Add files using upload-large-folder tool
Browse files- .gitattributes +1 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/APHead_AE_Flatten_Atomic65 KL.yaml +62 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase2/config.json +100 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase2/experiment_cfg/metadata.json +431 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase2/model.safetensors.index.json +0 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase2/trainer_state.json +0 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase2/training_args.bin +3 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/phase3/experiment_cfg/metadata.json +431 -0
- Atomic65/mgrkd_kl_ae_t0p1/default/resolved_config.yaml +63 -0
- Atomic65/mgrkd_kl_ae_t0p1/train_gpu01_20260626_135421.log +3 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/RowWiseAdaptiveProbMix.yaml +55 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/config.json +103 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/experiment_cfg/metadata.json +431 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/model.safetensors.index.json +0 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/trainer_state.json +0 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/training_args.bin +3 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/config.json +103 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/experiment_cfg/metadata.json +431 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/model.safetensors.index.json +0 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/trainer_state.json +0 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/training_args.bin +3 -0
- DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/resolved_config.yaml +56 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/TimeWeightedRowWiseDTW.yaml +55 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/config.json +103 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/experiment_cfg/metadata.json +431 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/model.safetensors.index.json +0 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/trainer_state.json +0 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/training_args.bin +3 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/config.json +103 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/experiment_cfg/metadata.json +431 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/model.safetensors.index.json +0 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/trainer_state.json +0 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/training_args.bin +3 -0
- DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/resolved_config.yaml +56 -0
- RKDv2.1/DA_MeanPool+LP/default/RKD v2.1 DA MF phase3 only.yaml +37 -0
- RKDv2.1/DA_MeanPool+LP/default/phase3/config.json +100 -0
- RKDv2.1/DA_MeanPool+LP/default/phase3/model.safetensors.index.json +0 -0
- RKDv2.1/DA_MeanPool+LP/default/phase3/trainer_state.json +0 -0
- RKDv2.1/DA_MeanPool+LP/default/phase3/training_args.bin +3 -0
- RKDv2.1/DA_MeanPool+LP/default/resolved_config.yaml +4 -29
- RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2.logpath +1 -0
- RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2.pid +1 -0
- RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2_20260626_103126.log +0 -0
.gitattributes
CHANGED
|
@@ -46,3 +46,4 @@ rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/defa
|
|
| 46 |
rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 47 |
RKDv2.1/rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 48 |
RKDv2.1/rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 46 |
rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 47 |
RKDv2.1/rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 48 |
RKDv2.1/rkd_v2.1[[:space:]]MoreSteps/rkd_v2_1_distance_angle_full_120k_cosine_decay/default/full_fm_rkd_da_cosine_decay_120k/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 49 |
+
Atomic65/mgrkd_kl_ae_t0p1/train_gpu01_20260626_135421.log filter=lfs diff=lfs merge=lfs -text
|
Atomic65/mgrkd_kl_ae_t0p1/default/APHead_AE_Flatten_Atomic65 KL.yaml
ADDED
|
@@ -0,0 +1,62 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: atomic65_mgrkd_kl_ae_t0p1
|
| 2 |
+
policy_type: groot_MGRKD
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/Atomic65_baseline/checkpoint-60000
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/Atomic65/mgrkd_kl_ae_t0p1
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: atomic_pretrain_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 13 |
+
phases:
|
| 14 |
+
- name: phase2_kl
|
| 15 |
+
max_steps: 30000
|
| 16 |
+
save_steps: 0
|
| 17 |
+
trainable:
|
| 18 |
+
preset: processing_line_only
|
| 19 |
+
tune_llm: false
|
| 20 |
+
tune_visual: false
|
| 21 |
+
tune_projector: false
|
| 22 |
+
tune_diffusion_model: false
|
| 23 |
+
losses:
|
| 24 |
+
rkd_enabled: true
|
| 25 |
+
rkd_fm_loss_weight: 0.0
|
| 26 |
+
rkd_loss_weight: 1.0
|
| 27 |
+
rkd_relation_mode: flatten
|
| 28 |
+
rkd_loss_type: kl
|
| 29 |
+
rkd_teacher_source: action_encoder
|
| 30 |
+
rkd_action_encoder_projector_enabled: true
|
| 31 |
+
rkd_action_encoder_projector_dim: 512
|
| 32 |
+
rkd_action_encoder_projector_pooling: flatten
|
| 33 |
+
rkd_student_reconstruction_head_enabled: true
|
| 34 |
+
rkd_student_reconstruction_head_dim: 512
|
| 35 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 36 |
+
rkd_distance_loss_weight: 1.0
|
| 37 |
+
rkd_angle_loss_weight: 2.0
|
| 38 |
+
rkd_exclude_diagonal: true
|
| 39 |
+
- name: phase3_fm_kl_0p5
|
| 40 |
+
max_steps: 30000
|
| 41 |
+
save_steps: 0
|
| 42 |
+
trainable:
|
| 43 |
+
tune_llm: false
|
| 44 |
+
tune_visual: false
|
| 45 |
+
tune_projector: true
|
| 46 |
+
tune_diffusion_model: true
|
| 47 |
+
losses:
|
| 48 |
+
rkd_enabled: true
|
| 49 |
+
rkd_fm_loss_weight: 1.0
|
| 50 |
+
rkd_loss_weight: 0.5
|
| 51 |
+
rkd_relation_mode: flatten
|
| 52 |
+
rkd_loss_type: kl
|
| 53 |
+
rkd_teacher_source: action_encoder
|
| 54 |
+
rkd_action_encoder_projector_enabled: true
|
| 55 |
+
rkd_action_encoder_projector_dim: 512
|
| 56 |
+
rkd_action_encoder_projector_pooling: flatten
|
| 57 |
+
rkd_student_reconstruction_head_enabled: true
|
| 58 |
+
rkd_student_reconstruction_head_dim: 512
|
| 59 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 60 |
+
rkd_distance_loss_weight: 1.0
|
| 61 |
+
rkd_angle_loss_weight: 2.0
|
| 62 |
+
rkd_exclude_diagonal: true
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase2/config.json
ADDED
|
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": true,
|
| 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": 0.0,
|
| 83 |
+
"rkd_loss_type": "kl",
|
| 84 |
+
"rkd_loss_weight": 1.0,
|
| 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": "flatten",
