Upload folder using huggingface_hub

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_10000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3496810b7e088ebbf3a9499d5f908de154e5c5535615fd2b971644839bc07910
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_15000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1c46a21ba90ad0a1d42830640f817396440e519ad1749f84e5dd41fc0f93773
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_20000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ef9b1b1d89d55bb21bdefce2d312c5891f93c110a692ab37dd52c1137c95492
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_25000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e61b9432e2bd7d371b20b4e8ef23aa88e55c06723b68278d91fa4bfb868f7fe
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_30000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d032d86a01e3d7a34aca50eee8d353700cfaa9d8e3e39627adfca1da6170c90e
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_35000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b61194d366f59d033c52d9d45df3a164d48dc2de03f71ac21256fed9b2fb15ef
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_40000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5de077c53050ef68f06eed06696ded8300cdb24f41b7bd639d6b52905e2eb1e6
+size 8604575530

intervla-m1/simple/G1WholebodyHandover/20260409_053020/checkpoints/steps_5000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b210d0f52e719aabadd09df6a20ffe8dc5c031d331a7f417f30f83bd2b9918ed
+size 8604574397

intervla-m1/simple/G1WholebodyHandover/20260409_053020/config.json

@@ -0,0 +1,151 @@
+{
+  "run_id": "G1WholebodyHandover",
+  "run_root_dir": "runs/InternVLA/Checkpoints",
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "wandb_entity": "jliu530-soochow-university",
+  "wandb_project": "psi",
+  "is_debug": false,
+  "framework": {
+    "framework_py": "InternVLA-M1",
+    "qwenvl": {
+      "base_vlm": "Qwen/Qwen2.5-VL-3B-Instruct",
+      "attn_implementation": "flash_attention_2",
+      "vl_hidden_dim": 2048
+    },
+    "dino": {
+      "dino_backbone": "dinov2_vits14"
+    },
+    "layer_qformer": {
+      "qformer_end_layer": 37,
+      "qformer_start_layer": 36,
+      "num_query_tokens": 64,
+      "input_dim": 2048,
+      "ouptput_dim": 768,
+      "grad_scale": 0.5
+    },
+    "action_model": {
+      "action_model_type": "DiT-B",
+      "action_hidden_dim": 768,
+      "action_dim": 36,
+      "use_ema": false,
+      "future_action_window_size": 15,
+      "past_action_window_size": 0,
+      "repeated_diffusion_steps": 8
+    },
+    "fm_head_config": {
+      "input_embedding_dim": 1536,
+      "hidden_size": 1024,
+      "add_pos_embed": true,
+      "max_seq_len": 1024,
+      "action_dim": 36,
+      "future_action_window_size": 15,
+      "action_horizon": 16,
+      "past_action_window_size": 0,
+      "noise_beta_alpha": 1.5,
+      "noise_beta_beta": 1.0,
+      "noise_s": 0.999,
+      "num_timestep_buckets": 1000,
+      "num_inference_timesteps": 4,
+      "num_target_vision_tokens": 32,
+      "diffusion_model_cfg": {
+        "attention_head_dim": 48,
+        "cross_attention_dim": 2048,
+        "dropout": 0.2,
+        "final_dropout": true,
+        "interleave_self_attention": true,
+        "norm_type": "ada_norm",
+        "num_attention_heads": 32,
+        "num_layers": 16,
+        "output_dim": 1024,
+        "positional_embeddings": null
+      }
+    }
+  },
+  "datasets": {
+    "vlm_data": {
+      "dataset_py": "vlm_datasets",
+      "dataformat": "llava_json",
+      "dataset_use": "asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en",
+      "eval_dataset": "aokvqa_cauldron_llava_format",
+      "data_flatten": false,
+      "base_interval": 2,
+      "max_pixels": 50176,
+      "min_pixels": 784,
+      "model_max_length": 2048,
+      "model_type": "qwen2.5vl",
+      "per_device_batch_size": 4
+    },
+    "vla_data": {
+      "dataset_py": "lerobot_datasets",
+      "data_root_dir": "/data/jliu/data/G1WholebodyHandover-v0",
+      "data_mix": "humanoid_",
+      "action_type": "abs_joints",
+      "CoT_prompt": "Your task is {instruction}. To identify the key objects for your task. Locate their bounding boxes in [x1,y1,x2,y2] format.",
+      "CoT_answer": "bbox",
+      "default_image_resolution": [
+        3,
+        224,
+        224
+      ],
+      "per_device_batch_size": 64,
+      "preload_all": true,
+      "load_all_data_for_training": true,
+      "obs": [
+        "image_0"
+      ],
+      "image_size": [
+        224,
+        224
+      ]
+    }
+  },
+  "trainer": {
+    "epochs": 100,
+    "max_train_steps": 40000,
+    "num_warmup_steps": 0,
+    "save_interval": 5000,
+    "eval_interval": 100,
+    "learning_rate": {
+      "base": 5e-05,
+      "qwen_vl_interface": 1e-05,
+      "action_model": 0.0001
+    },
+    "lr_scheduler_type": "cosine_with_min_lr",
+    "scheduler_specific_kwargs": {
+      "min_lr": 5e-07
+    },
+    "freeze_modules": "qwen_vl_interface,layer_qformer,dino_encoder,dino_pro",
+    "loss_scale": {
+      "vla": 1.0,
+      "vlm": 0.1
+    },
+    "pretrained_checkpoint": "/hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt",
+    "skip_reload_modules": "action_model",
+    "repeated_diffusion_steps": 4,
+    "max_grad_norm": 1.0,
+    "warmup_ratio": 0.1,
+    "weight_decay": 0.0,
+    "logging_frequency": 10,
+    "gradient_clipping": 1.0,
+    "gradient_accumulation_steps": 1,
+    "optimizer": {
+      "name": "AdamW",
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "weight_decay": 1e-08
+    },
+    "is_resume": false,
+    "resume_epoch": null,
+    "resume_step": null,
+    "enable_gradient_checkpointing": true,
+    "enable_mixed_precision_training": true
+  },
+  "output_dir": "runs/InternVLA/Checkpoints/G1WholebodyHandover/20260409_053020"
+}

