Add files using upload-large-folder tool
Browse files- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml +1 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml +1 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml +40 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/trainer_state.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/resolved_config.yaml +41 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml +1 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json +431 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/scheduler.pt +3 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json +101 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json +0 -0
- TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml +1 -0
- TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_20260702_121805.log +195 -0
- TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_tensorboard_20260702_122542.log +0 -0
- TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_20260702_121805.log +195 -0
- TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_tensorboard_20260702_122542.log +0 -0
- TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k_gpu0_bs128_numgpu1.pid +1 -0
- TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1.pid +1 -0
- TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1.pid +1 -0
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml
CHANGED
|
@@ -9,6 +9,7 @@ training:
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
|
|
|
| 12 |
model:
|
| 13 |
rkd_teacher_source: timewarp_action
|
| 14 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
|
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
model:
|
| 14 |
rkd_teacher_source: timewarp_action
|
| 15 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5000130901875584,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml
CHANGED
|
@@ -9,6 +9,7 @@ training:
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
|
|
|
| 12 |
model:
|
| 13 |
rkd_teacher_source: timewarp_action
|
| 14 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
|
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
model:
|
| 14 |
rkd_teacher_source: timewarp_action
|
| 15 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml
ADDED
|
@@ -0,0 +1,40 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD cosine
|
| 2 |
+
decay from GR00T base to 120k
|
| 3 |
+
policy_type: groot_RKDv3
|
| 4 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 5 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k
|
| 6 |
+
dataset:
|
| 7 |
+
dataset_soup: my_atomic26_human
|
| 8 |
+
training:
|
| 9 |
+
num_gpus: 1
|
| 10 |
+
batch_size: 128
|
| 11 |
+
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
+
model:
|
| 14 |
+
rkd_teacher_source: timewarp_action
|
| 15 |
+
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
| 16 |
+
rkd_relation_mode: token_pair_mean
|
| 17 |
+
rkd_loss_type: distance_angle
|
| 18 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 19 |
+
rkd_student_reconstruction_head_enabled: true
|
| 20 |
+
rkd_student_reconstruction_head_dim: 512
|
| 21 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 22 |
+
rkd_distance_loss_weight: 1.0
|
| 23 |
+
rkd_angle_loss_weight: 2.0
|
| 24 |
+
rkd_exclude_diagonal: true
|
| 25 |
+
phases:
|
| 26 |
+
- name: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k
|
| 27 |
+
max_steps: 120000
|
| 28 |
+
save_steps: 30000
|
| 29 |
+
trainable:
|
| 30 |
+
tune_llm: false
|
| 31 |
+
tune_visual: false
|
| 32 |
+
tune_projector: true
|
| 33 |
+
tune_diffusion_model: true
|
| 34 |
+
losses:
|
| 35 |
+
rkd_enabled: true
|
| 36 |
+
rkd_fm_loss_weight: 1.0
|
| 37 |
+
rkd_loss_weight: 1.0
|
| 38 |
+
rkd_loss_weight_schedule: cosine_decay
|
| 39 |
+
rkd_loss_weight_start: 1.0
|
| 40 |
+
rkd_loss_weight_end: 0.0
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.8535603325875607,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5000065450392368,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.14644892343511284,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/experiment_cfg/metadata.json
ADDED
|
@@ -0,0 +1,431 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"new_embodiment": {
|
| 3 |
+
"statistics": {
|
| 4 |
+
"state": {
|
| 5 |
+
"base_position": {
|
| 6 |
+
"max": [
|
| 7 |
+
7.3139495849609375,
|
| 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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/resolved_config.yaml
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD cosine
|
| 2 |
+
decay from GR00T base to 120k
|
| 3 |
+
policy_type: groot_RKDv3
|
| 4 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 5 |
+
output_root: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k
|
| 6 |
+
dataset:
|
| 7 |
+
dataset_soup: my_atomic26_human
|
| 8 |
+
training:
|
| 9 |
+
num_gpus: 1
|
| 10 |
+
batch_size: 128
|
| 11 |
+
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
+
model:
|
| 14 |
+
rkd_teacher_source: timewarp_action
|
| 15 |
+
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
| 16 |
+
rkd_relation_mode: token_pair_mean
|
| 17 |
+
rkd_loss_type: distance_angle
|
| 18 |
+
rkd_student_reconstruction_head_pooling: attention
|
| 19 |
+
rkd_student_reconstruction_head_enabled: true
|
| 20 |
+
rkd_student_reconstruction_head_dim: 512
|
| 21 |
+
rkd_student_reconstruction_head_hidden_dim: 512
|
| 22 |
+
rkd_distance_loss_weight: 1.0
|
| 23 |
+
rkd_angle_loss_weight: 2.0
|
| 24 |
+
rkd_exclude_diagonal: true
|
| 25 |
+
phases:
|
| 26 |
+