|
| 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": "action_encoder",
|
| 97 |
+
"rkd_vlm_temp": 0.1,
|
| 98 |
+
"torch_dtype": "bfloat16",
|
| 99 |
+
"transformers_version": "4.51.3"
|
| 100 |
+
}
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase2/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 |
+
}
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase2/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase2/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase2/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95c9dab7bef24312aea77317c134a8b06bed57c96b92cf016b2c7557c7ee7fab
|
| 3 |
+
size 5905
|
Atomic65/mgrkd_kl_ae_t0p1/default/phase3/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 |
+
}
|
Atomic65/mgrkd_kl_ae_t0p1/default/resolved_config.yaml
ADDED
|
@@ -0,0 +1,63 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: atomic65_mgrkd_kl_ae_t0p1
|
| 2 |
+
policy_type: groot_MGRKD
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/Atomic65_baseline/checkpoint-60000
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/Atomic65/mgrkd_kl_ae_t0p1
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: atomic_pretrain_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 13 |
+
phases:
|
| 14 |
+
- name: phase2_kl
|
| 15 |
+
max_steps: 30000
|
| 16 |
+
save_steps: 0
|
| 17 |
+
trainable:
|
| 18 |
+
preset: processing_line_only
|
| 19 |
+
tune_llm: false
|
| 20 |
+
tune_visual: false
|
| 21 |
+
tune_projector: false
|
| 22 |
+
tune_diffusion_model: false
|
| 23 |
+
losses:
|
| 24 |
+
rkd_enabled: true
|
| 25 |
+
rkd_fm_loss_weight: 0.0
|
| 26 |
+
rkd_loss_weight: 1.0
|
| 27 |
+
rkd_relation_mode: flatten
|
| 28 |
+
rkd_loss_type: kl
|
| 29 |
+
rkd_teacher_source: action_encoder
|
| 30 |
+
rkd_action_encoder_projector_enabled: true
|
| 31 |
+
rkd_action_encoder_projector_dim: 512
|
| 32 |
+
rkd_action_encoder_projector_pooling: flatten
|
| 33 |
+
rkd_student_reconstruction_head_enabled: true
|
| 34 |
+
rkd_student_reconstruction_head_dim: 512
|
| 35 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 36 |
+
rkd_distance_loss_weight: 1.0
|
| 37 |
+
rkd_angle_loss_weight: 2.0
|
| 38 |
+
rkd_exclude_diagonal: true
|
| 39 |
+
- name: phase3_fm_kl_0p5
|
| 40 |
+
max_steps: 30000
|
| 41 |
+
save_steps: 0
|
| 42 |
+
trainable:
|
| 43 |
+
tune_llm: false
|
| 44 |
+
tune_visual: false
|
| 45 |
+
tune_projector: true
|
| 46 |
+
tune_diffusion_model: true
|
| 47 |
+
losses:
|
| 48 |
+
rkd_enabled: true
|
| 49 |
+
rkd_fm_loss_weight: 1.0
|
| 50 |
+
rkd_loss_weight: 0.5
|
| 51 |
+
rkd_relation_mode: flatten
|
| 52 |
+
rkd_loss_type: kl
|
| 53 |
+
rkd_teacher_source: action_encoder
|
| 54 |
+
rkd_action_encoder_projector_enabled: true
|
| 55 |
+
rkd_action_encoder_projector_dim: 512
|
| 56 |
+
rkd_action_encoder_projector_pooling: flatten
|
| 57 |
+
rkd_student_reconstruction_head_enabled: true
|
| 58 |
+
rkd_student_reconstruction_head_dim: 512
|
| 59 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 60 |
+
rkd_distance_loss_weight: 1.0
|
| 61 |
+
rkd_angle_loss_weight: 2.0
|
| 62 |
+
rkd_exclude_diagonal: true
|
| 63 |
+
resolved_sweep: {}
|
Atomic65/mgrkd_kl_ae_t0p1/train_gpu01_20260626_135421.log
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:647908b220f773c0208676e8942c4092b33e1e1c8f7dad4cef1d4946a9f2425c
|
| 3 |
+
size 12982974
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/RowWiseAdaptiveProbMix.yaml
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: DTW from 60k - RowWiseAdaptiveProbMix
|
| 2 |
+
policy_type: groot_DTW
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: my_atomic26_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
dtw_enabled: true
|
| 13 |
+
dtw_relation_mode: flatten
|
| 14 |
+
dtw_loss_type: kl
|
| 15 |
+
dtw_action_temp: 0.1
|
| 16 |
+
dtw_student_temp: 0.1
|
| 17 |
+
dtw_student_reconstruction_head_enabled: true
|
| 18 |
+
dtw_student_reconstruction_head_dim: 512
|
| 19 |
+
dtw_student_reconstruction_head_hidden_dim: 512
|
| 20 |
+
dtw_student_reconstruction_head_pooling: attention
|
| 21 |
+
dtw_mix_mode: rowwise_adaptive_prob_mix
|
| 22 |
+
dtw_pos_weight: 1.0
|
| 23 |
+
dtw_rot_weight: 1.0
|
| 24 |
+
dtw_gripper_weight: 1.0
|
| 25 |
+
dtw_base_weight: 1.0
|
| 26 |
+
dtw_control_weight: 1.0
|
| 27 |
+
dtw_normalize_components: true
|
| 28 |
+
dtw_adaptive_mix_lambda: 1.0
|
| 29 |
+
dtw_adaptive_floor: 0.01
|
| 30 |
+
dtw_timing_log_interval: 1000
|
| 31 |
+
dtw_exclude_diagonal: true
|
| 32 |
+
phases:
|
| 33 |
+
- name: phase2_kl_dtw_aphead_rowwise_adaptive_only
|
| 34 |
+
max_steps: 30000
|
| 35 |
+
save_steps: 0
|
| 36 |
+
trainable:
|
| 37 |
+
preset: processing_line_only
|
| 38 |
+
tune_llm: false
|
| 39 |
+
tune_visual: false
|
| 40 |
+
tune_projector: false
|
| 41 |
+
tune_diffusion_model: false
|
| 42 |
+
losses:
|
| 43 |
+
dtw_fm_loss_weight: 0.0
|
| 44 |
+
dtw_kl_loss_weight: 1.0
|
| 45 |
+
- name: phase3_fm_kl_dtw_aphead_rowwise_adaptive_w0p5_f0p01
|
| 46 |
+
max_steps: 30000
|
| 47 |
+
save_steps: 0
|
| 48 |
+
trainable:
|
| 49 |
+
tune_llm: false
|
| 50 |
+
tune_visual: false
|
| 51 |
+
tune_projector: true
|
| 52 |
+
tune_diffusion_model: true
|
| 53 |
+
losses:
|
| 54 |
+
dtw_fm_loss_weight: 1.0
|
| 55 |
+
dtw_kl_loss_weight: 0.5
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/config.json
ADDED
|
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": 1.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.01,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 1.0,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_control_weight": 1.0,
|
| 75 |
+
"rkd_dtw_event_window_radius": 2,
|
| 76 |
+
"rkd_dtw_event_window_side": "both",
|
| 77 |
+
"rkd_dtw_gripper_weight": 1.0,
|
| 78 |
+
"rkd_dtw_mix_mode": "rowwise_adaptive_prob_mix",
|
| 79 |
+
"rkd_dtw_normalize_components": true,
|
| 80 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 81 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 82 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 83 |
+
"rkd_enabled": true,
|
| 84 |
+
"rkd_exclude_diagonal": true,
|
| 85 |
+
"rkd_fm_loss_weight": 0.0,
|
| 86 |
+
"rkd_loss_type": "kl",
|
| 87 |
+
"rkd_loss_weight": 1.0,
|
| 88 |
+
"rkd_loss_weight_end": 0.0,
|
| 89 |
+
"rkd_loss_weight_schedule": null,
|
| 90 |
+
"rkd_loss_weight_start": 0.1,
|
| 91 |
+
"rkd_raw_action_projector_dim": 512,
|
| 92 |
+
"rkd_raw_action_projector_enabled": false,
|
| 93 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 94 |
+
"rkd_relation_mode": "flatten",
|
| 95 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 96 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 97 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 98 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 99 |
+
"rkd_teacher_source": "dtw_action",
|
| 100 |
+
"rkd_vlm_temp": 0.1,
