
| 535 |
-
"consistency_delta_max_begin": 1,
|
| 536 |
-
"consistency_delta_max_end": 3,
|
| 537 |
-
"consistency_eps_begin": 0.02,
|
| 538 |
-
"consistency_eps_end": 0.1,
|
| 539 |
-
"consistency_layer_weights_begin": [
|
| 540 |
-
0.0,
|
| 541 |
-
0.0,
|
| 542 |
-
0.0,
|
| 543 |
-
0.0
|
| 544 |
-
],
|
| 545 |
-
"consistency_layer_weights_end": [
|
| 546 |
-
1.0,
|
| 547 |
-
0.7,
|
| 548 |
-
0.3,
|
| 549 |
-
0.1
|
| 550 |
-
],
|
| 551 |
-
"consistency_schedule_begin_fraction": 0.2,
|
| 552 |
-
"consistency_schedule_end_fraction": 0.8
|
| 553 |
-
},
|
| 554 |
-
"action_template": "{action}",
|
| 555 |
-
"num_residuals": 2,
|
| 556 |
-
"use_group_markers": true,
|
| 557 |
-
"group_order_shuffle": false,
|
| 558 |
-
"dropout_noop_parts": true,
|
| 559 |
-
"absent_key_fill_value": -100.0
|
| 560 |
-
},
|
| 561 |
-
"vla_training_strategy": "vla-full-train",
|
| 562 |
-
"backbone_lr_multiplier": 1.0,
|
| 563 |
-
"max_text_tokens": 160,
|
| 564 |
-
"num_input_images": 18,
|
| 565 |
-
"camera_size_config": {
|
| 566 |
-
"exterior": [
|
| 567 |
-
256,
|
| 568 |
-
256
|
| 569 |
-
],
|
| 570 |
-
"wrist_left": [
|
| 571 |
-
256,
|
| 572 |
-
256
|
| 573 |
-
],
|
| 574 |
-
"wrist_right": [
|
| 575 |
-
256,
|
| 576 |
-
256
|
| 577 |
-
]
|
| 578 |
-
},
|
| 579 |
-
"num_extra_image_tokens_per_camera": 0,
|
| 580 |
-
"use_lm_head": true,
|
| 581 |
-
"discrete_action": true,
|
| 582 |
-
"continuous_action": true,
|
| 583 |
-
"return_continuous_action": true,
|
| 584 |
-
"action_attend_cot": false,
|
| 585 |
-
"predict_cot": true,
|
| 586 |
-
"max_chunk_token_length": 1024,
|
| 587 |
-
"max_pad_token_length": null,
|
| 588 |
-
"input_preprocessor": {
|
| 589 |
-
"input_action_corruption": false,
|
| 590 |
-
"pred_eov": true,
|
| 591 |
-
"batchify_action": true,
|
| 592 |
-
"pi05_ft_mode": false
|
| 593 |
-
},
|
| 594 |
-
"ae_vlm_condition_mode": "cross_attn_only",
|
| 595 |
-
"mem_frame_drop_prob": 0.3,
|
| 596 |
-
"vision_lr_multiplier": 1.0,
|
| 597 |
-
"hf_processor_path": "checkpoints/qwen3_5_2b_base_processor"
|
| 598 |
-
},
|
| 599 |
-
"embodiment": "galaxea_r1lite",
|
| 600 |
-
"predict_cot": true,
|
| 601 |
-
"runtime_system": "system2",
|
| 602 |
-
"prompt_template": "<chat_user_prefix><image0_image_!><image1_image_!><image2_image_!><bos>Embodiment: <embodiment_text_!>; Task: <command_text_!_200> State: <proprio_proprio_!>;<chat_user_suffix><chat_assistant_prefix><prompt_text_!>\n<EOC><atomic_task_text>|Action: <EOV><action_action>|<eos>",
|
| 603 |
-
"num_prompt_images": 3,
|
| 604 |
-
"num_input_images": 18,
|
| 605 |
-
"camera_order": [
|
| 606 |
-
"observation.images.head_rgb",
|
| 607 |
-
"observation.images.left_wrist_rgb",
|
| 608 |
-
"observation.images.right_wrist_rgb"
|
| 609 |
-
],
|
| 610 |
-
"camera_sizes": {
|
| 611 |
-
"observation.images.head_rgb": [
|
| 612 |
-
256,
|
| 613 |