- name: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k
|
| 27 |
+
max_steps: 120000
|
| 28 |
+
save_steps: 30000
|
| 29 |
+
trainable:
|
| 30 |
+
tune_llm: false
|
| 31 |
+
tune_visual: false
|
| 32 |
+
tune_projector: true
|
| 33 |
+
tune_diffusion_model: true
|
| 34 |
+
losses:
|
| 35 |
+
rkd_enabled: true
|
| 36 |
+
rkd_fm_loss_weight: 1.0
|
| 37 |
+
rkd_loss_weight: 1.0
|
| 38 |
+
rkd_loss_weight_schedule: cosine_decay
|
| 39 |
+
rkd_loss_weight_start: 1.0
|
| 40 |
+
rkd_loss_weight_end: 0.0
|
| 41 |
+
resolved_sweep: {}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml
CHANGED
|
@@ -9,6 +9,7 @@ training:
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
|
|
|
| 12 |
model:
|
| 13 |
rkd_teacher_source: timewarp_action
|
| 14 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
|
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
model:
|
| 14 |
rkd_teacher_source: timewarp_action
|
| 15 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.5000130901875584,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/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 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ae8d5d8de3b387871d02c6360a92a63b63211c9bc3fc033261259c682861fdd
|
| 3 |
+
size 1465
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json
ADDED
|
@@ -0,0 +1,101 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"action_dim": 32,
|
| 3 |
+
"action_head_cfg": {
|
| 4 |
+
"action_dim": 32,
|
| 5 |
+
"action_horizon": 16,
|
| 6 |
+
"add_pos_embed": true,
|
| 7 |
+
"backbone_embedding_dim": 2048,
|
| 8 |
+
"diffusion_model_cfg": {
|
| 9 |
+
"attention_head_dim": 48,
|
| 10 |
+
"cross_attention_dim": 2048,
|
| 11 |
+
"dropout": 0.2,
|
| 12 |
+
"final_dropout": true,
|
| 13 |
+
"interleave_self_attention": true,
|
| 14 |
+
"norm_type": "ada_norm",
|
| 15 |
+
"num_attention_heads": 32,
|
| 16 |
+
"num_layers": 16,
|
| 17 |
+
"output_dim": 1024,
|
| 18 |
+
"positional_embeddings": null
|
| 19 |
+
},
|
| 20 |
+
"hidden_size": 1024,
|
| 21 |
+
"input_embedding_dim": 1536,
|
| 22 |
+
"max_action_dim": 32,
|
| 23 |
+
"max_state_dim": 64,
|
| 24 |
+
"model_dtype": "float32",
|
| 25 |
+
"noise_beta_alpha": 1.5,
|
| 26 |
+
"noise_beta_beta": 1.0,
|
| 27 |
+
"noise_s": 0.999,
|
| 28 |
+
"num_inference_timesteps": 4,
|
| 29 |
+
"num_target_vision_tokens": 32,
|
| 30 |
+
"num_timestep_buckets": 1000,
|
| 31 |
+
"tune_diffusion_model": true,
|
| 32 |
+
"tune_projector": true,
|
| 33 |
+
"use_vlln": true,
|
| 34 |
+
"vl_self_attention_cfg": {
|
| 35 |
+
"attention_head_dim": 64,
|
| 36 |
+
"dropout": 0.2,
|
| 37 |
+
"final_dropout": true,
|
| 38 |
+
"num_attention_heads": 32,
|
| 39 |
+
"num_layers": 4,
|
| 40 |
+
"positional_embeddings": null
|
| 41 |
+
}
|
| 42 |
+
},
|
| 43 |
+
"action_horizon": 16,
|
| 44 |
+
"architectures": [
|
| 45 |
+
"GR00T_N1_5_RKD"
|
| 46 |
+
],
|
| 47 |
+
"attn_implementation": null,
|
| 48 |
+
"backbone_cfg": {
|
| 49 |
+
"eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops",
|
| 50 |
+
"load_bf16": false,
|
| 51 |
+
"project_to_dim": null,
|
| 52 |
+
"reproject_vision": false,
|
| 53 |
+
"select_layer": 12,
|
| 54 |
+
"tune_llm": false,
|
| 55 |
+
"tune_visual": true,
|
| 56 |
+
"use_flash_attention": true
|
| 57 |
+
},
|
| 58 |
+
"compute_dtype": "bfloat16",
|
| 59 |
+
"hidden_size": 2048,
|
| 60 |
+
"model_dtype": "float32",
|
| 61 |
+
"model_type": "gr00t_n1_5",
|
| 62 |
+
"rkd_action_encoder_cls_ffn_hidden_dim": 2048,
|
| 63 |
+
"rkd_action_encoder_cls_num_heads": 8,
|
| 64 |
+
"rkd_action_encoder_cls_num_layers": 1,
|
| 65 |
+
"rkd_action_encoder_projector_dim": 512,
|
| 66 |
+
"rkd_action_encoder_projector_enabled": false,
|
| 67 |
+
"rkd_action_encoder_projector_pooling": "flatten",
|
| 68 |
+
"rkd_action_temp": 0.1,
|
| 69 |
+
"rkd_angle_loss_weight": 2.0,
|
| 70 |
+
"rkd_distance_loss_weight": 1.0,
|
| 71 |
+
"rkd_dtw_adaptive_floor": 0.05,
|
| 72 |
+
"rkd_dtw_adaptive_mix_lambda": 0.5,
|
| 73 |
+
"rkd_dtw_base_weight": 1.0,
|
| 74 |
+
"rkd_dtw_gripper_weight": 0.5,
|
| 75 |
+
"rkd_dtw_mix_mode": "distance_mix",
|
| 76 |
+
"rkd_dtw_normalize_components": true,
|
| 77 |
+
"rkd_dtw_pos_weight": 1.0,
|
| 78 |
+
"rkd_dtw_rot_weight": 1.0,
|
| 79 |
+
"rkd_dtw_timing_log_interval": 1000,
|
| 80 |
+
"rkd_enabled": true,
|
| 81 |
+
"rkd_exclude_diagonal": true,
|
| 82 |
+
"rkd_fm_loss_weight": 1.0,
|
| 83 |
+
"rkd_loss_type": "distance_angle",
|
| 84 |
+
"rkd_loss_weight": 0.0,
|
| 85 |
+
"rkd_loss_weight_end": 0.0,
|
| 86 |
+
"rkd_loss_weight_schedule": "cosine_decay",
|
| 87 |
+
"rkd_loss_weight_start": 1.0,
|
| 88 |
+
"rkd_raw_action_projector_dim": 512,
|
| 89 |
+
"rkd_raw_action_projector_enabled": false,
|
| 90 |
+
"rkd_raw_action_projector_pooling": "flatten",
|
| 91 |
+
"rkd_relation_mode": "token_pair_mean",
|
| 92 |
+
"rkd_student_reconstruction_head_dim": 512,
|
| 93 |
+
"rkd_student_reconstruction_head_enabled": true,
|
| 94 |
+
"rkd_student_reconstruction_head_hidden_dim": 512,
|
| 95 |
+
"rkd_student_reconstruction_head_pooling": "attention",
|
| 96 |
+
"rkd_teacher_source": "timewarp_action",
|
| 97 |
+
"rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46",
|
| 98 |
+
"rkd_vlm_temp": 0.1,
|
| 99 |
+
"torch_dtype": "bfloat16",
|
| 100 |
+
"transformers_version": "4.51.3"
|
| 101 |
+
}
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml
CHANGED
|
@@ -9,6 +9,7 @@ training:
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
|
|
|
| 12 |
model:
|
| 13 |
rkd_teacher_source: timewarp_action
|
| 14 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
|
|
|
| 9 |