|
| 101 |
+
"torch_dtype": "bfloat16",
|
| 102 |
+
"transformers_version": "4.51.3"
|
| 103 |
+
}
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.3139495849609375,
|
| 8 |
+
0.4876587688922882,
|
| 9 |
+
0.7196521759033203
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.94293737411499,
|
| 13 |
+
-6.890198230743408,
|
| 14 |
+
0.6996827125549316
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.749288365724937,
|
| 18 |
+
-2.0455216239909983,
|
| 19 |
+
0.700532551020533
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6786295897046262,
|
| 23 |
+
1.312897969434244,
|
| 24 |
+
0.00133751758906258
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-1.4241907881148037,
|
| 28 |
+
-5.1716547935950885,
|
| 29 |
+
0.6999670898768614
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.198111545831555,
|
| 33 |
+
-0.5873531334104489,
|
| 34 |
+
0.7038833014242587
|
| 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 |
+
0.23084999587450247,
|
| 54 |
+
0.6238431088581847
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6556404371644047,
|
| 60 |
+
0.3572252105732042
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-0.9999999999999999,
|
| 66 |
+
1.7482354593273349e-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 |
+
0.9014528393745422,
|
| 78 |
+
0.8003877401351929,
|
| 79 |
+
0.9829942584037781
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.37471598386764526,
|
| 83 |
+
-0.8472502827644348,
|
| 84 |
+
-0.25070279836654663
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.28667332601863704,
|
| 88 |
+
-0.038315473170876704,
|
| 89 |
+
0.45974658906777444
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.17067058828305154,
|
| 93 |
+
0.23569952037784195,
|
| 94 |
+
0.21950702354181487
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
0.015172948963784228,
|
| 98 |
+
-0.44450502063220615,
|
| 99 |
+
0.23314444103329338
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5609069689572412,
|
| 103 |
+
0.38454179128009547,
|
| 104 |
+
0.7028102736473852
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999998807907104,
|
| 110 |
+
0.9984455108642578,
|
| 111 |
+
0.9434727430343628,
|
| 112 |
+
0.9062229990959167
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999930262565613,
|
| 116 |
+
-0.9988337755203247,
|
| 117 |
+
-0.9618149995803833,
|
| 118 |
+
2.1872274658107926e-07
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.25672297401336,
|
| 122 |
+
0.024425335806687216,
|
| 123 |
+
-0.08740977164483847,
|
| 124 |
+
0.16608739644018397
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.7932585208392383,
|
| 128 |
+
0.3102260195580416,
|
| 129 |
+
0.3772472576315148,
|
| 130 |
+
0.17451959300158773
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.99379006987882,
|
| 134 |
+
-0.5478102050038828,
|
| 135 |
+
-0.6101777952971636,
|
| 136 |
+
0.002274085014134748
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.8804856038896288,
|
| 140 |
+
0.5910611092873037,
|
| 141 |
+
0.5216058042993562,
|
| 142 |
+
0.5231965974012255
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.055869169533252716,
|
| 148 |
+
0.010916369967162609
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011436971835792065,
|
| 152 |
+
-0.05664053186774254
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.031651551516052555,
|
| 156 |
+
-0.03161481387920421
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.013119769222217526,
|
| 160 |
+
0.01306078225192402
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.0060999246446777336,
|
| 164 |
+
-0.04062601653964198
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04054891621517283,
|
| 168 |
+
-0.0059163567113456345
|
| 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 |
+
0.008144896753718373,
|
| 188 |
+
-0.00018893135146078263,
|
| 189 |
+
-0.0008739062727221845,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11030045424364411,
|
| 194 |
+
0.10148082570313594,
|
| 195 |
+
0.08975698373467506,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.07287373067581518,
|
| 200 |
+
-0.09446994199569948,
|
| 201 |
+
-0.07702818259249572,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.04485485414407898,
|
| 206 |
+
0.09626941605965177,
|
| 207 |
+
0.07610665906122742,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9221974313425939
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.38670194342612674
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.384693946883099
|
| 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 |
+
0.01976129923017491,
|
| 244 |
+
-0.020870631344490034,
|
| 245 |
+
-0.05993476517349181
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4347024990804053,
|
| 249 |
+
0.4221662976569997,
|
| 250 |
+
0.3845506843044066
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.8835792939767807,
|
| 254 |
+
-0.9280919251713778,
|
| 255 |
+
-0.783361071470956
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7622084438449307,
|
| 259 |
+
0.8625125252609077,
|
| 260 |
+
0.7470974753250706
|
| 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 |
+
0.008089673068544335,
|
| 276 |
+
-0.026498254627927348,
|
| 277 |
+
0.0029083551743059005
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11216734503733773,
|
| 281 |
+
0.13206002613798629,
|
| 282 |
+
0.12615870412692864
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2814627972846672,
|
| 286 |
+
-0.4342386516115439,
|
| 287 |
+
-0.34489755628340574
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.31879353266939386,
|
| 291 |
+
0.28411797683220563,
|
| 292 |
+
0.3597718671669894
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3824228874769045
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9240015096469786
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.5597272305813592
|
| 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 |
+
}
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase2/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b3aca09852c2e57eed5caa3eed39589cae9470e967a467792cd0e6356fafed3
|
| 3 |
+
size 5905
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/config.json
ADDED
|
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": 1.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.01,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 1.0,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_control_weight": 1.0,
|
| 75 |
+
"rkd_dtw_event_window_radius": 2,
|
| 76 |
+
"rkd_dtw_event_window_side": "both",
|
| 77 |
+
"rkd_dtw_gripper_weight": 1.0,
|
| 78 |
+
"rkd_dtw_mix_mode": "rowwise_adaptive_prob_mix",
|
| 79 |
+
"rkd_dtw_normalize_components": true,
|
| 80 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 81 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 82 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 83 |
