Add files using upload-large-folder tool
Browse files- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json +431 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/optimizer.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/rng_state.pth +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/scheduler.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/training_args.bin +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json +431 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/optimizer.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/rng_state.pth +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/scheduler.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/training_args.bin +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json +431 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/optimizer.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/rng_state.pth +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/scheduler.pt +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/config.json +101 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00001-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00002-of-00002.safetensors +3 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model.safetensors.index.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/trainer_state.json +0 -0
- FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/training_args.bin +3 -0
- RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/base120_single_fm1-rkd0p5_atomic65_human.yaml +36 -0
- RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k/experiment_cfg/metadata.json +431 -0
- RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/resolved_config.yaml +37 -0
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.778976917266846,
|
| 8 |
+
0.33403685688972473,
|
| 9 |
+
0.7223060727119446
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.821934223175049,
|
| 13 |
+
-6.785110950469971,
|
| 14 |
+
0.6994727253913879
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.5886094736594543,
|
| 18 |
+
-2.1559664577316586,
|
| 19 |
+
0.7003580854548973
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6717149343420803,
|
| 23 |
+
1.3743695399555862,
|
| 24 |
+
0.000940069542359506
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-0.7517815060249613,
|
| 28 |
+
-5.497849509089705,
|
| 29 |
+
0.6999776914424374
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.060945690807028,
|
| 33 |
+
-0.5331499932493624,
|
| 34 |
+
0.7028864260472257
|
| 35 |
+
]
|
| 36 |
+
},
|
| 37 |
+
"base_rotation": {
|
| 38 |
+
"max": [
|
| 39 |
+
0.0,
|
| 40 |
+
0.0,
|
| 41 |
+
1.0,
|
| 42 |
+
1.0
|
| 43 |
+
],
|
| 44 |
+
"min": [
|
| 45 |
+
0.0,
|
| 46 |
+
0.0,
|
| 47 |
+
-1.0,
|
| 48 |
+
0.0
|
| 49 |
+
],
|
| 50 |
+
"mean": [
|
| 51 |
+
0.0,
|
| 52 |
+
0.0,
|
| 53 |
+
0.24488939171811902,
|
| 54 |
+
0.5968549204077488
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6659970734507764,
|
| 60 |
+
0.37443298392262264
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-0.9999999999999999,
|
| 66 |
+
4.149924358336069e-06
|
| 67 |
+
],
|
| 68 |
+
"q99": [
|
| 69 |
+
0.0,
|
| 70 |
+
0.0,
|
| 71 |
+
0.9999999999999999,
|
| 72 |
+
0.9999999999999999
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
"end_effector_position_relative": {
|
| 76 |
+
"max": [
|
| 77 |
+
0.8848381638526917,
|
| 78 |
+
0.8569459915161133,
|
| 79 |
+
1.053884744644165
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.5828549861907959,
|
| 83 |
+
-0.8486236333847046,
|
| 84 |
+
-0.3446122705936432
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.25087486209693133,
|
| 88 |
+
-0.005738920462036393,
|
| 89 |
+
0.43643948866361787
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.15741544532518012,
|
| 93 |
+
0.2547603106793105,
|
| 94 |
+
0.2050434492495558
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
-0.013990057288082561,
|
| 98 |
+
-0.4824599092623036,
|
| 99 |
+
0.18825220353500022
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5411989642841469,
|
| 103 |
+
0.4754923657245658,
|
| 104 |
+
0.6912323747882034
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999949932098389,
|
| 110 |
+
0.9997354745864868,
|
| 111 |
+
0.9900747537612915,
|
| 112 |
+
0.9735639095306396
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999914169311523,
|
| 116 |
+
-0.9999884366989136,
|
| 117 |
+
-0.9975369572639465,
|
| 118 |
+
6.754255821306288e-08
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.22290331318740714,
|
| 122 |
+
0.04579037037503981,
|
| 123 |
+
-0.05804094906527823,
|
| 124 |
+
0.12347219024917105
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.8519267931761344,
|
| 128 |
+
0.2949394732270141,
|
| 129 |
+
0.30026078832556913,
|
| 130 |
+
0.16333086331829588
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.9957959450418518,
|
| 134 |
+
-0.5678869493126506,
|
| 135 |
+
-0.5318135475761875,
|
| 136 |
+
0.0014160465686576505
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.9900692296898317,
|
| 140 |
+
0.6357358195673684,
|
| 141 |
+
0.5009847868015971,
|
| 142 |
+
0.4287044482210845
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.05289580300450325,
|
| 148 |
+
0.00937709677964449
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011669117026031017,
|
| 152 |
+
-0.06307756155729294
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.03203268930378034,
|
| 156 |
+
-0.032331639271968075
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.011607049173153697,
|
| 160 |
+
0.011306286745872529
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.006429109393656672,
|
| 164 |
+
-0.04056586943958638
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.040377860142451545,
|
| 168 |
+
-0.006593889698673258
|
| 169 |
+
]
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
"action": {
|
| 173 |
+
"base_motion": {
|
| 174 |
+
"max": [
|
| 175 |
+
1.0,
|
| 176 |
+