num_gpus: 1
|
| 10 |
batch_size: 128
|
| 11 |
seed: 42
|
| 12 |
+
report_to: wandb
|
| 13 |
model:
|
| 14 |
rkd_teacher_source: timewarp_action
|
| 15 |
rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46
|
TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_20260702_121805.log
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30)
|
| 2 |
+
[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40)
|
| 3 |
+
WARNING: mimicgen environments not imported since mimicgen is not installed!
|
| 4 |
+
/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1.
|
| 5 |
+
check_for_updates()
|
| 6 |
+
`use_fast` is set to `True` but the image processor class does not have a fast version. Falling back to the slow version.
|
| 7 |
+
|
| 8 |
+
==================================================
|
| 9 |
+
GR00T FINE-TUNING CONFIGURATION:
|
| 10 |
+
==================================================
|
| 11 |
+
config: /home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/Experiments_v2/DTW/TimewarpVAE/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml
|
| 12 |
+
dataset_soup: None
|
| 13 |
+
output_dir: /tmp/gr00t
|
| 14 |
+
output_root: None
|
| 15 |
+
data_config: panda_omron
|
| 16 |
+
batch_size: 128
|
| 17 |
+
max_steps: 300000
|
| 18 |
+
num_gpus: 1
|
| 19 |
+
save_steps: 20000
|
| 20 |
+
run_name: None
|
| 21 |
+
save_total_limit: 100
|
| 22 |
+
seed: 42
|
| 23 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 24 |
+
tune_llm: False
|
| 25 |
+
tune_visual: False
|
| 26 |
+
tune_projector: True
|
| 27 |
+
tune_diffusion_model: True
|
| 28 |
+
resume: False
|
| 29 |
+
learning_rate: 3e-05
|
| 30 |
+
weight_decay: 1e-05
|
| 31 |
+
warmup_ratio: 0.05
|
| 32 |
+
lora_rank: 0
|
| 33 |
+
lora_alpha: 16
|
| 34 |
+
lora_dropout: 0.1
|
| 35 |
+
lora_full_model: False
|
| 36 |
+
dataloader_num_workers: 8
|
| 37 |
+
report_to: wandb
|
| 38 |
+
embodiment_tag: new_embodiment
|
| 39 |
+
video_backend: opencv
|
| 40 |
+
balance_dataset_weights: True
|
| 41 |
+
balance_trajectory_weights: True
|
| 42 |
+
ds_weights_alpha: 0.4
|
| 43 |
+
==================================================
|
| 44 |
+
|
| 45 |
+
Using 1 GPUs
|
| 46 |
+
|
| 47 |
+
================================================================================
|
| 48 |
+
Starting sweep branch: default
|
| 49 |
+
Sweep vars: {}
|
| 50 |
+
================================================================================
|
| 51 |
+
|
| 52 |
+
--------------------------------------------------------------------------------
|
| 53 |
+
Running phase 1: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k
|
| 54 |
+
Policy type: groot_RKDv3
|
| 55 |
+
Base model path: nvidia/GR00T-N1.5-3B
|
| 56 |
+
Output dir: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k
|
| 57 |
+
Trainable preset: None
|
| 58 |
+
Policy overrides: {'rkd_teacher_source': 'timewarp_action', 'rkd_timewarp_vae_checkpoint': '${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46', 'rkd_relation_mode': 'token_pair_mean', 'rkd_loss_type': 'distance_angle', 'rkd_student_reconstruction_head_pooling': 'attention', 'rkd_student_reconstruction_head_enabled': True, 'rkd_student_reconstruction_head_dim': 512, 'rkd_student_reconstruction_head_hidden_dim': 512, 'rkd_distance_loss_weight': 1.0, 'rkd_angle_loss_weight': 2.0, 'rkd_exclude_diagonal': True, 'rkd_enabled': True, 'rkd_fm_loss_weight': 1.0, 'rkd_loss_weight': 1.0, 'rkd_loss_weight_schedule': 'cosine_decay', 'rkd_loss_weight_start': 1.0, 'rkd_loss_weight_end': 0.0}
|
| 59 |
+
--------------------------------------------------------------------------------
|
| 60 |
+
|
| 61 |
+
[{'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseBlenderLid/20250822/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseBlenderLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseFridge/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseFridge', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'CloseToasterOvenDoor', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CoffeeSetupMug/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'CoffeeSetupMug', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/NavigateKitchen/20250821/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'NavigateKitchen', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenCabinet/20250819/lerobot', 'horizon': 1050, 'filter_key': '100_demos', 'task': 'OpenCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenDrawer/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'OpenDrawer', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenStandMixerHead/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'OpenStandMixerHead', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceDrawerToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'PickPlaceSinkToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceToasterToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'PickPlaceCabinetToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnElectricKettle', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnMicrowave/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffMicrowave/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'OpenElectricKettleLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/SlideDishwasherRack/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'SlideDishwasherRack', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'TurnOnSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnSinkSpout/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnSinkSpout', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustWaterTemperature', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffStove/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'TurnOffStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnStove/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustToasterOvenTemperature', 'split': 'pretrain', 'source': 'human'}]