+
"rkd_enabled": true,
|
| 84 |
+
"rkd_exclude_diagonal": true,
|
| 85 |
+
"rkd_fm_loss_weight": 1.0,
|
| 86 |
+
"rkd_loss_type": "kl",
|
| 87 |
+
"rkd_loss_weight": 0.5,
|
| 88 |
+
"rkd_loss_weight_end": 0.0,
|
| 89 |
+
"rkd_loss_weight_schedule": null,
|
| 90 |
+
"rkd_loss_weight_start": 0.1,
|
| 91 |
+
"rkd_raw_action_projector_dim": 512,
|
| 92 |
+
"rkd_raw_action_projector_enabled": false,
|
| 93 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 94 |
+
"rkd_relation_mode": "flatten",
|
| 95 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 96 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 97 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 98 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 99 |
+
"rkd_teacher_source": "dtw_action",
|
| 100 |
+
"rkd_vlm_temp": 0.1,
|
| 101 |
+
"torch_dtype": "bfloat16",
|
| 102 |
+
"transformers_version": "4.51.3"
|
| 103 |
+
}
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.3139495849609375,
|
| 8 |
+
0.4876587688922882,
|
| 9 |
+
0.7196521759033203
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.94293737411499,
|
| 13 |
+
-6.890198230743408,
|
| 14 |
+
0.6996827125549316
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.749288365724937,
|
| 18 |
+
-2.0455216239909983,
|
| 19 |
+
0.700532551020533
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6786295897046262,
|
| 23 |
+
1.312897969434244,
|
| 24 |
+
0.00133751758906258
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-1.4241907881148037,
|
| 28 |
+
-5.1716547935950885,
|
| 29 |
+
0.6999670898768614
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.198111545831555,
|
| 33 |
+
-0.5873531334104489,
|
| 34 |
+
0.7038833014242587
|
| 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 |
+
0.23084999587450247,
|
| 54 |
+
0.6238431088581847
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6556404371644047,
|
| 60 |
+
0.3572252105732042
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-0.9999999999999999,
|
| 66 |
+
1.7482354593273349e-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 |
+
0.9014528393745422,
|
| 78 |
+
0.8003877401351929,
|
| 79 |
+
0.9829942584037781
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.37471598386764526,
|
| 83 |
+
-0.8472502827644348,
|
| 84 |
+
-0.25070279836654663
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.28667332601863704,
|
| 88 |
+
-0.038315473170876704,
|
| 89 |
+
0.45974658906777444
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.17067058828305154,
|
| 93 |
+
0.23569952037784195,
|
| 94 |
+
0.21950702354181487
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
0.015172948963784228,
|
| 98 |
+
-0.44450502063220615,
|
| 99 |
+
0.23314444103329338
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5609069689572412,
|
| 103 |
+
0.38454179128009547,
|
| 104 |
+
0.7028102736473852
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999998807907104,
|
| 110 |
+
0.9984455108642578,
|
| 111 |
+
0.9434727430343628,
|
| 112 |
+
0.9062229990959167
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999930262565613,
|
| 116 |
+
-0.9988337755203247,
|
| 117 |
+
-0.9618149995803833,
|
| 118 |
+
2.1872274658107926e-07
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.25672297401336,
|
| 122 |
+
0.024425335806687216,
|
| 123 |
+
-0.08740977164483847,
|
| 124 |
+
0.16608739644018397
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.7932585208392383,
|
| 128 |
+
0.3102260195580416,
|
| 129 |
+
0.3772472576315148,
|
| 130 |
+
0.17451959300158773
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.99379006987882,
|
| 134 |
+
-0.5478102050038828,
|
| 135 |
+
-0.6101777952971636,
|
| 136 |
+
0.002274085014134748
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.8804856038896288,
|
| 140 |
+
0.5910611092873037,
|
| 141 |
+
0.5216058042993562,
|
| 142 |
+
0.5231965974012255
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.055869169533252716,
|
| 148 |
+
0.010916369967162609
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011436971835792065,
|
| 152 |
+
-0.05664053186774254
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.031651551516052555,
|
| 156 |
+
-0.03161481387920421
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.013119769222217526,
|
| 160 |
+
0.01306078225192402
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.0060999246446777336,
|
| 164 |
+
-0.04062601653964198
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04054891621517283,
|
| 168 |
+
-0.0059163567113456345
|
| 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 |
+
0.008144896753718373,
|
| 188 |
+
-0.00018893135146078263,
|
| 189 |
+
-0.0008739062727221845,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11030045424364411,
|
| 194 |
+
0.10148082570313594,
|
| 195 |
+
0.08975698373467506,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.07287373067581518,
|
| 200 |
+
-0.09446994199569948,
|
| 201 |
+
-0.07702818259249572,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.04485485414407898,
|
| 206 |
+
0.09626941605965177,
|
| 207 |
+
0.07610665906122742,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9221974313425939
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.38670194342612674
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.384693946883099
|
| 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 |
+
0.01976129923017491,
|
| 244 |
+
-0.020870631344490034,
|
| 245 |
+
-0.05993476517349181
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4347024990804053,
|
| 249 |
+
0.4221662976569997,
|
| 250 |
+
0.3845506843044066
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.8835792939767807,
|
| 254 |
+
-0.9280919251713778,
|
| 255 |
+
-0.783361071470956
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7622084438449307,
|
| 259 |
+
0.8625125252609077,
|
| 260 |
+
0.7470974753250706
|
| 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 |
+
0.008089673068544335,
|
| 276 |
+
-0.026498254627927348,
|
| 277 |
+
0.0029083551743059005
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11216734503733773,
|
| 281 |
+
0.13206002613798629,
|
| 282 |
+
0.12615870412692864
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2814627972846672,
|
| 286 |
+
-0.4342386516115439,
|
| 287 |
+
-0.34489755628340574
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.31879353266939386,
|
| 291 |
+
0.28411797683220563,
|
| 292 |
+
0.3597718671669894
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3824228874769045
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9240015096469786
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.5597272305813592
|
| 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 |
+
}