1.0,
|
| 177 |
+
1.0,
|
| 178 |
+
0.0
|
| 179 |
+
],
|
| 180 |
+
"min": [
|
| 181 |
+
-1.0,
|
| 182 |
+
-1.0,
|
| 183 |
+
-1.0,
|
| 184 |
+
0.0
|
| 185 |
+
],
|
| 186 |
+
"mean": [
|
| 187 |
+
0.006050336852760097,
|
| 188 |
+
0.002728916466628117,
|
| 189 |
+
-0.0017236811393598838,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11032167711449951,
|
| 194 |
+
0.09824140301615687,
|
| 195 |
+
0.07825208047515658,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.154721262094005,
|
| 200 |
+
-0.18955302988697728,
|
| 201 |
+
-0.14454065867754942,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.24492955247744197,
|
| 206 |
+
0.2496797822190429,
|
| 207 |
+
0.11966742325336034,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9294448715752568
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.36897903503358326
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.21627562513311385
|
| 229 |
+
]
|
| 230 |
+
},
|
| 231 |
+
"end_effector_position": {
|
| 232 |
+
"max": [
|
| 233 |
+
1.0,
|
| 234 |
+
1.0,
|
| 235 |
+
1.0
|
| 236 |
+
],
|
| 237 |
+
"min": [
|
| 238 |
+
-1.0,
|
| 239 |
+
-1.0,
|
| 240 |
+
-1.0
|
| 241 |
+
],
|
| 242 |
+
"mean": [
|
| 243 |
+
-0.00826780165926883,
|
| 244 |
+
0.003161997359379232,
|
| 245 |
+
-0.08157453651841007
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4545430944464769,
|
| 249 |
+
0.41040151434986594,
|
| 250 |
+
0.4102644251536544
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.9653185255999107,
|
| 254 |
+
-0.9255635199318754,
|
| 255 |
+
-0.8208437313341718
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7987781084592441,
|
| 259 |
+
0.9260844232330088,
|
| 260 |
+
0.8357421812622201
|
| 261 |
+
]
|
| 262 |
+
},
|
| 263 |
+
"end_effector_rotation": {
|
| 264 |
+
"max": [
|
| 265 |
+
1.0,
|
| 266 |
+
1.0,
|
| 267 |
+
1.0
|
| 268 |
+
],
|
| 269 |
+
"min": [
|
| 270 |
+
-1.0,
|
| 271 |
+
-1.0,
|
| 272 |
+
-1.0
|
| 273 |
+
],
|
| 274 |
+
"mean": [
|
| 275 |
+
0.007367024641295137,
|
| 276 |
+
-0.02060756343728361,
|
| 277 |
+
-0.001884833075998489
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.1116768956941076,
|
| 281 |
+
0.13249042272165212,
|
| 282 |
+
0.12338908364725322
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.26894658663246845,
|
| 286 |
+
-0.39865607740339326,
|
| 287 |
+
-0.34393735191302655
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.3191674780171587,
|
| 291 |
+
0.31133630397431267,
|
| 292 |
+
0.33968476423051097
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3098994894722959
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9507718287498876
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.7545620899936146
|
| 313 |
+
]
|
| 314 |
+
}
|
| 315 |
+
}
|
| 316 |
+
},
|
| 317 |
+
"modalities": {
|
| 318 |
+
"video": {
|
| 319 |
+
"robot0_eye_in_hand": {
|
| 320 |
+
"resolution": [
|
| 321 |
+
256,
|
| 322 |
+
256
|
| 323 |
+
],
|
| 324 |
+
"channels": 3,
|
| 325 |
+
"fps": 20.0
|
| 326 |
+
},
|
| 327 |
+
"robot0_agentview_left": {
|
| 328 |
+
"resolution": [
|
| 329 |
+
256,
|
| 330 |
+
256
|
| 331 |
+
],
|
| 332 |
+
"channels": 3,
|
| 333 |
+
"fps": 20.0
|
| 334 |
+
},
|
| 335 |
+
"robot0_agentview_right": {
|
| 336 |
+
"resolution": [
|
| 337 |
+
256,
|
| 338 |
+
256
|
| 339 |
+
],
|
| 340 |
+
"channels": 3,
|
| 341 |
+
"fps": 20.0
|
| 342 |
+
}
|
| 343 |
+
},
|
| 344 |
+
"state": {
|
| 345 |
+
"base_position": {
|
| 346 |
+
"absolute": true,
|
| 347 |
+
"rotation_type": null,
|
| 348 |
+
"shape": [
|
| 349 |
+
3
|
| 350 |
+
],
|
| 351 |
+
"continuous": true
|
| 352 |
+
},
|
| 353 |
+
"base_rotation": {
|
| 354 |
+
"absolute": true,
|
| 355 |
+
"rotation_type": "quaternion",
|
| 356 |
+
"shape": [
|
| 357 |
+
4
|
| 358 |
+
],
|
| 359 |
+
"continuous": true
|
| 360 |
+
},
|
| 361 |
+
"end_effector_position_relative": {
|
| 362 |
+
"absolute": true,
|
| 363 |
+
"rotation_type": null,
|
| 364 |
+
"shape": [
|
| 365 |
+
3
|
| 366 |
+
],
|
| 367 |
+
"continuous": true
|
| 368 |
+
},
|
| 369 |
+
"end_effector_rotation_relative": {
|
| 370 |
+
"absolute": true,
|
| 371 |
+
"rotation_type": "quaternion",
|
| 372 |
+
"shape": [
|
| 373 |
+
4
|
| 374 |
+
],
|
| 375 |
+
"continuous": true
|
| 376 |
+
},
|
| 377 |
+
"gripper_qpos": {
|
| 378 |
+
"absolute": true,
|
| 379 |
+
"rotation_type": null,
|
| 380 |
+
"shape": [
|
| 381 |
+
2
|
| 382 |
+
],
|
| 383 |
+
"continuous": true
|
| 384 |
+
}
|
| 385 |
+
},
|
| 386 |
+
"action": {
|
| 387 |
+
"base_motion": {
|
| 388 |
+
"absolute": true,
|
| 389 |
+
"rotation_type": null,
|
| 390 |
+
"shape": [
|
| 391 |
+
4
|
| 392 |
+
],
|
| 393 |
+
"continuous": true
|
| 394 |
+
},
|
| 395 |
+
"control_mode": {
|
| 396 |
+
"absolute": true,
|
| 397 |
+
"rotation_type": null,
|
| 398 |
+
"shape": [
|
| 399 |
+
1
|
| 400 |
+
],
|
| 401 |
+
"continuous": true
|
| 402 |
+
},
|
| 403 |
+
"end_effector_position": {
|
| 404 |
+
"absolute": true,
|
| 405 |
+
"rotation_type": null,
|
| 406 |
+
"shape": [
|
| 407 |
+
3
|
| 408 |
+
],
|
| 409 |
+
"continuous": true
|
| 410 |
+
},
|
| 411 |
+
"end_effector_rotation": {
|
| 412 |
+
"absolute": true,
|
| 413 |
+
"rotation_type": "axis_angle",
|
| 414 |
+
"shape": [
|
| 415 |
+
3
|
| 416 |
+
],
|
| 417 |
+
"continuous": true
|
| 418 |
+
},
|
| 419 |
+
"gripper_close": {
|
| 420 |
+
"absolute": true,
|
| 421 |
+
"rotation_type": null,
|
| 422 |
+
"shape": [
|
| 423 |
+
1
|
| 424 |
+
],
|
| 425 |
+
"continuous": true
|
| 426 |
+
}
|
| 427 |
+
}
|
| 428 |
+
},
|
| 429 |
+
"embodiment_tag": "new_embodiment"
|
| 430 |
+
}
|
| 431 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dcbecde0056a2bcabba205866d2569d78defe9a1c298b5a74bc2dd650788b7f8
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b335aecb764efe0190c9a584c730cc341c0b92034c62e6d34aaaca37abaa0435
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c70c414c4729531fb412e583f2788ff8f89e272adf909ccd9542309b788eb86
|
| 3 |
+
size 8563326263