|
| 62 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 63 |
+
/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/data/transform/state_action.py:257: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor).
|
| 64 |
+
self.statistics[key] = torch.tensor(value)
|
| 65 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 66 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 67 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 68 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 69 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 70 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 71 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 72 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 73 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 74 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 75 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 76 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 77 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 78 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 79 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 80 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 81 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 82 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 83 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 84 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 85 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 86 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 87 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 88 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 89 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 90 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 91 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 92 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 93 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 94 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 95 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 96 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 97 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 98 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 99 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 100 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 101 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 102 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 103 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 104 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 105 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 106 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 107 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 108 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 109 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 110 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 111 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 112 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 113 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 114 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 115 |
+
dataset weights: [1. 0.88494669 0.76443286 0.83888607 0.73785464 1.00501644
|
| 116 |
+
0.80029956 0.66065525 0.83967491 0.83515002 0.95047759 0.86791692
|
| 117 |
+
0.88329477 0.78510653 0.64381843 0.67404766 0.69697218 0.60443684
|
| 118 |
+
0.78448103 0.83447486 0.61146215 0.64384058 0.79545025 0.93845856
|
| 119 |
+
0.75517122 0.7973985 ]
|
| 120 |
+
Loaded 26 datasets
|
| 121 |
+
Loading pretrained dual brain from nvidia/GR00T-N1.5-3B
|
| 122 |
+
Tune backbone vision tower: False
|
| 123 |
+
Tune backbone LLM: False
|
| 124 |
+
Tune action head projector: True
|
| 125 |
+
Tune action head DiT: True
|
| 126 |
+
|
| 127 |
+
Tune backbone llm: False
|
| 128 |
+
Tune backbone visual: True
|
| 129 |
+
Total number of DiT parameters: 550386688
|
| 130 |
+
Total number of SelfAttentionTransformer parameters: 201433088
|
| 131 |
+
Tune action head projector: True
|
| 132 |
+
Tune action head diffusion model: True
|
| 133 |
+
/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/torch/nn/modules/transformer.py:382: UserWarning: enable_nested_tensor is True, but self.use_nested_tensor is False because encoder_layer.norm_first was True
|
| 134 |
+
warnings.warn(
|
| 135 |
+
|
| 136 |
+
Some weights of GR00T_N1_5_RKD were not initialized from the model checkpoint at /home/ext_minje/.cache/huggingface/hub/models--nvidia--GR00T-N1.5-3B/snapshots/869830fc749c35f34771aa5209f923ac57e4564e and are newly initialized: ['rkd_action_encoder_cls_projector.cls_token', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight', 'rkd_action_encoder_cls_projector.in_proj.bias', 'rkd_action_encoder_cls_projector.in_proj.weight', 'rkd_action_encoder_cls_projector.norm.bias', 'rkd_action_encoder_cls_projector.norm.weight', 'rkd_action_encoder_cls_projector.position_embedding', 'rkd_action_encoder_projector.bias', 'rkd_action_encoder_projector.weight', 'rkd_raw_action_projector.bias', 'rkd_raw_action_projector.weight', 'rkd_student_mean_projector.bias', 'rkd_student_mean_projector.weight', 'rkd_student_reconstruction_head.attn_score.bias', 'rkd_student_reconstruction_head.attn_score.weight', 'rkd_student_reconstruction_head.out_proj.bias', 'rkd_student_reconstruction_head.out_proj.weight', 'rkd_student_reconstruction_head.token_mlp.0.bias', 'rkd_student_reconstruction_head.token_mlp.0.weight', 'rkd_student_reconstruction_head.token_mlp.2.bias', 'rkd_student_reconstruction_head.token_mlp.2.weight']