intervla-m1/simple/G1WholebodyHandover/20260409_053020/config.yaml

@@ -0,0 +1,130 @@
+run_id: G1WholebodyHandover
+run_root_dir: runs/InternVLA/Checkpoints
+seed: 42
+trackers:
+- jsonl
+- wandb
+wandb_entity: jliu530-soochow-university
+wandb_project: psi
+is_debug: false
+framework:
+  framework_py: InternVLA-M1
+  qwenvl:
+    base_vlm: Qwen/Qwen2.5-VL-3B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  dino:
+    dino_backbone: dinov2_vits14
+  layer_qformer:
+    qformer_end_layer: 37
+    qformer_start_layer: 36
+    num_query_tokens: 64
+    input_dim: 2048
+    ouptput_dim: 768
+    grad_scale: 0.5
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 768
+    action_dim: 36
+    use_ema: false
+    future_action_window_size: 15
+    past_action_window_size: 0
+    repeated_diffusion_steps: 8
+  fm_head_config:
+    input_embedding_dim: 1536
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 36
+    future_action_window_size: 15
+    action_horizon: 16
+    past_action_window_size: 0
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      attention_head_dim: 48
+      cross_attention_dim: 2048
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_attention_heads: 32
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+datasets:
+  vlm_data:
+    dataset_py: vlm_datasets
+    dataformat: llava_json
+    dataset_use: asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en
+    eval_dataset: aokvqa_cauldron_llava_format
+    data_flatten: false
+    base_interval: 2
+    max_pixels: 50176
+    min_pixels: 784
+    model_max_length: 2048
+    model_type: qwen2.5vl
+    per_device_batch_size: 4
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /data/jliu/data/G1WholebodyHandover-v0
+    data_mix: humanoid_
+    action_type: abs_joints
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    CoT_answer: bbox
+    default_image_resolution:
+    - 3
+    - 224
+    - 224
+    per_device_batch_size: 64
+    preload_all: true
+    load_all_data_for_training: true
+    obs:
+    - image_0
+    image_size:
+    - 224
+    - 224
+trainer:
+  epochs: 100
+  max_train_steps: 40000
+  num_warmup_steps: 0
+  save_interval: 5000
+  eval_interval: 100
+  learning_rate:
+    base: 5.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: qwen_vl_interface,layer_qformer,dino_encoder,dino_pro
+  loss_scale:
+    vla: 1.0
+    vlm: 0.1
+  pretrained_checkpoint: /hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt
+  skip_reload_modules: action_model
+  repeated_diffusion_steps: 4
+  max_grad_norm: 1.0
+  warmup_ratio: 0.1
+  weight_decay: 0.0
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  is_resume: false
+  resume_epoch: null
+  resume_step: null
+  enable_gradient_checkpointing: true
+  enable_mixed_precision_training: true
+output_dir: runs/InternVLA/Checkpoints/G1WholebodyHandover/20260409_053020