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/phase3/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:caf09b469d45d95a6d23d1dd33c6f8de88f5ffffef2313af0406f7c1353070f4
|
| 3 |
+
size 5905
|
DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix/default/resolved_config.yaml
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: DTW from 60k - RowWiseAdaptiveProbMix
|
| 2 |
+
policy_type: groot_DTW
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/Atomic26_baseline/checkpoint-60000
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/DTW/DTW_v1/from_60k/RowWiseAdaptiveProbMix
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: my_atomic26_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
dtw_enabled: true
|
| 13 |
+
dtw_relation_mode: flatten
|
| 14 |
+
dtw_loss_type: kl
|
| 15 |
+
dtw_action_temp: 0.1
|
| 16 |
+
dtw_student_temp: 0.1
|
| 17 |
+
dtw_student_reconstruction_head_enabled: true
|
| 18 |
+
dtw_student_reconstruction_head_dim: 512
|
| 19 |
+
dtw_student_reconstruction_head_hidden_dim: 512
|
| 20 |
+
dtw_student_reconstruction_head_pooling: attention
|
| 21 |
+
dtw_mix_mode: rowwise_adaptive_prob_mix
|
| 22 |
+
dtw_pos_weight: 1.0
|
| 23 |
+
dtw_rot_weight: 1.0
|
| 24 |
+
dtw_gripper_weight: 1.0
|
| 25 |
+
dtw_base_weight: 1.0
|
| 26 |
+
dtw_control_weight: 1.0
|
| 27 |
+
dtw_normalize_components: true
|
| 28 |
+
dtw_adaptive_mix_lambda: 1.0
|
| 29 |
+
dtw_adaptive_floor: 0.01
|
| 30 |
+
dtw_timing_log_interval: 1000
|
| 31 |
+
dtw_exclude_diagonal: true
|
| 32 |
+
phases:
|
| 33 |
+
- name: phase2_kl_dtw_aphead_rowwise_adaptive_only
|
| 34 |
+
max_steps: 30000
|
| 35 |
+
save_steps: 0
|
| 36 |
+
trainable:
|
| 37 |
+
preset: processing_line_only
|
| 38 |
+
tune_llm: false
|
| 39 |
+
tune_visual: false
|
| 40 |
+
tune_projector: false
|
| 41 |
+
tune_diffusion_model: false
|
| 42 |
+
losses:
|
| 43 |
+
dtw_fm_loss_weight: 0.0
|
| 44 |
+
dtw_kl_loss_weight: 1.0
|
| 45 |
+
- name: phase3_fm_kl_dtw_aphead_rowwise_adaptive_w0p5_f0p01
|
| 46 |
+
max_steps: 30000
|
| 47 |
+
save_steps: 0
|
| 48 |
+
trainable:
|
| 49 |
+
tune_llm: false
|
| 50 |
+
tune_visual: false
|
| 51 |
+
tune_projector: true
|
| 52 |
+
tune_diffusion_model: true
|
| 53 |
+
losses:
|
| 54 |
+
dtw_fm_loss_weight: 1.0
|
| 55 |
+
dtw_kl_loss_weight: 0.5
|
| 56 |
+
resolved_sweep: {}
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/TimeWeightedRowWiseDTW.yaml
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: DTW from base - TimeWeightedRowWiseDTW
|
| 2 |
+
policy_type: groot_DTW
|
| 3 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: my_atomic26_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
dtw_enabled: true
|
| 13 |
+
dtw_relation_mode: flatten
|
| 14 |
+
dtw_loss_type: kl
|
| 15 |
+
dtw_action_temp: 0.1
|
| 16 |
+
dtw_student_temp: 0.1
|
| 17 |
+
dtw_student_reconstruction_head_enabled: true
|
| 18 |
+
dtw_student_reconstruction_head_dim: 512
|
| 19 |
+
dtw_student_reconstruction_head_hidden_dim: 512
|
| 20 |
+
dtw_student_reconstruction_head_pooling: attention
|
| 21 |
+
dtw_mix_mode: time_weighted_rowwise_dtw
|
| 22 |
+
dtw_pos_weight: 1.0
|
| 23 |
+
dtw_rot_weight: 1.0
|
| 24 |
+
dtw_gripper_weight: 1.0
|
| 25 |
+
dtw_base_weight: 1.0
|
| 26 |
+
dtw_control_weight: 1.0
|
| 27 |
+
dtw_normalize_components: true
|
| 28 |
+
dtw_adaptive_mix_lambda: 1.0
|
| 29 |
+
dtw_adaptive_floor: 0.01
|
| 30 |
+
dtw_timing_log_interval: 1000
|
| 31 |
+
dtw_exclude_diagonal: true
|
| 32 |
+
phases:
|
| 33 |
+
- name: phase2_kl_dtw_aphead_time_weighted_only
|
| 34 |
+
max_steps: 30000
|
| 35 |
+
save_steps: 0
|
| 36 |
+
trainable:
|
| 37 |
+
preset: processing_line_only
|
| 38 |
+
tune_llm: false
|
| 39 |
+
tune_visual: false
|
| 40 |
+
tune_projector: false
|
| 41 |
+
tune_diffusion_model: false
|
| 42 |
+
losses:
|
| 43 |
+
dtw_fm_loss_weight: 0.0
|
| 44 |
+
dtw_kl_loss_weight: 1.0
|
| 45 |
+
- name: phase3_fm_kl_dtw_aphead_time_weighted_w0p5_f0p01
|
| 46 |
+
max_steps: 30000
|
| 47 |
+
save_steps: 0
|
| 48 |
+
trainable:
|
| 49 |
+
tune_llm: false
|
| 50 |
+
tune_visual: false
|
| 51 |
+
tune_projector: true
|
| 52 |
+
tune_diffusion_model: true
|
| 53 |
+
losses:
|
| 54 |
+
dtw_fm_loss_weight: 1.0
|
| 55 |
+
dtw_kl_loss_weight: 0.5
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/config.json
ADDED
|
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": 1.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.01,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 1.0,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_control_weight": 1.0,
|
| 75 |
+
"rkd_dtw_event_window_radius": 2,
|
| 76 |
+
"rkd_dtw_event_window_side": "both",
|
| 77 |
+
"rkd_dtw_gripper_weight": 1.0,
|
| 78 |
+
"rkd_dtw_mix_mode": "time_weighted_rowwise_dtw",
|
| 79 |
+
"rkd_dtw_normalize_components": true,
|
| 80 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 81 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 82 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 83 |
+
"rkd_enabled": true,
|
| 84 |
+
"rkd_exclude_diagonal": true,
|
| 85 |
+
"rkd_fm_loss_weight": 0.0,
|
| 86 |
+
"rkd_loss_type": "kl",
|
| 87 |
+
"rkd_loss_weight": 1.0,
|
| 88 |
+
"rkd_loss_weight_end": 0.0,
|
| 89 |
+
"rkd_loss_weight_schedule": null,
|
| 90 |
+
"rkd_loss_weight_start": 0.1,
|
| 91 |
+
"rkd_raw_action_projector_dim": 512,
|
| 92 |
+
"rkd_raw_action_projector_enabled": false,
|
| 93 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 94 |
+
"rkd_relation_mode": "flatten",
|
| 95 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 96 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 97 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 98 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 99 |
+
"rkd_teacher_source": "dtw_action",
|
| 100 |
+
"rkd_vlm_temp": 0.1,
|
| 101 |
+
"torch_dtype": "bfloat16",
|
| 102 |
+
"transformers_version": "4.51.3"
|
| 103 |
+
}
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.3139495849609375,
|
| 8 |
+
0.4876587688922882,
|
| 9 |
+
0.7196521759033203
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.94293737411499,
|
| 13 |
+
-6.890198230743408,
|
| 14 |
+
0.6996827125549316
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.749288365724937,
|
| 18 |
+
-2.0455216239909983,
|
| 19 |
+
0.700532551020533
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6786295897046262,
|
| 23 |
+
1.312897969434244,
|
| 24 |
+
0.00133751758906258
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-1.4241907881148037,
|
| 28 |
+
-5.1716547935950885,
|
| 29 |
+
0.6999670898768614
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.198111545831555,
|
| 33 |
+
-0.5873531334104489,
|
| 34 |
+
0.7038833014242587
|
| 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 |
+
0.23084999587450247,
|
| 54 |
+
0.6238431088581847
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6556404371644047,
|
| 60 |