|
| 137 |
+
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
|
| 138 |
+
Tune backbone llm: False
|
| 139 |
+
Tune backbone visual: False
|
| 140 |
+
Warning: No backbone trainable parameters found.
|
| 141 |
+
Tune action head projector: True
|
| 142 |
+
Tune action head diffusion model: True
|
| 143 |
+
Detected uninitialized RKD aux parameters; resetting 5 affected projector/head module(s).
|
| 144 |
+
Trainable summary: {'trainable_preset': 'default', 'trainable_modules': 'action_head.future_tokens(49,152), action_head.vlln(4,096), action_head.vl_self_attention(201,433,088), action_head.state_encoder(52,510,720), action_head.action_encoder(228,212,736), action_head.action_decoder(34,636,800), action_head.position_embedding(1,572,864), action_head.model(550,386,688), rkd_student_reconstruction_head(1,574,913)', 'trainable_module_names': 'action_head.future_tokens, action_head.vlln, action_head.vl_self_attention, action_head.state_encoder, action_head.action_encoder, action_head.action_decoder, action_head.position_embedding, action_head.model, rkd_student_reconstruction_head', 'trainable_module_param_counts': {'action_head.future_tokens': 49152, 'action_head.vlln': 4096, 'action_head.vl_self_attention': 201433088, 'action_head.state_encoder': 52510720, 'action_head.action_encoder': 228212736, 'action_head.action_decoder': 34636800, 'action_head.position_embedding': 1572864, 'action_head.model': 550386688, 'rkd_student_reconstruction_head': 1574913}, 'trainable_param_count': 1070381057, 'total_param_count': 2743583169, 'trainable_param_ratio': 0.39013982484450793}
|
| 145 |
+
Run name: RKDv3_TimewarpVAE_DA__Student_APHead__single-phase_FM_RKD_cosine_decay_from_GR00T_base_to_120k_fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k
|
| 146 |
+
train dataloader length: 3437
|
| 147 |
+
train dataset length: 439854
|
| 148 |
+
GPU memory before training: 7.1490302085876465 GB
|
| 149 |
+
TensorBoard logs will be saved to: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/runs
|
| 150 |
+
Aux loss weight schedule: rkd_loss_weight 1.0 -> 0.0 (cosine_decay)
|
| 151 |
+
Traceback (most recent call last):
|
| 152 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1158, in <module>
|
| 153 |
+
run_yaml_experiment(
|
| 154 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1117, in run_yaml_experiment
|
| 155 |
+
main(args, policy_type=policy_type, policy_overrides=policy_overrides, trainable_preset=preset)
|
| 156 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 847, in main
|
| 157 |
+
experiment.train()
|
| 158 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/runner.py", line 172, in train
|
| 159 |
+
self.trainer.train(resume_from_checkpoint=self.resume_from_checkpoint)
|
| 160 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/trainer.py", line 176, in train
|
| 161 |
+
return super().train(resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs)
|
| 162 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 163 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2245, in train
|
| 164 |
+
return inner_training_loop(
|
| 165 |
+
^^^^^^^^^^^^^^^^^^^^
|
| 166 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2469, in _inner_training_loop
|
| 167 |
+
self.control = self.callback_handler.on_train_begin(args, self.state, self.control)
|
| 168 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 169 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin
|
| 170 |
+
return self.call_event("on_train_begin", args, state, control)
|
| 171 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 172 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 556, in call_event
|
| 173 |
+
result = getattr(callback, event)(
|
| 174 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 175 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin
|
| 176 |
+
self.setup(args, state, model, **kwargs)
|
| 177 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 857, in setup
|
| 178 |
+
self._wandb.init(
|
| 179 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1595, in init
|
| 180 |
+
get_sentry().reraise(e)
|
| 181 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/analytics/sentry.py", line 190, in reraise
|
| 182 |
+
raise exc.with_traceback(tb)
|
| 183 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1516, in init
|
| 184 |
+
wi.maybe_login(init_settings)
|
| 185 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 193, in maybe_login
|
| 186 |
+
wandb_login._login(
|
| 187 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 190, in _login
|
| 188 |
+
auth = _find_or_prompt_for_key(
|
| 189 |
+
^^^^^^^^^^^^^^^^^^^^^^^^
|
| 190 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 266, in _find_or_prompt_for_key
|
| 191 |
+
auth = wbauth.authenticate_session(
|
| 192 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 193 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/lib/wbauth/authenticate.py", line 139, in authenticate_session
|
| 194 |
+
raise UsageError(
|
| 195 |
+
wandb.errors.errors.UsageError: No API key configured. Use `wandb login` to log in.