intervla-m1/simple/G1WholebodyHandover/20260409_053020/dataset_statistics.json

@@ -0,0 +1,480 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        -0.217879056930542,
+        0.33495163917541504,
+        0.33495163917541504,
+        -0.30543649196624756,
+        -0.7176172733306885,
+        -0.6976513862609863,
+        -0.7176172733306885,
+        -0.16062557697296143,
+        -0.22490878403186798,
+        -0.22490878403186798,
+        0.48187702894210815,
+        0.48187702894210815,
+        0.19276303052902222,
+        0.48187702894210815,
+        -0.06166350468993187,
+        0.2401788830757141,
+        0.0873611643910408,
+        -0.09128011763095856,
+        -0.06374146789312363,
+        -0.03256600350141525,
+        -0.012831903994083405,
+        -0.09480268508195877,
+        -0.2054116129875183,
+        0.14932410418987274,
+        -0.1460893154144287,
+        -0.02532443404197693,
+        0.13397766649723053,
+        0.06257354468107224,
+        -0.000577143335249275,
+        0.02542431280016899,
+        -0.01923290081322193,
+        0.7401506304740906,
+        0.009205126203596592,
+        0.023015793412923813,
+        0.00010224639845546335,
+        0.0025475628208369017
+      ],
+      "std": [
+        0.2612758278846741,
+        0.349688321352005,
+        0.349688321352005,
+        0.3383631706237793,
+        0.7491674423217773,
+        0.7378979921340942,
+        0.7491674423217773,
+        0.23291678726673126,
+        0.3260721266269684,
+        0.3260721266269684,
+        0.6987663507461548,
+        0.6987663507461548,
+        0.279427170753479,
+        0.6987663507461548,
+        0.09250971674919128,
+        0.08452950417995453,
+        0.408634215593338,
+        0.1649845391511917,
+        0.19536836445331573,
+        0.14846064150333405,
+        0.26794546842575073,
+        0.08737793564796448,
+        0.024206371977925228,
+        0.14024904370307922,
+        0.24467714130878448,
+        0.19793805480003357,
+        0.18220646679401398,
+        0.17332760989665985,
+        0.025831403210759163,
+        0.04043002799153328,
+        0.07628294080495834,
+        0.00015065705521256548,
+        0.09775390475988388,
+        0.09573811292648315,
+        0.04227162525057793,
+        0.026238275691866875
+      ],
+      "max": [
+        0.5,
+        0.699999988079071,
+        0.699999988079071,
+        0.0,
+        2.2146225653890418e-16,
+        2.2146225653890418e-16,
+        2.2146225653890418e-16,
+        1.2266071310501902e-19,
+        1.1078670818917075e-16,
+        1.1078670818917075e-16,
+        1.5,
+        1.5,
+        0.6000000238418579,
+        1.5,
+        0.2472410947084427,
+        0.7092280983924866,
+        1.2571598291397095,
+        0.42311304807662964,
+        0.8564174771308899,
+        0.5002086162567139,
+        0.5172277092933655,
+        0.16140148043632507,
+        -0.1900009959936142,
+        0.5362864136695862,
+        0.5715147256851196,
+        0.5002322196960449,
+        0.566592276096344,
+        0.6392397880554199,
+        0.1580466777086258,
+        0.2233395129442215,
+        0.2582152187824249,
+        0.7400000095367432,
+        0.5,
+        0.5,
+        0.3454970121383667,
+        0.2899305522441864
+      ],
+      "min": [
+        -0.5,
+        -1.1095792134107943e-16,
+        -1.1095792134107943e-16,
+        -1.5,
+        -1.5,
+        -1.5,
+        -1.5,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        -2.2166350627321588e-16,
+        -2.2166350627321588e-16,
+        0.0,
+        -2.2166350627321588e-16,
+        -0.47567468881607056,
+        0.1900009959936142,
+        -0.512170135974884,
+        -0.6265152096748352,
+        -0.5008617043495178,
+        -0.8220608830451965,
+        -0.9223371148109436,
+        -0.49507391452789307,
+        -0.3437551259994507,
+        -0.6871383190155029,
+        -0.7637607455253601,
+        -0.7568023204803467,
+        -0.576077401638031,
+        -0.4588268995285034,
+        -0.13876836001873016,
+        -0.10360867530107498,
+        -0.47856518626213074,
+        0.7400000095367432,
+        -0.5,
+        -0.5,
+        -0.26161932945251465,
+        -0.06718750298023224
+      ],
+      "q01": [
+        -0.5,
+        0.0,
+        0.0,
+        -1.5,
+        -1.5,
+        -1.5,
+        -1.5,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.3209294053912163,
+        0.1900009959936142,
+        -0.38803558617830275,
+        -0.5046620488166809,
+        -0.38927449703216555,
+        -0.47090124636888503,
+        -0.7163057714700699,
+        -0.3420322224497795,
+        -0.29934623271226884,
+        -0.26393272846937177,
+        -0.6705281788110733,
+        -0.5924341869354248,
+        -0.40590299278497693,
+        -0.32045080602169035,
+        -0.07038286864757538,
+        -0.07046280093491078,
+        -0.2561952766776085,
+        0.7400000095367432,
+        -0.3524305522441864,
+        0.0,
+        -0.1027187518030405,
+        0.0
+      ],
+      "q99": [
+        0.5,
+        0.699999988079071,
+        0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.5,
+        1.5,
+        0.6000000238418579,
+        1.5,
+        0.14089947253465648,
+        0.580883502960205,
+        1.1797874009609222,
+        0.25774784147739405,
+        0.5666770941019057,
+        0.34745706409215893,
+        0.339419822692871,
+        0.08011209599673746,
+        -0.1900009959936142,
+        0.40124923735857004,
+        0.3766537192463873,
+        0.41124969720840454,
+        0.48994380980730057,
+        0.4618227949738502,
+        0.09533960297703735,
+        0.12665506854653355,
+        0.11710006609559051,
+        0.7400000095367432,
+        0.4913194477558136,
+        0.5,
+        0.10739764258265483,
+        0.12076389044523239
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "state": {
+      "mean": [