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a8a05114ecb0bdd72a49b35902c9e4ae20e5a0a84102f27cff35c6a71fc9ed70
|
| 3 |
+
size 14645
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde
|
| 3 |
+
size 1465
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/checkpoint-5000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dcbecde0056a2bcabba205866d2569d78defe9a1c298b5a74bc2dd650788b7f8
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b335aecb764efe0190c9a584c730cc341c0b92034c62e6d34aaaca37abaa0435
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed0/default/train_5shot_seed0_fullactionhead_5k/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:87ff16bc6a1dabd0ec1cbda14c9c7677efea7066e6b379642c32d9370fc6a332
|
| 3 |
+
size 6033
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.778976917266846,
|
| 8 |
+
0.33403685688972473,
|
| 9 |
+
0.7223060727119446
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.821934223175049,
|
| 13 |
+
-6.785110950469971,
|
| 14 |
+
0.6994727253913879
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.5719731235192254,
|
| 18 |
+
-2.1489371002267235,
|
| 19 |
+
0.7003612741268546
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.669945884010784,
|
| 23 |
+
1.3739411381361726,
|
| 24 |
+
0.0009413282367754987
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-0.7944286339399775,
|
| 28 |
+
-5.486611518471242,
|
| 29 |
+
0.699977555971739
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.019886905263648,
|
| 33 |
+
-0.5229184210026855,
|
| 34 |
+
0.702930326282967
|
| 35 |
+
]
|
| 36 |
+
},
|
| 37 |
+
"base_rotation": {
|
| 38 |
+
"max": [
|
| 39 |
+
0.0,
|
| 40 |
+
0.0,
|
| 41 |
+
1.0,
|
| 42 |
+
1.0
|
| 43 |
+
],
|
| 44 |
+
"min": [
|
| 45 |
+
0.0,
|
| 46 |
+
0.0,
|
| 47 |
+
-1.0,
|
| 48 |
+
0.0
|
| 49 |
+
],
|
| 50 |
+
"mean": [
|
| 51 |
+
0.0,
|
| 52 |
+
0.0,
|
| 53 |
+
0.248543484335273,
|
| 54 |
+
0.5959317405459531
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6655550240074092,
|
| 60 |
+
0.374283995143559
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-1.0000000000000002,
|
| 66 |
+
4.503254968910557e-06
|
| 67 |
+
],
|
| 68 |
+
"q99": [
|
| 69 |
+
0.0,
|
| 70 |
+
0.0,
|
| 71 |
+
1.0000000000000002,
|
| 72 |
+
1.0000000000000002
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
"end_effector_position_relative": {
|
| 76 |
+
"max": [
|
| 77 |
+
0.8848381638526917,
|
| 78 |
+
0.8569459915161133,
|
| 79 |
+
1.053884744644165
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.5828549861907959,
|
| 83 |
+
-0.8486236333847046,
|
| 84 |
+
-0.3446122705936432
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.2465490874130862,
|
| 88 |
+
-0.004711658908481467,
|
| 89 |
+
0.43463835464312073
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.15798784670671595,
|
| 93 |
+
0.25624102400497073,
|
| 94 |
+
0.20836028727974915
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
-0.020752051134640608,
|
| 98 |
+
-0.4837933887347875,
|
| 99 |
+
0.1802942952289915
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5386784714844007,
|
| 103 |
+
0.4801424212024721,
|
| 104 |
+
0.695095732975766
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999949932098389,
|
| 110 |
+
0.9997354745864868,
|
| 111 |
+
0.9900747537612915,
|
| 112 |
+
0.9735639095306396
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999914169311523,
|
| 116 |
+
-0.9999884366989136,
|
| 117 |
+
-0.9975369572639465,
|
| 118 |
+
6.754255821306288e-08
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.22531859714056926,
|
| 122 |
+
0.046973702821026614,
|
| 123 |
+
-0.0568176967878771,
|
| 124 |
+
0.12364307042955697
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.8502144333612052,
|
| 128 |
+
0.29591250318147405,
|
| 129 |
+
0.30218053156837016,
|
| 130 |
+
0.1636031353412931
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.9959057387172566,
|
| 134 |
+
-0.5665384745573048,
|
| 135 |
+
-0.5329334620834977,
|
| 136 |
+
0.0014372279999397553
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.9904107951565839,
|
| 140 |
+
0.6359623612021396,
|
| 141 |
+
0.5026076022439377,
|
| 142 |
+
0.42955559234879914
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.05289580300450325,
|
| 148 |
+
0.00937709677964449
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011669117026031017,
|
| 152 |
+
-0.06307756155729294
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.03188757060429838,
|
| 156 |
+
-0.0321793333086053
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.01171052313737148,
|
| 160 |
+
0.011427336801203189
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.006462901319459925,
|
| 164 |
+
-0.040557140599419166
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04037619335388386,
|
| 168 |
+
-0.006523947835993347
|
| 169 |
+
]
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
"action": {
|
| 173 |
+
"base_motion": {
|
| 174 |
+
"max": [
|
| 175 |
+
1.0,
|
| 176 |
+
1.0,
|
| 177 |
+
1.0,
|
| 178 |
+
0.0
|
| 179 |
+
],
|
| 180 |
+
"min": [
|
| 181 |
+
-1.0,
|
| 182 |
+
-1.0,
|
| 183 |
+
-1.0,
|
| 184 |
+
0.0
|
| 185 |
+
],
|
| 186 |
+
"mean": [
|
| 187 |
+
0.006639007355857881,
|
| 188 |
+
0.0027330421795561874,
|
| 189 |
+
-0.0017774441970214892,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11841528114040736,
|
| 194 |
+
0.1041359083946544,
|
| 195 |
+
0.08346265521208351,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.17562359675270237,
|
| 200 |
+
-0.21706402321529558,
|
| 201 |
+
-0.16273392585723104,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.2572918701644887,
|
| 206 |
+
0.27022325575787115,
|
| 207 |
+
0.1369626751049278,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9205539608927732
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.39063657242534666
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.17711863866562935
|
| 229 |
+
]
|
| 230 |
+
},
|
| 231 |
+
"end_effector_position": {
|
| 232 |
+
"max": [
|
| 233 |