-
256
|
| 614 |
-
],
|
| 615 |
-
"observation.images.left_wrist_rgb": [
|
| 616 |
-
256,
|
| 617 |
-
256
|
| 618 |
-
],
|
| 619 |
-
"observation.images.right_wrist_rgb": [
|
| 620 |
-
256,
|
| 621 |
-
256
|
| 622 |
-
]
|
| 623 |
-
},
|
| 624 |
-
"policy_state_dim": 27,
|
| 625 |
-
"policy_action_dim": 27,
|
| 626 |
-
"chunk_size": 32,
|
| 627 |
-
"processor_metadata": {
|
| 628 |
-
"shape_meta": {
|
| 629 |
-
"action": [
|
| 630 |
-
{
|
| 631 |
-
"key": "left_arm",
|
| 632 |
-
"lerobot_key": "observation.state.left_arm",
|
| 633 |
-
"start_index": 0,
|
| 634 |
-
"raw_shape": 6,
|
| 635 |
-
"shape": 6,
|
| 636 |
-
"time_offset": 1
|
| 637 |
-
},
|
| 638 |
-
{
|
| 639 |
-
"key": "left_gripper",
|
| 640 |
-
"lerobot_key": "observation.state.left_gripper",
|
| 641 |
-
"start_index": 0,
|
| 642 |
-
"raw_shape": 1,
|
| 643 |
-
"shape": 1,
|
| 644 |
-
"time_offset": 1
|
| 645 |
-
},
|
| 646 |
-
{
|
| 647 |
-
"key": "right_arm",
|
| 648 |
-
"lerobot_key": "observation.state.right_arm",
|
| 649 |
-
"start_index": 0,
|
| 650 |
-
"raw_shape": 6,
|
| 651 |
-
"shape": 6,
|
| 652 |
-
"time_offset": 1
|
| 653 |
-
},
|
| 654 |
-
{
|
| 655 |
-
"key": "right_gripper",
|
| 656 |
-
"lerobot_key": "observation.state.right_gripper",
|
| 657 |
-
"start_index": 0,
|
| 658 |
-
"raw_shape": 1,
|
| 659 |
-
"shape": 1,
|
| 660 |
-
"time_offset": 1
|
| 661 |
-
}
|
| 662 |
-
],
|
| 663 |
-
"state": [
|
| 664 |
-
{
|
| 665 |
-
"key": "left_arm",
|
| 666 |
-
"lerobot_key": "observation.state.left_arm",
|
| 667 |
-
"start_index": 0,
|
| 668 |
-
"raw_shape": 6,
|
| 669 |
-
"shape": 6,
|
| 670 |
-
"time_offset": 0
|
| 671 |
-
},
|
| 672 |
-
{
|
| 673 |
-
"key": "left_gripper",
|
| 674 |
-
"lerobot_key": "observation.state.left_gripper",
|
| 675 |
-
"start_index": 0,
|
| 676 |
-
"raw_shape": 1,
|
| 677 |
-
"shape": 1,
|
| 678 |
-
"time_offset": 0
|
| 679 |
-
},
|
| 680 |
-
{
|
| 681 |
-
"key": "right_arm",
|
| 682 |
-
"lerobot_key": "observation.state.right_arm",
|
| 683 |
-
"start_index": 0,
|
| 684 |
-
"raw_shape": 6,
|
| 685 |
-
"shape": 6,
|
| 686 |
-
"time_offset": 0
|
| 687 |
-
},
|
| 688 |
-
{
|
| 689 |
-
"key": "right_gripper",
|
| 690 |
-
"lerobot_key": "observation.state.right_gripper",
|
| 691 |
-
"start_index": 0,
|
| 692 |
-
"raw_shape": 1,
|
| 693 |
-
"shape": 1,
|
| 694 |
-
"time_offset": 0
|
| 695 |
-
}
|
| 696 |
-
],
|
| 697 |
-
"images": [
|
| 698 |
-
{
|
| 699 |
-
"key": "head_rgb",
|
| 700 |
-
"camera_type": "exterior",
|
| 701 |
-
"lerobot_key": "observation.images.head_rgb",
|
| 702 |
-
"start_index": 0,
|
| 703 |
-
"raw_shape": [
|
| 704 |
-
3,
|
| 705 |
-
720,
|
| 706 |
-
1280
|
| 707 |
-
],
|
| 708 |
-
"shape": [
|
| 709 |
-
3,
|
| 710 |
-
224,
|
| 711 |
-
224
|
| 712 |
-
],
|
| 713 |
-
"time_offset": 0
|
| 714 |
-
},
|
| 715 |
-
{
|
| 716 |
-
"key": "left_wrist_rgb",
|
| 717 |
-