+        -0.2155175358057022,
+        -0.02219489961862564,
+        0.28812506794929504,
+        -0.1511061191558838,
+        -0.5701737403869629,
+        -0.21330925822257996,
+        -0.6113156676292419,
+        -0.17896264791488647,
+        -0.017788594588637352,
+        -0.20973540842533112,
+        0.1504911184310913,
+        0.4522649943828583,
+        0.16520608961582184,
+        0.4278402328491211,
+        -0.025824211537837982,
+        0.2251066416501999,
+        0.06837588548660278,
+        -0.006100596394389868,
+        -0.057470113039016724,
+        0.03059970773756504,
+        -0.014116800390183926,
+        -0.05716487765312195,
+        -0.1999409943819046,
+        0.16329504549503326,
+        -0.06425096839666367,
+        -0.031190501525998116,
+        0.18948617577552795,
+        0.06857924908399582,
+        -0.005298840347677469,
+        0.06115104258060455,
+        -0.01866256073117256,
+        0.7401506304740906
+      ],
+      "std": [
+        0.25221753120422363,
+        0.0899457111954689,
+        0.2992278039455414,
+        0.17489215731620789,
+        0.6001524329185486,
+        0.2285156548023224,
+        0.6428972482681274,
+        0.2302001267671585,
+        0.07764989882707596,
+        0.2707969546318054,
+        0.16321305930614471,
+        0.5522690415382385,
+        0.2336428165435791,
+        0.5221257209777832,
+        0.08843120187520981,
+        0.07218267768621445,
+        0.4016489088535309,
+        0.16209223866462708,
+        0.192921981215477,
+        0.14427520334720612,
+        0.26766741275787354,
+        0.08319389075040817,
+        0.02173873409628868,
+        0.13788215816020966,
+        0.24289271235466003,
+        0.1938791275024414,
+        0.17991188168525696,
+        0.174061119556427,
+        0.025640971958637238,
+        0.041976913809776306,
+        0.0752870962023735,
+        0.00015065705521256548
+      ],
+      "max": [
+        0.43566983938217163,
+        0.3739710748195648,
+        0.6575677990913391,
+        0.004060761071741581,
+        0.0005700877518393099,
+        0.0004725759499706328,
+        0.00010080631182063371,
+        1.310737025050912e-05,
+        0.21882089972496033,
+        0.0005271440604701638,
+        0.530737042427063,
+        1.4406861066818237,
+        1.4605127573013306,
+        1.4595911502838135,
+        0.2663630545139313,
+        0.657910943031311,
+        1.2515853643417358,
+        0.502498209476471,
+        0.8292973637580872,
+        0.5248894095420837,
+        0.4653257131576538,
+        0.18638382852077484,
+        -0.16696421802043915,
+        0.49318820238113403,
+        0.6363148093223572,
+        0.45773962140083313,
+        0.6238265037536621,
+        0.653800904750824,
+        0.1436084657907486,
+        0.25937986373901367,
+        0.26422709226608276,
+        0.7400000095367432
+      ],
+      "min": [
+        -0.5564982891082764,
+        -0.48307520151138306,
+        -0.0005447770818136632,
+        -0.8388738632202148,
+        -1.3970016241073608,
+        -0.8296014666557312,
+        -1.4599460363388062,
+        -0.5806806683540344,
+        -0.5149835348129272,
+        -0.6775947213172913,
+        -0.001480442238971591,
+        -0.0002713006397243589,
+        -0.000914653530344367,
+        -0.00019419840828049928,
+        -0.4206617772579193,
+        0.13972464203834534,
+        -0.546251654624939,
+        -0.5596316456794739,
+        -0.4764360189437866,
+        -0.7253566384315491,
+        -0.9443663954734802,
+        -0.4381798803806305,
+        -0.3338131606578827,
+        -0.667724072933197,
+        -0.6881827116012573,
+        -0.7544379830360413,
+        -0.5189417600631714,
+        -0.4484957158565521,
+        -0.13709338009357452,
+        -0.07360810041427612,
+        -0.4748336970806122,
+        0.7400000095367432
+      ],
+      "q01": [
+        -0.545208849310875,
+        -0.42749745190143584,
+        -0.0005246381351025775,
+        -0.6480066239833832,
+        -1.3562620949745179,
+        -0.7743040478229523,
+        -1.4221707606315612,
+        -0.5719072341918945,
+        -0.324733624458313,
+        -0.6709954166412353,
+        -1.057923989264964e-06,
+        -1.457349050326684e-07,
+        -1.5106486750937617e-06,
+        -2.423548727392699e-07,
+        -0.27424134463071825,
+        0.15975838720798494,
+        -0.40353597432374955,
+        -0.4082282695174217,
+        -0.3712728089094162,
+        -0.40305238008499145,
+        -0.7383889842033386,
+        -0.2908404359221458,
+        -0.2874874520301819,
+        -0.24262819081544876,
+        -0.6050335317850113,
+        -0.5883933693170548,
+        -0.33852073848247527,
+        -0.31927637457847596,
+        -0.07735681585967541,
+        -0.042391608729958535,
+        -0.25428820788860323,
+        0.7400000095367432
+      ],
+      "q99": [
+        0.40327705115079837,
+        0.2348814429342737,
+        0.6518104630708694,
+        0.00030119536590063946,
+        0.0005192926508607343,
+        0.0001505175937199965,
+        2.3595025995746256e-05,
+        5.185912332308361e-06,
+        0.08334404386579981,
+        0.00022139013104606418,
+        0.44970364242792127,
+        1.3850609183311462,
+        1.4514530301094055,
+        1.4533516800403594,
+        0.16941204354166983,
+        0.5226882100105286,
+        1.1606279826164243,
+        0.33347084760665896,
+        0.5583706372976303,
+        0.3959252551198003,
+        0.3333164182305336,
+        0.1102284654974937,
+        -0.1776231697201729,
+        0.40905793011188507,
+        0.44925396174192406,
+        0.3893393576145172,
+        0.5468983370065689,
+        0.474879567325115,
+        0.08610126286745068,
+        0.1693275338411331,
+        0.11660626158118245,
+        0.7400000095367432
+      ]
+    },
+    "num_transitions": 45530,
+    "num_trajectories": 100
+  }
+}