+
1.0,
|
| 234 |
+
1.0,
|
| 235 |
+
1.0
|
| 236 |
+
],
|
| 237 |
+
"min": [
|
| 238 |
+
-1.0,
|
| 239 |
+
-1.0,
|
| 240 |
+
-1.0
|
| 241 |
+
],
|
| 242 |
+
"mean": [
|
| 243 |
+
-0.00924803964621179,
|
| 244 |
+
0.004053841761390575,
|
| 245 |
+
-0.08385115578460103
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.45413072518223824,
|
| 249 |
+
0.40158612714634856,
|
| 250 |
+
0.41068634292849304
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.9669817145875343,
|
| 254 |
+
-0.9241307799783718,
|
| 255 |
+
-0.8264079503068825
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.8021511975722166,
|
| 259 |
+
0.9241346582748478,
|
| 260 |
+
0.8348116781502298
|
| 261 |
+
]
|
| 262 |
+
},
|
| 263 |
+
"end_effector_rotation": {
|
| 264 |
+
"max": [
|
| 265 |
+
1.0,
|
| 266 |
+
1.0,
|
| 267 |
+
1.0
|
| 268 |
+
],
|
| 269 |
+
"min": [
|
| 270 |
+
-1.0,
|
| 271 |
+
-1.0,
|
| 272 |
+
-1.0
|
| 273 |
+
],
|
| 274 |
+
"mean": [
|
| 275 |
+
0.007644157504353809,
|
| 276 |
+
-0.019962543093886335,
|
| 277 |
+
-0.002047110452602617
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11076012154791087,
|
| 281 |
+
0.13176735140285795,
|
| 282 |
+
0.12286528665895614
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2659105439401192,
|
| 286 |
+
-0.3961694033813533,
|
| 287 |
+
-0.3434870832468936
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.3145786959266369,
|
| 291 |
+
0.31309430361227547,
|
| 292 |
+
0.34107754401942986
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.29870612267533575
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9543487555375691
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.7610674802484976
|
| 313 |
+
]
|
| 314 |
+
}
|
| 315 |
+
}
|
| 316 |
+
},
|
| 317 |
+
"modalities": {
|
| 318 |
+
"video": {
|
| 319 |
+
"robot0_eye_in_hand": {
|
| 320 |
+
"resolution": [
|
| 321 |
+
256,
|
| 322 |
+
256
|
| 323 |
+
],
|
| 324 |
+
"channels": 3,
|
| 325 |
+
"fps": 20.0
|
| 326 |
+
},
|
| 327 |
+
"robot0_agentview_left": {
|
| 328 |
+
"resolution": [
|
| 329 |
+
256,
|
| 330 |
+
256
|
| 331 |
+
],
|
| 332 |
+
"channels": 3,
|
| 333 |
+
"fps": 20.0
|
| 334 |
+
},
|
| 335 |
+
"robot0_agentview_right": {
|
| 336 |
+
"resolution": [
|
| 337 |
+
256,
|
| 338 |
+
256
|
| 339 |
+
],
|
| 340 |
+
"channels": 3,
|
| 341 |
+
"fps": 20.0
|
| 342 |
+
}
|
| 343 |
+
},
|
| 344 |
+
"state": {
|
| 345 |
+
"base_position": {
|
| 346 |
+
"absolute": true,
|
| 347 |
+
"rotation_type": null,
|
| 348 |
+
"shape": [
|
| 349 |
+
3
|
| 350 |
+
],
|
| 351 |
+
"continuous": true
|
| 352 |
+
},
|
| 353 |
+
"base_rotation": {
|
| 354 |
+
"absolute": true,
|
| 355 |
+
"rotation_type": "quaternion",
|
| 356 |
+
"shape": [
|
| 357 |
+
4
|
| 358 |
+
],
|
| 359 |
+
"continuous": true
|
| 360 |
+
},
|
| 361 |
+
"end_effector_position_relative": {
|
| 362 |
+
"absolute": true,
|
| 363 |
+
"rotation_type": null,
|
| 364 |
+
"shape": [
|
| 365 |
+
3
|
| 366 |
+
],
|
| 367 |
+
"continuous": true
|
| 368 |
+
},
|
| 369 |
+
"end_effector_rotation_relative": {
|
| 370 |
+
"absolute": true,
|
| 371 |
+
"rotation_type": "quaternion",
|
| 372 |
+
"shape": [
|
| 373 |
+
4
|
| 374 |
+
],
|
| 375 |
+
"continuous": true
|
| 376 |
+
},
|
| 377 |
+
"gripper_qpos": {
|
| 378 |
+
"absolute": true,
|
| 379 |
+
"rotation_type": null,
|
| 380 |
+
"shape": [
|
| 381 |
+
2
|
| 382 |
+
],
|
| 383 |
+
"continuous": true
|
| 384 |
+
}
|
| 385 |
+
},
|
| 386 |
+
"action": {
|
| 387 |
+
"base_motion": {
|
| 388 |
+
"absolute": true,
|
| 389 |
+
"rotation_type": null,
|
| 390 |
+
"shape": [
|
| 391 |
+
4
|
| 392 |
+
],
|
| 393 |
+
"continuous": true
|
| 394 |
+
},
|
| 395 |
+
"control_mode": {
|
| 396 |
+
"absolute": true,
|
| 397 |
+
"rotation_type": null,
|
| 398 |
+
"shape": [
|
| 399 |
+
1
|
| 400 |
+
],
|
| 401 |
+
"continuous": true
|
| 402 |
+
},
|
| 403 |
+
"end_effector_position": {
|
| 404 |
+
"absolute": true,
|
| 405 |
+
"rotation_type": null,
|
| 406 |
+
"shape": [
|
| 407 |
+
3
|
| 408 |
+
],
|
| 409 |
+
"continuous": true
|
| 410 |
+
},
|
| 411 |
+
"end_effector_rotation": {
|
| 412 |
+
"absolute": true,
|
| 413 |
+
"rotation_type": "axis_angle",
|
| 414 |
+
"shape": [
|
| 415 |
+
3
|
| 416 |
+
],
|
| 417 |
+
"continuous": true
|
| 418 |
+
},
|
| 419 |
+
"gripper_close": {
|
| 420 |
+
"absolute": true,
|
| 421 |
+
"rotation_type": null,
|
| 422 |
+
"shape": [
|
| 423 |
+
1
|
| 424 |
+
],
|
| 425 |
+
"continuous": true
|
| 426 |
+
}
|
| 427 |
+
}
|
| 428 |
+
},
|
| 429 |
+
"embodiment_tag": "new_embodiment"
|
| 430 |
+
}
|
| 431 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3cb77271d1a9fce833f206b6913da1e083de735183e622ded352d7b12681fb1
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e1d911f2e4331b03e74d36773631a8e6ac5b184267257c9c86e233b51cf80f3
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:82811a21c277b2bc53ee264d227d38451f2c82a8ed18435eacd8a4ae5beade78
|
| 3 |
+
size 8563326263
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54f950b8bb9b0c32184ffeb79794242a6b8e05bbd5987343fab0e5c1c77ad754
|
| 3 |
+
size 14645
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde
|
| 3 |
+
size 1465
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/checkpoint-5000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3cb77271d1a9fce833f206b6913da1e083de735183e622ded352d7b12681fb1
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8e1d911f2e4331b03e74d36773631a8e6ac5b184267257c9c86e233b51cf80f3
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed1/default/train_5shot_seed1_fullactionhead_5k/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ca330dac47870f5c0b9e52222b90eb78c3ac30f4764dd89362b10daaeda47df
|
| 3 |
+
size 6033
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.778976917266846,
|
| 8 |
+
0.33403685688972473,
|
| 9 |
+
0.7223060727119446
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.821934223175049,
|
| 13 |
+
-6.785110950469971,
|
| 14 |
+
0.6994727253913879