+
0.3572252105732042
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-0.9999999999999999,
|
| 66 |
+
1.7482354593273349e-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 |
+
0.9014528393745422,
|
| 78 |
+
0.8003877401351929,
|
| 79 |
+
0.9829942584037781
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.37471598386764526,
|
| 83 |
+
-0.8472502827644348,
|
| 84 |
+
-0.25070279836654663
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.28667332601863704,
|
| 88 |
+
-0.038315473170876704,
|
| 89 |
+
0.45974658906777444
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.17067058828305154,
|
| 93 |
+
0.23569952037784195,
|
| 94 |
+
0.21950702354181487
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
0.015172948963784228,
|
| 98 |
+
-0.44450502063220615,
|
| 99 |
+
0.23314444103329338
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5609069689572412,
|
| 103 |
+
0.38454179128009547,
|
| 104 |
+
0.7028102736473852
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999998807907104,
|
| 110 |
+
0.9984455108642578,
|
| 111 |
+
0.9434727430343628,
|
| 112 |
+
0.9062229990959167
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999930262565613,
|
| 116 |
+
-0.9988337755203247,
|
| 117 |
+
-0.9618149995803833,
|
| 118 |
+
2.1872274658107926e-07
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.25672297401336,
|
| 122 |
+
0.024425335806687216,
|
| 123 |
+
-0.08740977164483847,
|
| 124 |
+
0.16608739644018397
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.7932585208392383,
|
| 128 |
+
0.3102260195580416,
|
| 129 |
+
0.3772472576315148,
|
| 130 |
+
0.17451959300158773
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.99379006987882,
|
| 134 |
+
-0.5478102050038828,
|
| 135 |
+
-0.6101777952971636,
|
| 136 |
+
0.002274085014134748
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.8804856038896288,
|
| 140 |
+
0.5910611092873037,
|
| 141 |
+
0.5216058042993562,
|
| 142 |
+
0.5231965974012255
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.055869169533252716,
|
| 148 |
+
0.010916369967162609
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011436971835792065,
|
| 152 |
+
-0.05664053186774254
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.031651551516052555,
|
| 156 |
+
-0.03161481387920421
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.013119769222217526,
|
| 160 |
+
0.01306078225192402
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.0060999246446777336,
|
| 164 |
+
-0.04062601653964198
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04054891621517283,
|
| 168 |
+
-0.0059163567113456345
|
| 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 |
+
0.008144896753718373,
|
| 188 |
+
-0.00018893135146078263,
|
| 189 |
+
-0.0008739062727221845,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11030045424364411,
|
| 194 |
+
0.10148082570313594,
|
| 195 |
+
0.08975698373467506,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.07287373067581518,
|
| 200 |
+
-0.09446994199569948,
|
| 201 |
+
-0.07702818259249572,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.04485485414407898,
|
| 206 |
+
0.09626941605965177,
|
| 207 |
+
0.07610665906122742,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9221974313425939
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.38670194342612674
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.384693946883099
|
| 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 |
+
0.01976129923017491,
|
| 244 |
+
-0.020870631344490034,
|
| 245 |
+
-0.05993476517349181
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4347024990804053,
|
| 249 |
+
0.4221662976569997,
|
| 250 |
+
0.3845506843044066
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.8835792939767807,
|
| 254 |
+
-0.9280919251713778,
|
| 255 |
+
-0.783361071470956
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7622084438449307,
|
| 259 |
+
0.8625125252609077,
|
| 260 |
+
0.7470974753250706
|
| 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 |
+
0.008089673068544335,
|
| 276 |
+
-0.026498254627927348,
|
| 277 |
+
0.0029083551743059005
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11216734503733773,
|
| 281 |
+
0.13206002613798629,
|
| 282 |
+
0.12615870412692864
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2814627972846672,
|
| 286 |
+
-0.4342386516115439,
|
| 287 |
+
-0.34489755628340574
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.31879353266939386,
|
| 291 |
+
0.28411797683220563,
|
| 292 |
+
0.3597718671669894
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3824228874769045
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9240015096469786
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.5597272305813592
|
| 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 |
+
}
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase2/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8667d70e6e24259b9d1fdabf20a4f3af370926650a253b6116783824521ed0a
|
| 3 |
+
size 5905
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/config.json
ADDED
|
@@ -0,0 +1,103 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": 1.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.01,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 1.0,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_control_weight": 1.0,
|
| 75 |
+
"rkd_dtw_event_window_radius": 2,
|
| 76 |
+
"rkd_dtw_event_window_side": "both",
|
| 77 |
+
"rkd_dtw_gripper_weight": 1.0,
|
| 78 |
+
"rkd_dtw_mix_mode": "time_weighted_rowwise_dtw",
|
| 79 |
+
"rkd_dtw_normalize_components": true,
|
| 80 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 81 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 82 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 83 |
+
"rkd_enabled": true,
|
| 84 |
+
"rkd_exclude_diagonal": true,
|
| 85 |
+
"rkd_fm_loss_weight": 1.0,
|
| 86 |
+
"rkd_loss_type": "kl",
|
| 87 |
+
"rkd_loss_weight": 0.5,
|
| 88 |
+
"rkd_loss_weight_end": 0.0,
|
| 89 |
+
"rkd_loss_weight_schedule": null,
|
| 90 |
+
"rkd_loss_weight_start": 0.1,
|
| 91 |
+
"rkd_raw_action_projector_dim": 512,
|
| 92 |
+
"rkd_raw_action_projector_enabled": false,
|
| 93 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 94 |
+
"rkd_relation_mode": "flatten",
|
| 95 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 96 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 97 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 98 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 99 |
+
"rkd_teacher_source": "dtw_action",
|
| 100 |
+
"rkd_vlm_temp": 0.1,
|
| 101 |
+
"torch_dtype": "bfloat16",
|
| 102 |
+
"transformers_version": "4.51.3"
|
| 103 |
+
}
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.3139495849609375,
|
| 8 |
+
0.4876587688922882,
|
| 9 |
+
0.7196521759033203
|
| 10 |