"camera_type": "wrist_left",
|
| 718 |
-
"lerobot_key": "observation.images.left_wrist_rgb",
|
| 719 |
-
"start_index": 0,
|
| 720 |
-
"raw_shape": [
|
| 721 |
-
3,
|
| 722 |
-
720,
|
| 723 |
-
1280
|
| 724 |
-
],
|
| 725 |
-
"shape": [
|
| 726 |
-
3,
|
| 727 |
-
224,
|
| 728 |
-
224
|
| 729 |
-
],
|
| 730 |
-
"time_offset": 0
|
| 731 |
-
},
|
| 732 |
-
{
|
| 733 |
-
"key": "right_wrist_rgb",
|
| 734 |
-
"camera_type": "wrist_right",
|
| 735 |
-
"lerobot_key": "observation.images.right_wrist_rgb",
|
| 736 |
-
"start_index": 0,
|
| 737 |
-
"raw_shape": [
|
| 738 |
-
3,
|
| 739 |
-
720,
|
| 740 |
-
1280
|
| 741 |
-
],
|
| 742 |
-
"shape": [
|
| 743 |
-
3,
|
| 744 |
-
224,
|
| 745 |
-
224
|
| 746 |
-
],
|
| 747 |
-
"time_offset": 0
|
| 748 |
-
}
|
| 749 |
-
]
|
| 750 |
-
},
|
| 751 |
-
"action_state_transforms": [
|
| 752 |
-
{
|
| 753 |
-
"_target_": "g05.data_processor.transforms.relative_action.RelativeJointTransform",
|
| 754 |
-
"keys": [
|
| 755 |
-
"left_arm",
|
| 756 |
-
"right_arm"
|
| 757 |
-
]
|
| 758 |
-
}
|
| 759 |
-
],
|
| 760 |
-
"train_transforms": {
|
| 761 |
-
"head_rgb": [
|
| 762 |
-
{
|
| 763 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 764 |
-
},
|
| 765 |
-
{
|
| 766 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 767 |
-
"mean": [
|
| 768 |
-
0.5,
|
| 769 |
-
0.5,
|
| 770 |
-
0.5
|
| 771 |
-
],
|
| 772 |
-
"std": [
|
| 773 |
-
0.5,
|
| 774 |
-
0.5,
|
| 775 |
-
0.5
|
| 776 |
-
]
|
| 777 |
-
}
|
| 778 |
-
],
|
| 779 |
-
"left_wrist_rgb": [
|
| 780 |
-
{
|
| 781 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 782 |
-
},
|
| 783 |
-
{
|
| 784 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 785 |
-
"mean": [
|
| 786 |
-
0.5,
|
| 787 |
-
0.5,
|
| 788 |
-
0.5
|
| 789 |
-
],
|
| 790 |
-
"std": [
|
| 791 |
-
0.5,
|
| 792 |
-
0.5,
|
| 793 |
-
0.5
|
| 794 |
-
]
|
| 795 |
-
}
|
| 796 |
-
],
|
| 797 |
-
"right_wrist_rgb": [
|
| 798 |
-
{
|
| 799 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 800 |
-
},
|
| 801 |
-
{
|
| 802 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 803 |
-
"mean": [
|
| 804 |
-
0.5,
|
| 805 |
-
0.5,
|
| 806 |
-
0.5
|
| 807 |
-
],
|
| 808 |
-
"std": [
|
| 809 |
-
0.5,
|
| 810 |
-
0.5,
|
| 811 |
-
0.5
|
| 812 |
-
]
|
| 813 |
-
}
|
| 814 |
-
]
|
| 815 |
-
},
|
| 816 |
-
"val_transforms": {
|
| 817 |
-
"head_rgb": [
|
| 818 |
-
{
|
| 819 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 820 |
-
},
|
| 821 |
-
{
|
| 822 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 823 |
-
"mean": [
|
| 824 |
-
0.5,
|
| 825 |
-
0.5,
|
| 826 |
-
0.5
|
| 827 |
-
],
|
| 828 |
-
"std": [
|
| 829 |
-
0.5,
|
| 830 |
-
0.5,
|
| 831 |
-
0.5
|
| 832 |
-
]
|
| 833 |
-
}
|
| 834 |
-
],
|
| 835 |
-
"left_wrist_rgb": [
|
| 836 |
-
{
|
| 837 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 838 |
-
},
|
| 839 |
-
{
|