intervla-m1/simple/G1WholebodyHandover/20260409_053020/final_model/pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e208035472e5474bac095beed97fb98fabc9423207afb4ce872a77f68efb0418
+size 8604557774

intervla-m1/simple/G1WholebodyHandover/20260409_053020/summary.jsonl

@@ -0,0 +1,8 @@
+{"steps": 5000}
+{"steps": 10000}
+{"steps": 15000}
+{"steps": 20000}
+{"steps": 25000}
+{"steps": 30000}
+{"steps": 35000}
+{"steps": 40000}

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_10000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b22aeeb726c35af16a4b2648560d30867903090efb032011d21b73597620d0b
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_15000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ff18dfa42f23cffbe8bf20767cb45455f8e4ebaee7e2983f665081809ba5054
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_20000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d8b75f78f58f3529a74a8da5d91c29e7ab83a9e54f59bccc26407dfd649e58f
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_25000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:593e5f1f120302800b15e759aa5d3fcdcdd715adb08a38b9b8dc5b7c328f756b
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_30000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9bb2ca0788a6ffcc07c0c6f6b4a4212127f349b161cffc15aa3d26c7e5164555
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_35000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df7251b81bf2de426cd479000063b168734004441353b65c638703039ea295df
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_40000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ed48be8d9fdae3547ac2f4b9bedc3dc822e1bf72d72d65c3f62976f44fcaf8d
+size 8604575530

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/checkpoints/steps_5000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f1d99ec43b59b95843747d77328a8d4d2abd51115f47e3daa4cfdfda1e07b03
+size 8604574397