+
],
|
| 11 |
+
"min": [
|
| 12 |
+
-4.94293737411499,
|
| 13 |
+
-6.890198230743408,
|
| 14 |
+
0.6996827125549316
|
| 15 |
+
],
|
| 16 |
+
"mean": [
|
| 17 |
+
2.749288365724937,
|
| 18 |
+
-2.0455216239909983,
|
| 19 |
+
0.700532551020533
|
| 20 |
+
],
|
| 21 |
+
"std": [
|
| 22 |
+
1.6786295897046262,
|
| 23 |
+
1.312897969434244,
|
| 24 |
+
0.00133751758906258
|
| 25 |
+
],
|
| 26 |
+
"q01": [
|
| 27 |
+
-1.4241907881148037,
|
| 28 |
+
-5.1716547935950885,
|
| 29 |
+
0.6999670898768614
|
| 30 |
+
],
|
| 31 |
+
"q99": [
|
| 32 |
+
6.198111545831555,
|
| 33 |
+
-0.5873531334104489,
|
| 34 |
+
0.7038833014242587
|
| 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 |
+
0.23084999587450247,
|
| 54 |
+
0.6238431088581847
|
| 55 |
+
],
|
| 56 |
+
"std": [
|
| 57 |
+
0.0,
|
| 58 |
+
0.0,
|
| 59 |
+
0.6556404371644047,
|
| 60 |
+
0.3572252105732042
|
| 61 |
+
],
|
| 62 |
+
"q01": [
|
| 63 |
+
0.0,
|
| 64 |
+
0.0,
|
| 65 |
+
-0.9999999999999999,
|
| 66 |
+
1.7482354593273349e-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 |
+
0.9014528393745422,
|
| 78 |
+
0.8003877401351929,
|
| 79 |
+
0.9829942584037781
|
| 80 |
+
],
|
| 81 |
+
"min": [
|
| 82 |
+
-0.37471598386764526,
|
| 83 |
+
-0.8472502827644348,
|
| 84 |
+
-0.25070279836654663
|
| 85 |
+
],
|
| 86 |
+
"mean": [
|
| 87 |
+
0.28667332601863704,
|
| 88 |
+
-0.038315473170876704,
|
| 89 |
+
0.45974658906777444
|
| 90 |
+
],
|
| 91 |
+
"std": [
|
| 92 |
+
0.17067058828305154,
|
| 93 |
+
0.23569952037784195,
|
| 94 |
+
0.21950702354181487
|
| 95 |
+
],
|
| 96 |
+
"q01": [
|
| 97 |
+
0.015172948963784228,
|
| 98 |
+
-0.44450502063220615,
|
| 99 |
+
0.23314444103329338
|
| 100 |
+
],
|
| 101 |
+
"q99": [
|
| 102 |
+
0.5609069689572412,
|
| 103 |
+
0.38454179128009547,
|
| 104 |
+
0.7028102736473852
|
| 105 |
+
]
|
| 106 |
+
},
|
| 107 |
+
"end_effector_rotation_relative": {
|
| 108 |
+
"max": [
|
| 109 |
+
0.9999998807907104,
|
| 110 |
+
0.9984455108642578,
|
| 111 |
+
0.9434727430343628,
|
| 112 |
+
0.9062229990959167
|
| 113 |
+
],
|
| 114 |
+
"min": [
|
| 115 |
+
-0.9999930262565613,
|
| 116 |
+
-0.9988337755203247,
|
| 117 |
+
-0.9618149995803833,
|
| 118 |
+
2.1872274658107926e-07
|
| 119 |
+
],
|
| 120 |
+
"mean": [
|
| 121 |
+
-0.25672297401336,
|
| 122 |
+
0.024425335806687216,
|
| 123 |
+
-0.08740977164483847,
|
| 124 |
+
0.16608739644018397
|
| 125 |
+
],
|
| 126 |
+
"std": [
|
| 127 |
+
0.7932585208392383,
|
| 128 |
+
0.3102260195580416,
|
| 129 |
+
0.3772472576315148,
|
| 130 |
+
0.17451959300158773
|
| 131 |
+
],
|
| 132 |
+
"q01": [
|
| 133 |
+
-0.99379006987882,
|
| 134 |
+
-0.5478102050038828,
|
| 135 |
+
-0.6101777952971636,
|
| 136 |
+
0.002274085014134748
|
| 137 |
+
],
|
| 138 |
+
"q99": [
|
| 139 |
+
0.8804856038896288,
|
| 140 |
+
0.5910611092873037,
|
| 141 |
+
0.5216058042993562,
|
| 142 |
+
0.5231965974012255
|
| 143 |
+
]
|
| 144 |
+
},
|
| 145 |
+
"gripper_qpos": {
|
| 146 |
+
"max": [
|
| 147 |
+
0.055869169533252716,
|
| 148 |
+
0.010916369967162609
|
| 149 |
+
],
|
| 150 |
+
"min": [
|
| 151 |
+
-0.011436971835792065,
|
| 152 |
+
-0.05664053186774254
|
| 153 |
+
],
|
| 154 |
+
"mean": [
|
| 155 |
+
0.031651551516052555,
|
| 156 |
+
-0.03161481387920421
|
| 157 |
+
],
|
| 158 |
+
"std": [
|
| 159 |
+
0.013119769222217526,
|
| 160 |
+
0.01306078225192402
|
| 161 |
+
],
|
| 162 |
+
"q01": [
|
| 163 |
+
0.0060999246446777336,
|
| 164 |
+
-0.04062601653964198
|
| 165 |
+
],
|
| 166 |
+
"q99": [
|
| 167 |
+
0.04054891621517283,
|
| 168 |
+
-0.0059163567113456345
|
| 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 |
+
0.008144896753718373,
|
| 188 |
+
-0.00018893135146078263,
|
| 189 |
+
-0.0008739062727221845,
|
| 190 |
+
0.0
|
| 191 |
+
],
|
| 192 |
+
"std": [
|
| 193 |
+
0.11030045424364411,
|
| 194 |
+
0.10148082570313594,
|
| 195 |
+
0.08975698373467506,
|
| 196 |
+
0.0
|
| 197 |
+
],
|
| 198 |
+
"q01": [
|
| 199 |
+
-0.07287373067581518,
|
| 200 |
+
-0.09446994199569948,
|
| 201 |
+
-0.07702818259249572,
|
| 202 |
+
0.0
|
| 203 |
+
],
|
| 204 |
+
"q99": [
|
| 205 |
+
0.04485485414407898,
|
| 206 |
+
0.09626941605965177,
|
| 207 |
+
0.07610665906122742,
|
| 208 |
+
0.0
|
| 209 |
+
]
|
| 210 |
+
},
|
| 211 |
+
"control_mode": {
|
| 212 |
+
"max": [
|
| 213 |
+
1.0
|
| 214 |
+
],
|
| 215 |
+
"min": [
|
| 216 |
+
-1.0
|
| 217 |
+
],
|
| 218 |
+
"mean": [
|
| 219 |
+
-0.9221974313425939
|
| 220 |
+
],
|
| 221 |
+
"std": [
|
| 222 |
+
0.38670194342612674
|
| 223 |
+
],
|
| 224 |
+
"q01": [
|
| 225 |
+
-1.0
|
| 226 |
+
],
|
| 227 |
+
"q99": [
|
| 228 |
+
-0.384693946883099
|
| 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 |
+
0.01976129923017491,
|
| 244 |
+
-0.020870631344490034,
|
| 245 |
+
-0.05993476517349181
|
| 246 |
+
],
|
| 247 |
+
"std": [
|
| 248 |
+
0.4347024990804053,
|
| 249 |
+
0.4221662976569997,
|
| 250 |
+
0.3845506843044066
|
| 251 |
+
],
|
| 252 |
+
"q01": [
|
| 253 |
+
-0.8835792939767807,
|
| 254 |
+
-0.9280919251713778,
|
| 255 |
+
-0.783361071470956
|
| 256 |
+
],
|
| 257 |
+
"q99": [
|
| 258 |
+
0.7622084438449307,
|
| 259 |
+
0.8625125252609077,
|
| 260 |
+
0.7470974753250706
|
| 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 |
+
0.008089673068544335,
|
| 276 |
+
-0.026498254627927348,
|
| 277 |
+
0.0029083551743059005
|
| 278 |
+
],
|
| 279 |
+
"std": [
|
| 280 |
+
0.11216734503733773,
|
| 281 |
+
0.13206002613798629,
|
| 282 |
+
0.12615870412692864
|
| 283 |
+
],
|
| 284 |
+
"q01": [
|
| 285 |
+
-0.2814627972846672,
|
| 286 |
+
-0.4342386516115439,
|
| 287 |
+
-0.34489755628340574
|
| 288 |
+
],
|
| 289 |
+
"q99": [
|
| 290 |
+
0.31879353266939386,
|
| 291 |
+
0.28411797683220563,
|
| 292 |
+
0.3597718671669894
|
| 293 |
+
]
|
| 294 |
+
},
|
| 295 |
+
"gripper_close": {
|
| 296 |
+
"max": [
|
| 297 |
+
1.0
|
| 298 |
+
],
|
| 299 |
+
"min": [
|
| 300 |
+
-1.0
|
| 301 |
+
],
|
| 302 |
+
"mean": [
|
| 303 |
+
-0.3824228874769045
|
| 304 |
+
],
|
| 305 |
+
"std": [
|
| 306 |
+
0.9240015096469786
|
| 307 |
+
],
|
| 308 |
+
"q01": [
|
| 309 |
+
-1.0
|
| 310 |
+
],
|
| 311 |
+
"q99": [
|
| 312 |
+
0.5597272305813592
|
| 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 |
+
}
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/phase3/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7feaa55fa94489a4456f9b9e1fc29086bf99cca15f2b801ade819d01457d3c32
|
| 3 |
+
size 5905
|
DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW/default/resolved_config.yaml
ADDED
|
@@ -0,0 +1,56 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: DTW from base - TimeWeightedRowWiseDTW
|
| 2 |
+
policy_type: groot_DTW
|
| 3 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/DTW/DTW_v1/from_base/TimeWeightedRowWiseDTW
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: my_atomic26_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
dtw_enabled: true