|
TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_tensorboard_20260702_122542.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_20260702_121805.log
ADDED
|
@@ -0,0 +1,195 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30)
|
| 2 |
+
[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40)
|
| 3 |
+
WARNING: mimicgen environments not imported since mimicgen is not installed!
|
| 4 |
+
/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1.
|
| 5 |
+
check_for_updates()
|
| 6 |
+
`use_fast` is set to `True` but the image processor class does not have a fast version. Falling back to the slow version.
|
| 7 |
+
|
| 8 |
+
==================================================
|
| 9 |
+
GR00T FINE-TUNING CONFIGURATION:
|
| 10 |
+
==================================================
|
| 11 |
+
config: /home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/Experiments_v2/DTW/TimewarpVAE/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml
|
| 12 |
+
dataset_soup: None
|
| 13 |
+
output_dir: /tmp/gr00t
|
| 14 |
+
output_root: None
|
| 15 |
+
data_config: panda_omron
|
| 16 |
+
batch_size: 128
|
| 17 |
+
max_steps: 300000
|
| 18 |
+
num_gpus: 1
|
| 19 |
+
save_steps: 20000
|
| 20 |
+
run_name: None
|
| 21 |
+
save_total_limit: 100
|
| 22 |
+
seed: 42
|
| 23 |
+
base_model_path: nvidia/GR00T-N1.5-3B
|
| 24 |
+
tune_llm: False
|
| 25 |
+
tune_visual: False
|
| 26 |
+
tune_projector: True
|
| 27 |
+
tune_diffusion_model: True
|
| 28 |
+
resume: False
|
| 29 |
+
learning_rate: 3e-05
|
| 30 |
+
weight_decay: 1e-05
|
| 31 |
+
warmup_ratio: 0.05
|
| 32 |
+
lora_rank: 0
|
| 33 |
+
lora_alpha: 16
|
| 34 |
+
lora_dropout: 0.1
|
| 35 |
+
lora_full_model: False
|
| 36 |
+
dataloader_num_workers: 8
|
| 37 |
+
report_to: wandb
|
| 38 |
+
embodiment_tag: new_embodiment
|
| 39 |
+
video_backend: opencv
|
| 40 |
+
balance_dataset_weights: True
|
| 41 |
+
balance_trajectory_weights: True
|
| 42 |
+
ds_weights_alpha: 0.4
|
| 43 |
+
==================================================
|
| 44 |
+
|
| 45 |
+
Using 1 GPUs
|
| 46 |
+
|
| 47 |
+
================================================================================
|
| 48 |
+
Starting sweep branch: default
|
| 49 |
+
Sweep vars: {}
|
| 50 |
+
================================================================================
|
| 51 |
+
|
| 52 |
+
--------------------------------------------------------------------------------
|
| 53 |
+
Running phase 1: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k
|
| 54 |
+
Policy type: groot_RKDv3
|
| 55 |
+
Base model path: nvidia/GR00T-N1.5-3B
|
| 56 |
+
Output dir: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k
|
| 57 |
+
Trainable preset: None
|
| 58 |
+
Policy overrides: {'rkd_teacher_source': 'timewarp_action', 'rkd_timewarp_vae_checkpoint': '${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46', 'rkd_relation_mode': 'token_pair_mean', 'rkd_loss_type': 'distance_angle', 'rkd_student_reconstruction_head_pooling': 'attention', 'rkd_student_reconstruction_head_enabled': True, 'rkd_student_reconstruction_head_dim': 512, 'rkd_student_reconstruction_head_hidden_dim': 512, 'rkd_distance_loss_weight': 1.0, 'rkd_angle_loss_weight': 2.0, 'rkd_exclude_diagonal': True, 'rkd_enabled': True, 'rkd_fm_loss_weight': 1.0, 'rkd_loss_weight': 1.0, 'rkd_loss_weight_schedule': 'cosine_decay', 'rkd_loss_weight_start': 1.0, 'rkd_loss_weight_end': 0.0}
|
| 59 |
+
--------------------------------------------------------------------------------
|
| 60 |
+
|
| 61 |
+
[{'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseBlenderLid/20250822/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseBlenderLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseFridge/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseFridge', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'CloseToasterOvenDoor', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CoffeeSetupMug/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'CoffeeSetupMug', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/NavigateKitchen/20250821/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'NavigateKitchen', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenCabinet/20250819/lerobot', 'horizon': 1050, 'filter_key': '100_demos', 'task': 'OpenCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenDrawer/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'OpenDrawer', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenStandMixerHead/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'OpenStandMixerHead', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceDrawerToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'PickPlaceSinkToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceToasterToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'PickPlaceCabinetToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnElectricKettle', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnMicrowave/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffMicrowave/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'OpenElectricKettleLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/SlideDishwasherRack/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'SlideDishwasherRack', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'TurnOnSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnSinkSpout/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnSinkSpout', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustWaterTemperature', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffStove/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'TurnOffStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnStove/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustToasterOvenTemperature', 'split': 'pretrain', 'source': 'human'}]
|
| 62 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 63 |
+
/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/data/transform/state_action.py:257: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor).