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/config.json

@@ -0,0 +1,151 @@
+{
+  "run_id": "G1WholebodyLocomotionPickBetweenTablesTeleop",
+  "run_root_dir": "runs/InternVLA/Checkpoints",
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "wandb_entity": "jliu530-soochow-university",
+  "wandb_project": "psi",
+  "is_debug": false,
+  "framework": {
+    "framework_py": "InternVLA-M1",
+    "qwenvl": {
+      "base_vlm": "Qwen/Qwen2.5-VL-3B-Instruct",
+      "attn_implementation": "flash_attention_2",
+      "vl_hidden_dim": 2048
+    },
+    "dino": {
+      "dino_backbone": "dinov2_vits14"
+    },
+    "layer_qformer": {
+      "qformer_end_layer": 37,
+      "qformer_start_layer": 36,
+      "num_query_tokens": 64,
+      "input_dim": 2048,
+      "ouptput_dim": 768,
+      "grad_scale": 0.5
+    },
+    "action_model": {
+      "action_model_type": "DiT-B",
+      "action_hidden_dim": 768,
+      "action_dim": 36,
+      "use_ema": false,
+      "future_action_window_size": 15,
+      "past_action_window_size": 0,
+      "repeated_diffusion_steps": 8
+    },
+    "fm_head_config": {
+      "input_embedding_dim": 1536,
+      "hidden_size": 1024,
+      "add_pos_embed": true,
+      "max_seq_len": 1024,
+      "action_dim": 36,
+      "future_action_window_size": 15,
+      "action_horizon": 16,
+      "past_action_window_size": 0,
+      "noise_beta_alpha": 1.5,
+      "noise_beta_beta": 1.0,
+      "noise_s": 0.999,
+      "num_timestep_buckets": 1000,
+      "num_inference_timesteps": 4,
+      "num_target_vision_tokens": 32,
+      "diffusion_model_cfg": {
+        "attention_head_dim": 48,
+        "cross_attention_dim": 2048,
+        "dropout": 0.2,
+        "final_dropout": true,
+        "interleave_self_attention": true,
+        "norm_type": "ada_norm",
+        "num_attention_heads": 32,
+        "num_layers": 16,
+        "output_dim": 1024,
+        "positional_embeddings": null
+      }
+    }
+  },
+  "datasets": {
+    "vlm_data": {
+      "dataset_py": "vlm_datasets",
+      "dataformat": "llava_json",
+      "dataset_use": "asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en",
+      "eval_dataset": "aokvqa_cauldron_llava_format",
+      "data_flatten": false,
+      "base_interval": 2,
+      "max_pixels": 50176,
+      "min_pixels": 784,
+      "model_max_length": 2048,
+      "model_type": "qwen2.5vl",
+      "per_device_batch_size": 4
+    },
+    "vla_data": {
+      "dataset_py": "lerobot_datasets",
+      "data_root_dir": "/data/jliu/data/G1WholebodyLocomotionPickBetweenTablesTeleop-v0",
+      "data_mix": "humanoid_",
+      "action_type": "abs_joints",
+      "CoT_prompt": "Your task is {instruction}. To identify the key objects for your task. Locate their bounding boxes in [x1,y1,x2,y2] format.",
+      "CoT_answer": "bbox",
+      "default_image_resolution": [
+        3,
+        224,
+        224
+      ],
+      "per_device_batch_size": 64,
+      "preload_all": true,
+      "load_all_data_for_training": true,
+      "obs": [
+        "image_0"
+      ],
+      "image_size": [
+        224,
+        224
+      ]
+    }
+  },
+  "trainer": {
+    "epochs": 100,
+    "max_train_steps": 40000,
+    "num_warmup_steps": 0,
+    "save_interval": 5000,
+    "eval_interval": 100,
+    "learning_rate": {
+      "base": 5e-05,
+      "qwen_vl_interface": 1e-05,
+      "action_model": 0.0001
+    },
+    "lr_scheduler_type": "cosine_with_min_lr",
+    "scheduler_specific_kwargs": {
+      "min_lr": 5e-07
+    },
+    "freeze_modules": "qwen_vl_interface,layer_qformer,dino_encoder,dino_pro",
+    "loss_scale": {
+      "vla": 1.0,
+      "vlm": 0.1
+    },
+    "pretrained_checkpoint": "/hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt",
+    "skip_reload_modules": "action_model",
+    "repeated_diffusion_steps": 4,
+    "max_grad_norm": 1.0,
+    "warmup_ratio": 0.1,
+    "weight_decay": 0.0,
+    "logging_frequency": 10,
+    "gradient_clipping": 1.0,
+    "gradient_accumulation_steps": 1,
+    "optimizer": {
+      "name": "AdamW",
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "weight_decay": 1e-08
+    },
+    "is_resume": false,
+    "resume_epoch": null,
+    "resume_step": null,
+    "enable_gradient_checkpointing": true,
+    "enable_mixed_precision_training": true
+  },
+  "output_dir": "runs/InternVLA/Checkpoints/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110"
+}