| 840 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 841 |
-
"mean": [
|
| 842 |
-
0.5,
|
| 843 |
-
0.5,
|
| 844 |
-
0.5
|
| 845 |
-
],
|
| 846 |
-
"std": [
|
| 847 |
-
0.5,
|
| 848 |
-
0.5,
|
| 849 |
-
0.5
|
| 850 |
-
]
|
| 851 |
-
}
|
| 852 |
-
],
|
| 853 |
-
"right_wrist_rgb": [
|
| 854 |
-
{
|
| 855 |
-
"_target_": "g05.data_processor.transforms.image.ToTensor"
|
| 856 |
-
},
|
| 857 |
-
{
|
| 858 |
-
"_target_": "torchvision.transforms.Normalize",
|
| 859 |
-
"mean": [
|
| 860 |
-
0.5,
|
| 861 |
-
0.5,
|
| 862 |
-
0.5
|
| 863 |
-
],
|
| 864 |
-
"std": [
|
| 865 |
-
0.5,
|
| 866 |
-
0.5,
|
| 867 |
-
0.5
|
| 868 |
-
]
|
| 869 |
-
}
|
| 870 |
-
]
|
| 871 |
-
},
|
| 872 |
-
"norm_default_mode": "z-score-tail",
|
| 873 |
-
"norm_exception_mode": {
|
| 874 |
-
"action": {
|
| 875 |
-
"left_gripper": "q01/q99",
|
| 876 |
-
"right_gripper": "q01/q99"
|
| 877 |
-
},
|
| 878 |
-
"state": {
|
| 879 |
-
"left_gripper": "q01/q99",
|
| 880 |
-
"right_gripper": "q01/q99"
|
| 881 |
-
}
|
| 882 |
-
},
|
| 883 |
-
"action_filter": {
|
| 884 |
-
"_target_": "g05.data_processor.processor.galaxea_action_processor.R1LiteJointActionFilter",
|
| 885 |
-
"velocity_threshold": 0.1,
|
| 886 |
-
"dim_thresholds": {
|
| 887 |
-
"left_arm": [
|
| 888 |
-
0.00756,
|
| 889 |
-
0.00662,
|
| 890 |
-
0.0432,
|
| 891 |
-
0.00732,
|
| 892 |
-
0.00493,
|
| 893 |
-
0.00612
|
| 894 |
-
],
|
| 895 |
-
"right_arm": [
|
| 896 |
-
0.0115,
|
| 897 |
-
0.00717,
|
| 898 |
-
0.0619,
|
| 899 |
-
0.012,
|
| 900 |
-
0.00737,
|
| 901 |
-
0.00918
|
| 902 |
-
]
|
| 903 |
-
}
|
| 904 |
-
},
|
| 905 |
-
"action_state_merger": {
|
| 906 |
-
"_target_": "g05.data_processor.transforms.action_state_merger.GroupedPaddingMerger",
|
| 907 |
-
"merge": true,
|
| 908 |
-
"max_action_shape_meta": {
|
| 909 |
-
"left_arm": 7,
|
| 910 |
-
"right_arm": 7,
|
| 911 |
-
"left_gripper": 1,
|
| 912 |
-
"right_gripper": 1,
|
| 913 |
-
"left_ee_pose": 9,
|
| 914 |
-
"right_ee_pose": 9,
|
| 915 |
-
"lower_body": 7
|
| 916 |
-
},
|
| 917 |
-
"max_state_shape_meta": {
|
| 918 |
-
"left_arm": 7,
|
| 919 |
-
"right_arm": 7,
|
| 920 |
-
"left_gripper": 1,
|
| 921 |
-
"right_gripper": 1,
|
| 922 |
-
"left_ee_pose": 9,
|
| 923 |
-
"right_ee_pose": 9,
|
| 924 |
-
"lower_body": 7
|
| 925 |
-
},
|
| 926 |
-
"merge_spec": {
|
| 927 |
-
"left_control": [
|
| 928 |
-
"left_arm",
|
| 929 |
-
"left_ee_pose"
|
| 930 |
-
],
|
| 931 |
-
"left_gripper": [
|
| 932 |
-
"left_gripper"
|
| 933 |
-
],
|
| 934 |
-
"right_control": [
|
| 935 |
-
"right_arm",
|
| 936 |
-
"right_ee_pose"
|
| 937 |
-
],
|
| 938 |
-
"right_gripper": [
|
| 939 |
-
"right_gripper"
|
| 940 |
-
],
|
| 941 |
-
"lower_body": [
|
| 942 |
-
"lower_body"
|
| 943 |
-
]
|
| 944 |
-
}
|
| 945 |
-
},
|
| 946 |
-
"num_obs_steps": 6,
|
| 947 |
-
"use_stepwise_action_norm": true,
|
| 948 |
-
"drop_high_level_prob": 1.0,
|
| 949 |
-