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.577927505176127,
|
| 18 |
+
-2.1525618086425604,
|
| 19 |
+
0.7003541659101092
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6720365948940972,
|
| 23 |
+
1.3768414767482617,
|
| 24 |
+
0.0009102921174316669
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-0.7605057497562686,
|
| 28 |
+
-5.518671947805291,
|
| 29 |
+
0.6999764703640262
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.038879433964306,
|
| 33 |
+
-0.5168299981147954,
|
| 34 |
+
0.7029182461269848
|
| 35 |
+
]
|
| 36 |
+
},
|
| 37 |
+
"base_rotation": {
|
| 38 |
+
"max": [
|
| 39 |
+
0.0,
|
| 40 |
+
0.0,
|
| 41 |
+
1.0,
|
| 42 |
+
1.0
|
| 43 |
+
],
|
| 44 |
+
"min": [
|
| 45 |
+
0.0,
|
| 46 |
+
0.0,
|
| 47 |
+
-1.0,
|
| 48 |
+
0.0
|
| 49 |
+
],
|
| 50 |
+
"mean": [
|
| 51 |
+
0.0,
|
| 52 |
+
0.0,
|
| 53 |
+
0.24782371371751144,
|
| 54 |
+
0.5971093380149053
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6642559499983499,
|
| 60 |
+
0.3751891785184018
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-1.0,
|
| 66 |
+
4.194193576416142e-06
|
| 67 |
+
],
|
| 68 |
+
"q99": [
|
| 69 |
+
0.0,
|
| 70 |
+
0.0,
|
| 71 |
+
1.0,
|
| 72 |
+
1.0
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
"end_effector_position_relative": {
|
| 76 |
+
"max": [
|
| 77 |
+
0.8848381638526917,
|
| 78 |
+
0.8569459915161133,
|
| 79 |
+
1.053884744644165
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.5828549861907959,
|
| 83 |
+
-0.8486236333847046,
|
| 84 |
+
-0.3446122705936432
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.25006971579930243,
|
| 88 |
+
-0.005716176032514759,
|
| 89 |
+
0.4426027721228362
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.15838384751435522,
|
| 93 |
+
0.25141504139800513,
|
| 94 |
+
0.20646816234620438
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
-0.017003721077380903,
|
| 98 |
+
-0.47738320165276543,
|
| 99 |
+
0.1917563409753314
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5401029199137674,
|
| 103 |
+
0.4706646609268599,
|
| 104 |
+
0.6965586537529516
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999949932098389,
|
| 110 |
+
0.9997354745864868,
|
| 111 |
+
0.9900747537612915,
|
| 112 |
+
0.9735639095306396
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999914169311523,
|
| 116 |
+
-0.9999884366989136,
|
| 117 |
+
-0.9975369572639465,
|
| 118 |
+
6.754255821306288e-08
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.22531628862537245,
|
| 122 |
+
0.045669797538445966,
|
| 123 |
+
-0.06098055586474368,
|
| 124 |
+
0.12292005493282424
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.8506341437448487,
|
| 128 |
+
0.2880763358322842,
|
| 129 |
+
0.30816573781739026,
|
| 130 |
+
0.16366165154486362
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.9957510593732032,
|
| 134 |
+
-0.5584741016891311,
|
| 135 |
+
-0.5371647381159943,
|
| 136 |
+
0.0013784815838340492
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.9899375209218226,
|
| 140 |
+
0.6327878884078038,
|
| 141 |
+
0.5071539024531994,
|
| 142 |
+
0.43555975763914345
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.05289580300450325,
|
| 148 |
+
0.00937709677964449
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011669117026031017,
|
| 152 |
+
-0.06307756155729294
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.03196506153371015,
|
| 156 |
+
-0.03229791582989581
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.01171640503534049,
|
| 160 |
+
0.01140016755425199
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.006290779996034005,
|
| 164 |
+
-0.04059717954551848
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04038165838477291,
|
| 168 |
+
-0.006363076424932449
|
| 169 |
+
]
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
"action": {
|
| 173 |
+
"base_motion": {
|
| 174 |
+
"max": [
|
| 175 |
+
1.0,
|
| 176 |
+
1.0,
|
| 177 |
+
1.0,
|
| 178 |
+
0.0
|
| 179 |
+
],
|
| 180 |
+
"min": [
|
| 181 |
+
-1.0,
|
| 182 |
+
-1.0,
|
| 183 |
+
-1.0,
|
| 184 |
+
0.0
|
| 185 |
+
],
|
| 186 |
+
"mean": [
|
| 187 |
+
0.006112990930475949,
|
| 188 |
+
0.002504955933349057,
|
| 189 |
+
-0.001639409958610569,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11478077265783658,
|
| 194 |
+
0.1002662368589144,
|
| 195 |
+
0.0807258815139021,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.16969970921613756,
|
| 200 |
+
-0.19919537649977875,
|
| 201 |
+
-0.15012139866371593,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.2466625453958827,
|
| 206 |
+
0.25040842772679267,
|
| 207 |
+
0.12777701094549537,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9253067479599785
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.3792397014797737
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.1974093068624371
|
| 229 |
+
]
|
| 230 |
+
},
|
| 231 |
+
"end_effector_position": {
|
| 232 |
+
"max": [
|
| 233 |
+
1.0,
|
| 234 |
+
1.0,
|
| 235 |
+
1.0
|
| 236 |
+
],
|
| 237 |
+
"min": [
|
| 238 |
+
-1.0,
|
| 239 |
+
-1.0,
|
| 240 |
+
-1.0
|
| 241 |
+
],
|
| 242 |
+
"mean": [
|
| 243 |
+
-0.006754222051360652,
|
| 244 |
+
0.0027854853184013867,
|
| 245 |
+
-0.08087572130330477
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.45779905420789546,
|
| 249 |
+
0.40578975219522756,
|
| 250 |
+
0.4062694098132682
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.968295960290036,
|
| 254 |
+
-0.9240402337368848,
|
| 255 |
+
-0.8249359802202183
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.8046634915851651,
|
| 259 |
+
0.9212243025654695,
|
| 260 |
+
0.838433711437593
|
| 261 |
+
]
|
| 262 |
+
},
|
| 263 |
+
"end_effector_rotation": {
|
| 264 |
+
"max": [
|
| 265 |
+
1.0,
|
| 266 |
+
1.0,
|
| 267 |
+
1.0
|
| 268 |
+
],
|
| 269 |
+
"min": [
|
| 270 |
+
-1.0,
|
| 271 |
+
-1.0,
|
| 272 |
+
-1.0
|
| 273 |
+
],
|
| 274 |
+
"mean": [
|
| 275 |
+
0.008633749251857347,
|
| 276 |
+
-0.020406457524911718,
|
| 277 |
+
-0.0027639446362883006