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/config.yaml

@@ -0,0 +1,130 @@
+run_id: G1WholebodyLocomotionPickBetweenTablesTeleop
+run_root_dir: runs/InternVLA/Checkpoints
+seed: 42
+trackers:
+- jsonl
+- wandb
+wandb_entity: jliu530-soochow-university
+wandb_project: psi
+is_debug: false
+framework:
+  framework_py: InternVLA-M1
+  qwenvl:
+    base_vlm: Qwen/Qwen2.5-VL-3B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  dino:
+    dino_backbone: dinov2_vits14
+  layer_qformer:
+    qformer_end_layer: 37
+    qformer_start_layer: 36
+    num_query_tokens: 64
+    input_dim: 2048
+    ouptput_dim: 768
+    grad_scale: 0.5
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 768
+    action_dim: 36
+    use_ema: false
+    future_action_window_size: 15
+    past_action_window_size: 0
+    repeated_diffusion_steps: 8
+  fm_head_config:
+    input_embedding_dim: 1536
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 36
+    future_action_window_size: 15
+    action_horizon: 16
+    past_action_window_size: 0
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      attention_head_dim: 48
+      cross_attention_dim: 2048
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_attention_heads: 32
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+datasets:
+  vlm_data:
+    dataset_py: vlm_datasets
+    dataformat: llava_json
+    dataset_use: asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en
+    eval_dataset: aokvqa_cauldron_llava_format
+    data_flatten: false
+    base_interval: 2
+    max_pixels: 50176
+    min_pixels: 784
+    model_max_length: 2048
+    model_type: qwen2.5vl
+    per_device_batch_size: 4
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /data/jliu/data/G1WholebodyLocomotionPickBetweenTablesTeleop-v0
+    data_mix: humanoid_
+    action_type: abs_joints
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    CoT_answer: bbox
+    default_image_resolution:
+    - 3
+    - 224
+    - 224
+    per_device_batch_size: 64
+    preload_all: true
+    load_all_data_for_training: true
+    obs:
+    - image_0
+    image_size:
+    - 224
+    - 224
+trainer:
+  epochs: 100
+  max_train_steps: 40000
+  num_warmup_steps: 0
+  save_interval: 5000
+  eval_interval: 100
+  learning_rate:
+    base: 5.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: qwen_vl_interface,layer_qformer,dino_encoder,dino_pro
+  loss_scale:
+    vla: 1.0
+    vlm: 0.1
+  pretrained_checkpoint: /hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt
+  skip_reload_modules: action_model
+  repeated_diffusion_steps: 4
+  max_grad_norm: 1.0
+  warmup_ratio: 0.1
+  weight_decay: 0.0
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  is_resume: false
+  resume_epoch: null
+  resume_step: null
+  enable_gradient_checkpointing: true
+  enable_mixed_precision_training: true
+output_dir: runs/InternVLA/Checkpoints/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/dataset_statistics.json