|
| 64 |
+
self.statistics[key] = torch.tensor(value)
|
| 65 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 66 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 67 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 68 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 69 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 70 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 71 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 72 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 73 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 74 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 75 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 76 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 77 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 78 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 79 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 80 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 81 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 82 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 83 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 84 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 85 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 86 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 87 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 88 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 89 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 90 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 91 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 92 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 93 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 94 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 95 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 96 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 97 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 98 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 99 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 100 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 101 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 102 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 103 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 104 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 105 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 106 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 107 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 108 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 109 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 110 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 111 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 112 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 113 |
+
Using 100 subset demos for filter_key: 100_demos
|
| 114 |
+
Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT
|
| 115 |
+
dataset weights: [1. 0.88494669 0.76443286 0.83888607 0.73785464 1.00501644
|
| 116 |
+
0.80029956 0.66065525 0.83967491 0.83515002 0.95047759 0.86791692
|
| 117 |
+
0.88329477 0.78510653 0.64381843 0.67404766 0.69697218 0.60443684
|
| 118 |
+
0.78448103 0.83447486 0.61146215 0.64384058 0.79545025 0.93845856
|
| 119 |
+
0.75517122 0.7973985 ]
|
| 120 |
+
Loaded 26 datasets
|
| 121 |
+
Loading pretrained dual brain from nvidia/GR00T-N1.5-3B
|
| 122 |
+
Tune backbone vision tower: False
|
| 123 |
+
Tune backbone LLM: False
|
| 124 |
+
Tune action head projector: True
|
| 125 |
+
Tune action head DiT: True
|
| 126 |
+
|
| 127 |
+
Tune backbone llm: False
|
| 128 |
+
Tune backbone visual: True
|
| 129 |
+
Total number of DiT parameters: 550386688
|
| 130 |
+
Total number of SelfAttentionTransformer parameters: 201433088
|
| 131 |
+
Tune action head projector: True
|
| 132 |
+
Tune action head diffusion model: True
|
| 133 |
+
/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/torch/nn/modules/transformer.py:382: UserWarning: enable_nested_tensor is True, but self.use_nested_tensor is False because encoder_layer.norm_first was True
|
| 134 |
+
warnings.warn(
|
| 135 |
+
|
| 136 |
+
Some weights of GR00T_N1_5_RKD were not initialized from the model checkpoint at /home/ext_minje/.cache/huggingface/hub/models--nvidia--GR00T-N1.5-3B/snapshots/869830fc749c35f34771aa5209f923ac57e4564e and are newly initialized: ['rkd_action_encoder_cls_projector.cls_token', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias', 'rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight', 'rkd_action_encoder_cls_projector.in_proj.bias', 'rkd_action_encoder_cls_projector.in_proj.weight', 'rkd_action_encoder_cls_projector.norm.bias', 'rkd_action_encoder_cls_projector.norm.weight', 'rkd_action_encoder_cls_projector.position_embedding', 'rkd_action_encoder_projector.bias', 'rkd_action_encoder_projector.weight', 'rkd_raw_action_projector.bias', 'rkd_raw_action_projector.weight', 'rkd_student_mean_projector.bias', 'rkd_student_mean_projector.weight', 'rkd_student_reconstruction_head.attn_score.bias', 'rkd_student_reconstruction_head.attn_score.weight', 'rkd_student_reconstruction_head.out_proj.bias', 'rkd_student_reconstruction_head.out_proj.weight', 'rkd_student_reconstruction_head.token_mlp.0.bias', 'rkd_student_reconstruction_head.token_mlp.0.weight', 'rkd_student_reconstruction_head.token_mlp.2.bias', 'rkd_student_reconstruction_head.token_mlp.2.weight']
|
| 137 |
+
You should probably TRAIN this model on a down-stream task to be able to use it for predictions and inference.