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11393972872407104,
|
| 281 |
+
0.13300186268009506,
|
| 282 |
+
0.12192288302920105
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2668579353776187,
|
| 286 |
+
-0.40299251206771697,
|
| 287 |
+
-0.34712957892172586
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.32748281264383927,
|
| 291 |
+
0.3179028969292779,
|
| 292 |
+
0.33459405510046736
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3091635769453286
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9510122432876397
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.7643545615216344
|
| 313 |
+
]
|
| 314 |
+
}
|
| 315 |
+
}
|
| 316 |
+
},
|
| 317 |
+
"modalities": {
|
| 318 |
+
"video": {
|
| 319 |
+
"robot0_eye_in_hand": {
|
| 320 |
+
"resolution": [
|
| 321 |
+
256,
|
| 322 |
+
256
|
| 323 |
+
],
|
| 324 |
+
"channels": 3,
|
| 325 |
+
"fps": 20.0
|
| 326 |
+
},
|
| 327 |
+
"robot0_agentview_left": {
|
| 328 |
+
"resolution": [
|
| 329 |
+
256,
|
| 330 |
+
256
|
| 331 |
+
],
|
| 332 |
+
"channels": 3,
|
| 333 |
+
"fps": 20.0
|
| 334 |
+
},
|
| 335 |
+
"robot0_agentview_right": {
|
| 336 |
+
"resolution": [
|
| 337 |
+
256,
|
| 338 |
+
256
|
| 339 |
+
],
|
| 340 |
+
"channels": 3,
|
| 341 |
+
"fps": 20.0
|
| 342 |
+
}
|
| 343 |
+
},
|
| 344 |
+
"state": {
|
| 345 |
+
"base_position": {
|
| 346 |
+
"absolute": true,
|
| 347 |
+
"rotation_type": null,
|
| 348 |
+
"shape": [
|
| 349 |
+
3
|
| 350 |
+
],
|
| 351 |
+
"continuous": true
|
| 352 |
+
},
|
| 353 |
+
"base_rotation": {
|
| 354 |
+
"absolute": true,
|
| 355 |
+
"rotation_type": "quaternion",
|
| 356 |
+
"shape": [
|
| 357 |
+
4
|
| 358 |
+
],
|
| 359 |
+
"continuous": true
|
| 360 |
+
},
|
| 361 |
+
"end_effector_position_relative": {
|
| 362 |
+
"absolute": true,
|
| 363 |
+
"rotation_type": null,
|
| 364 |
+
"shape": [
|
| 365 |
+
3
|
| 366 |
+
],
|
| 367 |
+
"continuous": true
|
| 368 |
+
},
|
| 369 |
+
"end_effector_rotation_relative": {
|
| 370 |
+
"absolute": true,
|
| 371 |
+
"rotation_type": "quaternion",
|
| 372 |
+
"shape": [
|
| 373 |
+
4
|
| 374 |
+
],
|
| 375 |
+
"continuous": true
|
| 376 |
+
},
|
| 377 |
+
"gripper_qpos": {
|
| 378 |
+
"absolute": true,
|
| 379 |
+
"rotation_type": null,
|
| 380 |
+
"shape": [
|
| 381 |
+
2
|
| 382 |
+
],
|
| 383 |
+
"continuous": true
|
| 384 |
+
}
|
| 385 |
+
},
|
| 386 |
+
"action": {
|
| 387 |
+
"base_motion": {
|
| 388 |
+
"absolute": true,
|
| 389 |
+
"rotation_type": null,
|
| 390 |
+
"shape": [
|
| 391 |
+
4
|
| 392 |
+
],
|
| 393 |
+
"continuous": true
|
| 394 |
+
},
|
| 395 |
+
"control_mode": {
|
| 396 |
+
"absolute": true,
|
| 397 |
+
"rotation_type": null,
|
| 398 |
+
"shape": [
|
| 399 |
+
1
|
| 400 |
+
],
|
| 401 |
+
"continuous": true
|
| 402 |
+
},
|
| 403 |
+
"end_effector_position": {
|
| 404 |
+
"absolute": true,
|
| 405 |
+
"rotation_type": null,
|
| 406 |
+
"shape": [
|
| 407 |
+
3
|
| 408 |
+
],
|
| 409 |
+
"continuous": true
|
| 410 |
+
},
|
| 411 |
+
"end_effector_rotation": {
|
| 412 |
+
"absolute": true,
|
| 413 |
+
"rotation_type": "axis_angle",
|
| 414 |
+
"shape": [
|
| 415 |
+
3
|
| 416 |
+
],
|
| 417 |
+
"continuous": true
|
| 418 |
+
},
|
| 419 |
+
"gripper_close": {
|
| 420 |
+
"absolute": true,
|
| 421 |
+
"rotation_type": null,
|
| 422 |
+
"shape": [
|
| 423 |
+
1
|
| 424 |
+
],
|
| 425 |
+
"continuous": true
|
| 426 |
+
}
|
| 427 |
+
}
|
| 428 |
+
},
|
| 429 |
+
"embodiment_tag": "new_embodiment"
|
| 430 |
+
}
|
| 431 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77aaacc7ee85470538018578fecaded4f70641bd8f27d08ebee25a9a605e7a99
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e239dee05215eab5d4798f2b6cf01904a026fd99af2c3f053cabc4a4b3f41d74
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a19dbc17222c8452fc2f1b373a4cba22d0aea3a7022f8f94dcd55ff210982b33
|
| 3 |
+
size 8563326263
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/rng_state.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:52d8b7794775502c36c9f18edf928415953e0f0e15962e98a47a94f03bf16dba
|
| 3 |
+
size 14645
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39f93e93eb5653777630c590a20e98e9373f48f53da024592383b624fc2dcbde
|
| 3 |
+
size 1465
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/checkpoint-5000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": null,
|
| 87 |
+
"rkd_loss_weight_start": 0.1,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00001-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77aaacc7ee85470538018578fecaded4f70641bd8f27d08ebee25a9a605e7a99
|
| 3 |
+
size 4999367032
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model-00002-of-00002.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e239dee05215eab5d4798f2b6cf01904a026fd99af2c3f053cabc4a4b3f41d74
|
| 3 |
+
size 2664540348
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
FewShot/39tasks/5shot/FromRKD-Timewarp_FullactionHead/seed2/default/train_5shot_seed2_fullactionhead_5k/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d4b4312a6d79ef8fa73c97bc91f00d99c6408f7c452df91b898b0233d20c14c
|
| 3 |
+
size 6033
|
RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/base120_single_fm1-rkd0p5_atomic65_human.yaml
ADDED
|
@@ -0,0 +1,36 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD 1:0.5 from
|
| 2 |
+
GR00T base to 120k on Atomic65 human
|
| 3 |
+
policy_type: groot_RKDv3
|
| 4 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 5 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv3/base120_single_fm1-rkd0p5_atomic65_human
|
| 6 |
+
dataset:
|
| 7 |
+
dataset_soup: atomic_pretrain_human
|
| 8 |
+
training:
|
| 9 |
+
num_gpus: 2
|
| 10 |
+
batch_size: 64
|
| 11 |
+
seed: 42
|
| 12 |
+
model:
|
| 13 |
+
rkd_teacher_source: timewarp_action
|
| 14 |
+
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
| 15 |
+
rkd_relation_mode: flatten
|
| 16 |
+
rkd_loss_type: distance_angle
|
| 17 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 18 |
+