|
| 13 |
+
dtw_relation_mode: flatten
|
| 14 |
+
dtw_loss_type: kl
|
| 15 |
+
dtw_action_temp: 0.1
|
| 16 |
+
dtw_student_temp: 0.1
|
| 17 |
+
dtw_student_reconstruction_head_enabled: true
|
| 18 |
+
dtw_student_reconstruction_head_dim: 512
|
| 19 |
+
dtw_student_reconstruction_head_hidden_dim: 512
|
| 20 |
+
dtw_student_reconstruction_head_pooling: attention
|
| 21 |
+
dtw_mix_mode: time_weighted_rowwise_dtw
|
| 22 |
+
dtw_pos_weight: 1.0
|
| 23 |
+
dtw_rot_weight: 1.0
|
| 24 |
+
dtw_gripper_weight: 1.0
|
| 25 |
+
dtw_base_weight: 1.0
|
| 26 |
+
dtw_control_weight: 1.0
|
| 27 |
+
dtw_normalize_components: true
|
| 28 |
+
dtw_adaptive_mix_lambda: 1.0
|
| 29 |
+
dtw_adaptive_floor: 0.01
|
| 30 |
+
dtw_timing_log_interval: 1000
|
| 31 |
+
dtw_exclude_diagonal: true
|
| 32 |
+
phases:
|
| 33 |
+
- name: phase2_kl_dtw_aphead_time_weighted_only
|
| 34 |
+
max_steps: 30000
|
| 35 |
+
save_steps: 0
|
| 36 |
+
trainable:
|
| 37 |
+
preset: processing_line_only
|
| 38 |
+
tune_llm: false
|
| 39 |
+
tune_visual: false
|
| 40 |
+
tune_projector: false
|
| 41 |
+
tune_diffusion_model: false
|
| 42 |
+
losses:
|
| 43 |
+
dtw_fm_loss_weight: 0.0
|
| 44 |
+
dtw_kl_loss_weight: 1.0
|
| 45 |
+
- name: phase3_fm_kl_dtw_aphead_time_weighted_w0p5_f0p01
|
| 46 |
+
max_steps: 30000
|
| 47 |
+
save_steps: 0
|
| 48 |
+
trainable:
|
| 49 |
+
tune_llm: false
|
| 50 |
+
tune_visual: false
|
| 51 |
+
tune_projector: true
|
| 52 |
+
tune_diffusion_model: true
|
| 53 |
+
losses:
|
| 54 |
+
dtw_fm_loss_weight: 1.0
|
| 55 |
+
dtw_kl_loss_weight: 0.5
|
| 56 |
+
resolved_sweep: {}
|
RKDv2.1/DA_MeanPool+LP/default/RKD v2.1 DA MF phase3 only.yaml
ADDED
|
@@ -0,0 +1,37 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: RKD v2.1 DA, Student:MeanPool + LP phase3
|
| 2 |
+
policy_type: groot_MGRKD
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP/default/phase2
|
| 4 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP
|
| 5 |
+
dataset:
|
| 6 |
+
dataset_soup: my_atomic26_human
|
| 7 |
+
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
+
seed: 42
|
| 11 |
+
model:
|
| 12 |
+
rkd_student_reconstruction_head_pooling: mean
|
| 13 |
+
phases:
|
| 14 |
+
- name: phase3_fm_mgrkd_da_student_meanpool_fixed_0p5
|
| 15 |
+
max_steps: 30000
|
| 16 |
+
save_steps: 0
|
| 17 |
+
trainable:
|
| 18 |
+
tune_llm: false
|
| 19 |
+
tune_visual: false
|
| 20 |
+
tune_projector: true
|
| 21 |
+
tune_diffusion_model: true
|
| 22 |
+
losses:
|
| 23 |
+
rkd_enabled: true
|
| 24 |
+
rkd_fm_loss_weight: 1.0
|
| 25 |
+
rkd_loss_weight: 0.5
|
| 26 |
+
rkd_relation_mode: flatten
|
| 27 |
+
rkd_loss_type: distance_angle
|
| 28 |
+
rkd_teacher_source: action_encoder
|
| 29 |
+
rkd_action_encoder_projector_enabled: true
|
| 30 |
+
rkd_action_encoder_projector_dim: 512
|
| 31 |
+
rkd_action_encoder_projector_pooling: flatten
|
| 32 |
+
rkd_student_reconstruction_head_enabled: true
|
| 33 |
+
rkd_student_reconstruction_head_dim: 512
|
| 34 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 35 |
+
rkd_distance_loss_weight: 1.0
|
| 36 |
+
rkd_angle_loss_weight: 2.0
|
| 37 |
+
rkd_exclude_diagonal: true
|
RKDv2.1/DA_MeanPool+LP/default/phase3/config.json
ADDED
|
@@ -0,0 +1,100 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 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": true,
|
| 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": "flatten",
|
| 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": "mean",
|
| 96 |
+
"rkd_teacher_source": "action_encoder",
|
| 97 |
+
"rkd_vlm_temp": 0.1,
|
| 98 |
+
"torch_dtype": "bfloat16",
|
| 99 |
+
"transformers_version": "4.51.3"
|
| 100 |
+
}
|
RKDv2.1/DA_MeanPool+LP/default/phase3/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
RKDv2.1/DA_MeanPool+LP/default/phase3/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
RKDv2.1/DA_MeanPool+LP/default/phase3/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0fd387214b8bb86f373a7906c705a1a28392832667c23e2aa8f593e67bc2f2ef
|
| 3 |
+
size 5905
|
RKDv2.1/DA_MeanPool+LP/default/resolved_config.yaml
CHANGED
|
@@ -1,41 +1,16 @@
|
|
| 1 |
-
experiment_name: RKD v2.1 DA, Student:MeanPool + LP
|
| 2 |
policy_type: groot_MGRKD
|
| 3 |
-
base_model_path: ${HOME}/groot_robocasa/
|
| 4 |
output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP
|
| 5 |
dataset:
|
| 6 |
dataset_soup: my_atomic26_human
|
| 7 |
training:
|
| 8 |
-
num_gpus:
|
| 9 |
-
batch_size:
|
| 10 |
seed: 42
|
| 11 |
model:
|
| 12 |
rkd_student_reconstruction_head_pooling: mean
|
| 13 |
phases:
|
| 14 |
-
- name: phase2_mgrkd_da_student_meanpool_only
|
| 15 |
-
max_steps: 30000
|
| 16 |
-
save_steps: 0
|
| 17 |
-
trainable:
|
| 18 |
-
preset: processing_line_only
|
| 19 |
-
tune_llm: false
|
| 20 |
-
tune_visual: false
|
| 21 |
-
tune_projector: false
|
| 22 |
-
tune_diffusion_model: false
|
| 23 |
-
losses:
|
| 24 |
-
rkd_enabled: true
|
| 25 |
-
rkd_fm_loss_weight: 0.0
|
| 26 |
-
rkd_loss_weight: 1.0
|
| 27 |
-
rkd_relation_mode: flatten
|
| 28 |
-
rkd_loss_type: distance_angle
|
| 29 |
-
rkd_teacher_source: action_encoder
|
| 30 |
-
rkd_action_encoder_projector_enabled: true
|
| 31 |
-
rkd_action_encoder_projector_dim: 512
|
| 32 |
-
rkd_action_encoder_projector_pooling: flatten
|
| 33 |
-
rkd_student_reconstruction_head_enabled: true
|
| 34 |
-
rkd_student_reconstruction_head_dim: 512
|
| 35 |
-
rkd_student_reconstruction_head_hidden_dim: 512
|
| 36 |
-
rkd_distance_loss_weight: 1.0
|
| 37 |
-
rkd_angle_loss_weight: 2.0
|
| 38 |
-
rkd_exclude_diagonal: true
|
| 39 |
- name: phase3_fm_mgrkd_da_student_meanpool_fixed_0p5
|
| 40 |
max_steps: 30000
|
| 41 |
save_steps: 0
|
|
|
|
| 1 |
+
experiment_name: RKD v2.1 DA, Student:MeanPool + LP phase3
|
| 2 |
policy_type: groot_MGRKD
|
| 3 |
+
base_model_path: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP/default/phase2
|
| 4 |
output_root: ${HOME}/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP
|
| 5 |
dataset:
|
| 6 |
dataset_soup: my_atomic26_human
|
| 7 |
training:
|
| 8 |
+
num_gpus: 2
|
| 9 |
+
batch_size: 64
|
| 10 |
seed: 42
|
| 11 |
model:
|
| 12 |
rkd_student_reconstruction_head_pooling: mean
|
| 13 |
phases:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 14 |
- name: phase3_fm_mgrkd_da_student_meanpool_fixed_0p5
|
| 15 |
max_steps: 30000
|
| 16 |
save_steps: 0
|
RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2.logpath
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
/home/seonho/groot_robocasa/robocasa_v2/RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2_20260626_103126.log
|
RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2.pid
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
2909377
|
RKDv2.1/DA_MeanPool+LP/launch_logs/train_phase3_gpu2-3_bs64_numgpu2_20260626_103126.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|