|
| 138 |
+
Tune backbone llm: False
|
| 139 |
+
Tune backbone visual: False
|
| 140 |
+
Warning: No backbone trainable parameters found.
|
| 141 |
+
Tune action head projector: True
|
| 142 |
+
Tune action head diffusion model: True
|
| 143 |
+
Detected uninitialized RKD aux parameters; resetting 5 affected projector/head module(s).
|
| 144 |
+
Trainable summary: {'trainable_preset': 'default', 'trainable_modules': 'action_head.future_tokens(49,152), action_head.vlln(4,096), action_head.vl_self_attention(201,433,088), action_head.state_encoder(52,510,720), action_head.action_encoder(228,212,736), action_head.action_decoder(34,636,800), action_head.position_embedding(1,572,864), action_head.model(550,386,688), rkd_student_reconstruction_head(1,574,913)', 'trainable_module_names': 'action_head.future_tokens, action_head.vlln, action_head.vl_self_attention, action_head.state_encoder, action_head.action_encoder, action_head.action_decoder, action_head.position_embedding, action_head.model, rkd_student_reconstruction_head', 'trainable_module_param_counts': {'action_head.future_tokens': 49152, 'action_head.vlln': 4096, 'action_head.vl_self_attention': 201433088, 'action_head.state_encoder': 52510720, 'action_head.action_encoder': 228212736, 'action_head.action_decoder': 34636800, 'action_head.position_embedding': 1572864, 'action_head.model': 550386688, 'rkd_student_reconstruction_head': 1574913}, 'trainable_param_count': 1070381057, 'total_param_count': 2743583169, 'trainable_param_ratio': 0.39013982484450793}
|
| 145 |
+
Run name: RKDv3_TimewarpVAE_DA__Student_APHead__single-phase_FM_RKD_cosine_decay_from_GR00T_base_to_60k_fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k
|
| 146 |
+
train dataloader length: 3437
|
| 147 |
+
train dataset length: 439854
|
| 148 |
+
GPU memory before training: 7.1490302085876465 GB
|
| 149 |
+
TensorBoard logs will be saved to: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/runs
|
| 150 |
+
Aux loss weight schedule: rkd_loss_weight 1.0 -> 0.0 (cosine_decay)
|
| 151 |
+
Traceback (most recent call last):
|
| 152 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1158, in <module>
|
| 153 |
+
run_yaml_experiment(
|
| 154 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1117, in run_yaml_experiment
|
| 155 |
+
main(args, policy_type=policy_type, policy_overrides=policy_overrides, trainable_preset=preset)
|
| 156 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 847, in main
|
| 157 |
+
experiment.train()
|
| 158 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/runner.py", line 172, in train
|
| 159 |
+
self.trainer.train(resume_from_checkpoint=self.resume_from_checkpoint)
|
| 160 |
+
File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/trainer.py", line 176, in train
|
| 161 |
+
return super().train(resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs)
|
| 162 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 163 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2245, in train
|
| 164 |
+
return inner_training_loop(
|
| 165 |
+
^^^^^^^^^^^^^^^^^^^^
|
| 166 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2469, in _inner_training_loop
|
| 167 |
+
self.control = self.callback_handler.on_train_begin(args, self.state, self.control)
|
| 168 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 169 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin
|
| 170 |
+
return self.call_event("on_train_begin", args, state, control)
|
| 171 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 172 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 556, in call_event
|
| 173 |
+
result = getattr(callback, event)(
|
| 174 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 175 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin
|
| 176 |
+
self.setup(args, state, model, **kwargs)
|
| 177 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 857, in setup
|
| 178 |
+
self._wandb.init(
|
| 179 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1595, in init
|
| 180 |
+
get_sentry().reraise(e)
|
| 181 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/analytics/sentry.py", line 190, in reraise
|
| 182 |
+
raise exc.with_traceback(tb)
|
| 183 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1516, in init
|
| 184 |
+
wi.maybe_login(init_settings)
|
| 185 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 193, in maybe_login
|
| 186 |
+
wandb_login._login(
|
| 187 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 190, in _login
|
| 188 |
+
auth = _find_or_prompt_for_key(
|
| 189 |
+
^^^^^^^^^^^^^^^^^^^^^^^^
|
| 190 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 266, in _find_or_prompt_for_key
|
| 191 |
+
auth = wbauth.authenticate_session(
|
| 192 |
+
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
| 193 |
+
File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/lib/wbauth/authenticate.py", line 139, in authenticate_session
|
| 194 |
+
raise UsageError(
|
| 195 |
+
wandb.errors.errors.UsageError: No API key configured. Use `wandb login` to log in.
|
TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_tensorboard_20260702_122542.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k_gpu0_bs128_numgpu1.pid
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
346449
|
TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1.pid
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
1986170
|
TimewarpVAE/pids/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1.pid
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
1986169
|