rkd_student_reconstruction_head_enabled: true
|
| 19 |
+
rkd_student_reconstruction_head_dim: 512
|
| 20 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 21 |
+
rkd_distance_loss_weight: 1.0
|
| 22 |
+
rkd_angle_loss_weight: 2.0
|
| 23 |
+
rkd_exclude_diagonal: true
|
| 24 |
+
phases:
|
| 25 |
+
- name: fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k
|
| 26 |
+
max_steps: 120000
|
| 27 |
+
save_steps: 60000
|
| 28 |
+
trainable:
|
| 29 |
+
tune_llm: false
|
| 30 |
+
tune_visual: false
|
| 31 |
+
tune_projector: true
|
| 32 |
+
tune_diffusion_model: true
|
| 33 |
+
losses:
|
| 34 |
+
rkd_enabled: true
|
| 35 |
+
rkd_fm_loss_weight: 1.0
|
| 36 |
+
rkd_loss_weight: 0.5
|
RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.778976917266846,
|
| 8 |
+
0.4876587688922882,
|
| 9 |
+
0.7223060727119446
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.94293737411499,
|
| 13 |
+
-6.890198230743408,
|
| 14 |
+
0.6994727253913879
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.645730005800653,
|
| 18 |
+
-2.110066447108802,
|
| 19 |
+
0.7004254938507055
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.674798582069234,
|
| 23 |
+
1.350392482318102,
|
| 24 |
+
0.001107956680671985
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-1.0132281893317567,
|
| 28 |
+
-5.366714195503661,
|
| 29 |
+
0.699973412653353
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.099414437993715,
|
| 33 |
+
-0.5498886012762193,
|
| 34 |
+
0.7032860957123939
|
| 35 |
+
]
|
| 36 |
+
},
|
| 37 |
+
"base_rotation": {
|
| 38 |
+
"max": [
|
| 39 |
+
0.0,
|
| 40 |
+
0.0,
|
| 41 |
+
1.0,
|
| 42 |
+
1.0
|
| 43 |
+
],
|
| 44 |
+
"min": [
|
| 45 |
+
0.0,
|
| 46 |
+
0.0,
|
| 47 |
+
-1.0,
|
| 48 |
+
0.0
|
| 49 |
+
],
|
| 50 |
+
"mean": [
|
| 51 |
+
0.0,
|
| 52 |
+
0.0,
|
| 53 |
+
0.24103797327895668,
|
| 54 |
+
0.6071393324239354
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6615962784298529,
|
| 60 |
+
0.36814045893518177
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-1.0,
|
| 66 |
+
3.3913522062507517e-06
|
| 67 |
+
],
|
| 68 |
+
"q99": [
|
| 69 |
+
0.0,
|
| 70 |
+
0.0,
|
| 71 |
+
1.0,
|
| 72 |
+
1.0
|
| 73 |
+
]
|
| 74 |
+
},
|
| 75 |
+
"end_effector_position_relative": {
|
| 76 |
+
"max": [
|
| 77 |
+
0.9014528393745422,
|
| 78 |
+
0.8569459915161133,
|
| 79 |
+
1.053884744644165
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.5828549861907959,
|
| 83 |
+
-0.8486236333847046,
|
| 84 |
+
-0.3446122705936432
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.2641355221895152,
|
| 88 |
+
-0.01768191244102862,
|
| 89 |
+
0.4458847037294584
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.16437879912713016,
|
| 93 |
+
0.24835923712456218,
|
| 94 |
+
0.2124190803465676
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
-0.004962867024380738,
|
| 98 |
+
-0.46862868595225443,
|
| 99 |
+
0.20316236863116172
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5495510296644974,
|
| 103 |
+
0.44201455499146414,
|
| 104 |
+
0.6984316542098391
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999998807907104,
|
| 110 |
+
0.9997354745864868,
|
| 111 |
+
0.9900747537612915,
|
| 112 |
+
0.9735639095306396
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999930262565613,
|
| 116 |
+
-0.9999884366989136,
|
| 117 |
+
-0.9975369572639465,
|
| 118 |
+
6.754255821306288e-08
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.23536101367097556,
|
| 122 |
+
0.0384247508727208,
|
| 123 |
+
-0.06854891687606551,
|
| 124 |
+
0.1400254132624415
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.8296216819146821,
|
| 128 |
+
0.30067208350163366,
|
| 129 |
+
0.33390768096919515,
|
| 130 |
+
0.16926885514344236
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.9950540525525874,
|
| 134 |
+
-0.5590028004550295,
|
| 135 |
+
-0.5635208777456633,
|
| 136 |
+
0.001747048046951923
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.9472459325196646,
|
| 140 |
+
0.6183763159959191,
|
| 141 |
+
0.5108060516185887,
|
| 142 |
+
0.46646525101830405
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.055869169533252716,
|
| 148 |
+
0.010916369967162609
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011669117026031017,
|
| 152 |
+
-0.06307756155729294
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.031849369408926076,
|
| 156 |
+
-0.03200748595341724
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.012248040928874927,
|
| 160 |
+
0.01206276030599065
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.006329614505051435,
|
| 164 |
+
-0.0405872199089872
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.040445365520021676,
|
| 168 |
+
-0.006282430560590391
|
| 169 |
+
]
|
| 170 |
+
}
|
| 171 |
+
},
|
| 172 |
+
"action": {
|
| 173 |
+
"base_motion": {
|
| 174 |
+
"max": [
|
| 175 |
+
1.0,
|
| 176 |
+
1.0,
|
| 177 |
+
1.0,
|
| 178 |
+
0.0
|
| 179 |
+
],
|
| 180 |
+
"min": [
|
| 181 |
+
-1.0,
|
| 182 |
+
-1.0,
|
| 183 |
+
-1.0,
|
| 184 |
+
0.0
|
| 185 |
+
],
|
| 186 |
+
"mean": [
|
| 187 |
+
0.0071452133647496785,
|
| 188 |
+
0.0015603604277540566,
|
| 189 |
+
-0.0014052511062474704,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11471999232420287,
|
| 194 |
+
0.10209228015027566,
|
| 195 |
+
0.0855718789814838,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.13329522793794643,
|
| 200 |
+
-0.16270717101832055,
|
| 201 |
+
-0.12619798329427676,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.1721366453818688,
|
| 206 |
+
0.1972873452547982,
|
| 207 |
+
0.11128571706089413,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9218734936549906
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.3874975703570591
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.26692269247002653
|
| 229 |
+
]
|
| 230 |
+
},
|
| 231 |
+
"end_effector_position": {
|
| 232 |
+
"max": [
|
| 233 |
+
1.0,
|
| 234 |
+
1.0,
|
| 235 |
+
1.0
|
| 236 |
+
],
|
| 237 |
+
"min": [
|
| 238 |
+
-1.0,
|
| 239 |