"use_zh_instruction": false,
|
| 950 |
-
"_target_": "g05.data_processor.processor.galaxea_cot_processor.GalaxeaCoTProcessor",
|
| 951 |
-
"discrete_action": true,
|
| 952 |
-
"num_output_cameras": 18,
|
| 953 |
-
"pad_token_id": 0,
|
| 954 |
-
"image_token_index": 248056,
|
| 955 |
-
"tokenizer_params": {
|
| 956 |
-
"pretrained_model_name_or_path": null,
|
| 957 |
-
"local_files_only": true,
|
| 958 |
-
"token": null
|
| 959 |
-
},
|
| 960 |
-
"max_text_tokens": 160,
|
| 961 |
-
"num_input_cameras": 3,
|
| 962 |
-
"camera_size_config": {
|
| 963 |
-
"exterior": [
|
| 964 |
-
256,
|
| 965 |
-
256
|
| 966 |
-
],
|
| 967 |
-
"wrist_left": [
|
| 968 |
-
256,
|
| 969 |
-
256
|
| 970 |
-
],
|
| 971 |
-
"wrist_right": [
|
| 972 |
-
256,
|
| 973 |
-
256
|
| 974 |
-
]
|
| 975 |
-
},
|
| 976 |
-
"samples_builder": {
|
| 977 |
-
"_target_": "g05.data_processor.processor.samples_builder.MixedSamplesBuilder",
|
| 978 |
-
"_partial_": true,
|
| 979 |
-
"_recursive_": false,
|
| 980 |
-
"candidates": [
|
| 981 |
-
{
|
| 982 |
-
"_target_": "g05.data_processor.processor.samples_builder.HighLevelAtomicTaskCoTBuilder",
|
| 983 |
-
"weight": 1.0
|
| 984 |
-
},
|
| 985 |
-
{
|
| 986 |
-
"_target_": "g05.data_processor.processor.samples_builder.SubtaskActionHintCoTBuilder",
|
| 987 |
-
"weight": 1.0
|
| 988 |
-
},
|
| 989 |
-
{
|
| 990 |
-
"_target_": "g05.data_processor.processor.samples_builder.AtomicTaskBaseSamplesBuilder",
|
| 991 |
-
"weight": 1.0
|
| 992 |
-
},
|
| 993 |
-
{
|
| 994 |
-
"_target_": "g05.data_processor.processor.samples_builder.Trace2DCoTBuilder",
|
| 995 |
-
"weight": 1.0
|
| 996 |
-
},
|
| 997 |
-
{
|
| 998 |
-
"_target_": "g05.data_processor.processor.samples_builder.BBoxSubtaskCoTBuilder",
|
| 999 |
-
"weight": 1.0
|
| 1000 |
-
},
|
| 1001 |
-
{
|
| 1002 |
-
"_target_": "g05.data_processor.processor.samples_builder.BBoxCoTBuilder",
|
| 1003 |
-
"weight": 1.0
|
| 1004 |
-
},
|
| 1005 |
-
{
|
| 1006 |
-
"_target_": "g05.data_processor.processor.samples_builder.SubtaskCoTBuilder",
|
| 1007 |
-
"weight": 2.0
|
| 1008 |
-
},
|
| 1009 |
-
{
|
| 1010 |
-
"_target_": "g05.data_processor.processor.samples_builder.BaseSamplesBuilder",
|
| 1011 |
-
"weight": 1.0
|
| 1012 |
-
}
|
| 1013 |
-
],
|
| 1014 |
-
"eval_builder": {
|
| 1015 |
-
"_target_": "g05.data_processor.processor.samples_builder.BaseSamplesBuilder"
|
| 1016 |
-
}
|
| 1017 |
-
},
|
| 1018 |
-
"vlm_input_action_norm_default_mode": null,
|
| 1019 |
-
"vlm_input_action_norm_exception_mode": null
|
| 1020 |
-
},
|
| 1021 |
-
"cot_bbox_camera": null
|
| 1022 |
-
},
|
| 1023 |
"revision": null
|
| 1024 |
}
|
| 1025 |
}
|
|
|
|
| 251 |
"registry_name": "g05_tokenizer",
|
| 252 |
"config": {
|
| 253 |
"checkpoint_path": "lerobot/g05_base",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 254 |
"revision": null
|
| 255 |
}
|
| 256 |
}
|