+
-1.0,
|
| 240 |
+
-1.0
|
| 241 |
+
],
|
| 242 |
+
"mean": [
|
| 243 |
+
0.002920238070268553,
|
| 244 |
+
-0.0060913649045883315,
|
| 245 |
+
-0.07322447393298565
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4482114094812734,
|
| 249 |
+
0.41233706159751987,
|
| 250 |
+
0.4006084454909327
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.934104737554819,
|
| 254 |
+
-0.9255792016234818,
|
| 255 |
+
-0.8085822192859831
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7874185230421324,
|
| 259 |
+
0.9004049161564378,
|
| 260 |
+
0.8018332495119154
|
| 261 |
+
]
|
| 262 |
+
},
|
| 263 |
+
"end_effector_rotation": {
|
| 264 |
+
"max": [
|
| 265 |
+
1.0,
|
| 266 |
+
1.0,
|
| 267 |
+
1.0
|
| 268 |
+
],
|
| 269 |
+
"min": [
|
| 270 |
+
-1.0,
|
| 271 |
+
-1.0,
|
| 272 |
+
-1.0
|
| 273 |
+
],
|
| 274 |
+
"mean": [
|
| 275 |
+
0.0077772595328458915,
|
| 276 |
+
-0.022620543966259083,
|
| 277 |
+
-0.00029332176634943376
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11155576533045976,
|
| 281 |
+
0.13202104364511658,
|
| 282 |
+
0.1240098754411023
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2726786325766535,
|
| 286 |
+
-0.4125887853047325,
|
| 287 |
+
-0.3445183934868137
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.31803481867756295,
|
| 291 |
+
0.3015153519871841,
|
| 292 |
+
0.34730986869917346
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.33422761717726496
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9425001619711008
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.681821585371573
|
| 313 |
+
]
|
| 314 |
+
}
|
| 315 |
+
}
|
| 316 |
+
},
|
| 317 |
+
"modalities": {
|
| 318 |
+
"video": {
|
| 319 |
+
"robot0_eye_in_hand": {
|
| 320 |
+
"resolution": [
|
| 321 |
+
256,
|
| 322 |
+
256
|
| 323 |
+
],
|
| 324 |
+
"channels": 3,
|
| 325 |
+
"fps": 20.0
|
| 326 |
+
},
|
| 327 |
+
"robot0_agentview_left": {
|
| 328 |
+
"resolution": [
|
| 329 |
+
256,
|
| 330 |
+
256
|
| 331 |
+
],
|
| 332 |
+
"channels": 3,
|
| 333 |
+
"fps": 20.0
|
| 334 |
+
},
|
| 335 |
+
"robot0_agentview_right": {
|
| 336 |
+
"resolution": [
|
| 337 |
+
256,
|
| 338 |
+
256
|
| 339 |
+
],
|
| 340 |
+
"channels": 3,
|
| 341 |
+
"fps": 20.0
|
| 342 |
+
}
|
| 343 |
+
},
|
| 344 |
+
"state": {
|
| 345 |
+
"base_position": {
|
| 346 |
+
"absolute": true,
|
| 347 |
+
"rotation_type": null,
|
| 348 |
+
"shape": [
|
| 349 |
+
3
|
| 350 |
+
],
|
| 351 |
+
"continuous": true
|
| 352 |
+
},
|
| 353 |
+
"base_rotation": {
|
| 354 |
+
"absolute": true,
|
| 355 |
+
"rotation_type": "quaternion",
|
| 356 |
+
"shape": [
|
| 357 |
+
4
|
| 358 |
+
],
|
| 359 |
+
"continuous": true
|
| 360 |
+
},
|
| 361 |
+
"end_effector_position_relative": {
|
| 362 |
+
"absolute": true,
|
| 363 |
+
"rotation_type": null,
|
| 364 |
+
"shape": [
|
| 365 |
+
3
|
| 366 |
+
],
|
| 367 |
+
"continuous": true
|
| 368 |
+
},
|
| 369 |
+
"end_effector_rotation_relative": {
|
| 370 |
+
"absolute": true,
|
| 371 |
+
"rotation_type": "quaternion",
|
| 372 |
+
"shape": [
|
| 373 |
+
4
|
| 374 |
+
],
|
| 375 |
+
"continuous": true
|
| 376 |
+
},
|
| 377 |
+
"gripper_qpos": {
|
| 378 |
+
"absolute": true,
|
| 379 |
+
"rotation_type": null,
|
| 380 |
+
"shape": [
|
| 381 |
+
2
|
| 382 |
+
],
|
| 383 |
+
"continuous": true
|
| 384 |
+
}
|
| 385 |
+
},
|
| 386 |
+
"action": {
|
| 387 |
+
"base_motion": {
|
| 388 |
+
"absolute": true,
|
| 389 |
+
"rotation_type": null,
|
| 390 |
+
"shape": [
|
| 391 |
+
4
|
| 392 |
+
],
|
| 393 |
+
"continuous": true
|
| 394 |
+
},
|
| 395 |
+
"control_mode": {
|
| 396 |
+
"absolute": true,
|
| 397 |
+
"rotation_type": null,
|
| 398 |
+
"shape": [
|
| 399 |
+
1
|
| 400 |
+
],
|
| 401 |
+
"continuous": true
|
| 402 |
+
},
|
| 403 |
+
"end_effector_position": {
|
| 404 |
+
"absolute": true,
|
| 405 |
+
"rotation_type": null,
|
| 406 |
+
"shape": [
|
| 407 |
+
3
|
| 408 |
+
],
|
| 409 |
+
"continuous": true
|
| 410 |
+
},
|
| 411 |
+
"end_effector_rotation": {
|
| 412 |
+
"absolute": true,
|
| 413 |
+
"rotation_type": "axis_angle",
|
| 414 |
+
"shape": [
|
| 415 |
+
3
|
| 416 |
+
],
|
| 417 |
+
"continuous": true
|
| 418 |
+
},
|
| 419 |
+
"gripper_close": {
|
| 420 |
+
"absolute": true,
|
| 421 |
+
"rotation_type": null,
|
| 422 |
+
"shape": [
|
| 423 |
+
1
|
| 424 |
+
],
|
| 425 |
+
"continuous": true
|
| 426 |
+
}
|
| 427 |
+
}
|
| 428 |
+
},
|
| 429 |
+
"embodiment_tag": "new_embodiment"
|
| 430 |
+
}
|
| 431 |
+
}
|
RKDv3/base120_single_fm1-rkd0p5_atomic65_human/default/resolved_config.yaml
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD 1:0.5 from
|
| 2 |
+
GR00T base to 120k on Atomic65 human
|
| 3 |
+
policy_type: groot_RKDv3
|
| 4 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 5 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv3/base120_single_fm1-rkd0p5_atomic65_human
|
| 6 |
+
dataset:
|
| 7 |
+
dataset_soup: atomic_pretrain_human
|
| 8 |
+
training:
|
| 9 |
+
num_gpus: 2
|
| 10 |
+
batch_size: 64
|
| 11 |
+
seed: 42
|
| 12 |
+
model:
|
| 13 |
+
rkd_teacher_source: timewarp_action
|
| 14 |
+
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
| 15 |
+
rkd_relation_mode: flatten
|
| 16 |
+
rkd_loss_type: distance_angle
|
| 17 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 18 |
+
rkd_student_reconstruction_head_enabled: true
|
| 19 |
+
rkd_student_reconstruction_head_dim: 512
|
| 20 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 21 |
+
rkd_distance_loss_weight: 1.0
|
| 22 |
+
rkd_angle_loss_weight: 2.0
|
| 23 |
+
rkd_exclude_diagonal: true
|
| 24 |
+
phases:
|
| 25 |
+
- name: fm_rkdv3_timewarp_vae_da_aphead_fixed_0p5_120k
|
| 26 |
+
max_steps: 120000
|
| 27 |
+
save_steps: 60000
|
| 28 |
+
trainable:
|
| 29 |
+
tune_llm: false
|
| 30 |
+
tune_visual: false
|
| 31 |
+
tune_projector: true
|
| 32 |
+
tune_diffusion_model: true
|
| 33 |
+
losses:
|
| 34 |
+
rkd_enabled: true
|
| 35 |
+
rkd_fm_loss_weight: 1.0
|
| 36 |
+
rkd_loss_weight: 0.5
|
| 37 |
+
resolved_sweep: {}
|