@@ -0,0 +1,480 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        0.0007652958738617599,
+        0.0010714149102568626,
+        0.0010714149102568626,
+        -0.0022958877962082624,
+        -0.0022958877962082624,
+        -0.0009183543152175844,
+        -0.0022958877962082624,
+        -0.4180760681629181,
+        -0.5853561758995056,
+        -0.5853561758995056,
+        1.254742980003357,
+        1.2550007104873657,
+        0.5019800662994385,
+        1.2550007104873657,
+        -0.06184714287519455,
+        0.19450722634792328,
+        -0.12890596687793732,
+        -0.0004327417991589755,
+        -0.11649200320243835,
+        -0.022241855040192604,
+        0.14975687861442566,
+        -0.2271433174610138,
+        -0.28912046551704407,
+        -0.12231507897377014,
+        -0.23417018353939056,
+        0.2349066436290741,
+        0.10561700165271759,
+        -0.03762618452310562,
+        -0.007439092267304659,
+        0.04567578062415123,
+        0.0297609381377697,
+        0.7396373152732849,
+        0.12372121214866638,
+        0.013200674206018448,
+        -0.2235538214445114,
+        0.4495043158531189
+      ],
+      "std": [
+        0.01948833279311657,
+        0.027283955365419388,
+        0.027283955365419388,
+        0.05846178159117699,
+        0.05846178159117699,
+        0.02337435446679592,
+        0.05846178159117699,
+        0.1841326802968979,
+        0.2574934661388397,
+        0.2574934661388397,
+        0.5518325567245483,
+        0.5516064763069153,
+        0.22092333436012268,
+        0.5516064763069153,
+        0.08286559581756592,
+        0.013414013199508157,
+        0.14786243438720703,
+        0.08076202124357224,
+        0.06461654603481293,
+        0.08045286685228348,
+        0.10112713277339935,
+        0.11457119882106781,
+        0.1058330088853836,
+        0.2477482408285141,
+        0.16770882904529572,
+        0.21804748475551605,
+        0.15766489505767822,
+        0.1760520040988922,
+        0.052497588098049164,
+        0.024612687528133392,
+        0.057495709508657455,
+        0.0003625280806534051,
+        0.2089398205280304,
+        0.08656619489192963,
+        0.3711760640144348,
+        2.2139852046966553
+      ],
+      "max": [
+        0.5,
+        0.699999988079071,
+        0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        6.930528109384597e-19,
+        1.108467615016421e-16,
+        1.108467615016421e-16,
+        1.5,
+        1.5,
+        1.0,
+        1.5,
+        0.2557959258556366,
+        0.35884979367256165,
+        0.5090755820274353,
+        0.19132143259048462,
+        0.21249642968177795,
+        0.3992660641670227,
+        0.4283020794391632,
+        0.1457289606332779,
+        -0.1900009959936142,
+        0.6150448322296143,
+        0.35468167066574097,
+        0.8703295588493347,
+        0.7531875371932983,
+        0.971237301826477,
+        0.13985762000083923,
+        0.15686897933483124,
+        0.4661160111427307,
+        0.7400000095367432,
+        0.5,
+        0.5,
+        1.0,
+        3.1414895057678223
+      ],
+      "min": [
+        0.0,
+        0.0,
+        0.0,
+        -1.5,
+        -1.5,
+        -0.6000000238418579,
+        -1.5,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        -2.216935230032842e-16,
+        -2.216935230032842e-16,
+        -4.0845591349633594e-18,
+        -2.216935230032842e-16,
+        -0.4883034825325012,
+        0.1900009959936142,
+        -0.5470856428146362,
+        -0.34318920969963074,
+        -0.35952919721603394,
+        -0.35302427411079407,
+        -0.4469815790653229,
+        -0.6371198296546936,
+        -0.7683824300765991,
+        -1.0653810501098633,
+        -0.8479154706001282,
+        -1.0297260284423828,
+        -0.42936205863952637,
+        -0.5147944092750549,
+        -0.16820405423641205,
+        -0.045328833162784576,
+        -0.13282617926597595,
+        0.7400000095367432,
+        -0.5,
+        -0.5,
+        -1.0,
+        -3.138223648071289
+      ],
+      "q01": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.30536221772432326,
+        0.1900009959936142,
+        -0.4495888948440552,
+        -0.23494456708431244,
+        -0.27246662437915803,
+        -0.2315385288000107,
+        -0.18485171496868133,
+        -0.4921060320734978,
+        -0.6122316139936447,
+        -0.7731428289413452,
+        -0.5696775823831558,
+        -0.4394104504585266,
+        -0.28442258715629576,
+        -0.3538160628080368,
+        -0.11734950572252273,
+        -0.015320314802229404,
+        -0.09260479986667633,
+        0.7400000095367432,
+        0.0,
+        -0.3567708432674408,
+        -1.0,
+        -3.124027729034424
+      ],
+      "q99": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.5,
+        1.5,
+        0.6000000238418579,
+        1.5,
+        0.12895929232239725,
+        0.26329000800848007,
+        0.2936864292621614,
+        0.14898110926151276,
+        0.06174419380724448,
+        0.2529342502355577,
+        0.3428380289673807,
+        -0.0013759000797290315,
+        -0.1900009959936142,
+        0.34379526853561404,
+        0.19408822178840665,
+        0.6508016681671143,
+        0.5123037415742882,
+        0.49184119641780855,
+        0.11030469514429586,
+        0.11221159264445316,
+        0.21934302642941514,
+        0.7400000095367432,
+        0.5,
+        0.35676584899425506,
+        0.19804689854383473,
+        3.1301991939544678
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "state": {
+      "mean": [
+        0.0007934353780001402,
+        0.0010695882374420762,
+        0.0012370680924504995,
+        -0.0009144614450633526,
+        -0.002343796193599701,
+        -0.002245287410914898,
+        -0.0023474614135921,
+        -0.43587779998779297,
+        0.009345951490104198,
+        -0.5239402651786804,
+        0.347458153963089,
+        1.1157641410827637,
+        0.3141929805278778,
+        1.061142086982727,
+        -0.02466614544391632,
+        0.18699687719345093,
+        -0.14352792501449585,
+        0.0824379175901413,
+        -0.1188783347606659,
+        0.038459114730358124,
+        0.14665654301643372,
+        -0.18040531873703003,
+        -0.2714536190032959,
+        -0.09729836881160736,
+        -0.149928018450737,
+        0.2267490029335022,
+        0.16860823333263397,
+        -0.03799568489193916,
+        -0.00529489666223526,
+        0.08189272880554199,
+        0.029461175203323364,
+        0.7396373152732849
+      ],
+      "std": [
+        0.01693989522755146,
+        0.02404647134244442,
+        0.02378660999238491,
+        0.017665348947048187,
+        0.04505283385515213,
+        0.045055001974105835,
+        0.04504767060279846,
+        0.19141361117362976,
+        0.08607088029384613,
+        0.22821539640426636,
+        0.15562357008457184,
+        0.497448593378067,
+        0.16223013401031494,
+        0.47353243827819824,
+        0.07905217260122299,
+        0.013432762585580316,
+        0.1468047797679901,
+        0.07942581176757812,
+        0.06510384380817413,
+        0.0791940838098526,
+        0.10252271592617035,
+        0.11002297699451447,
+        0.09499681740999222,
+        0.24013079702854156,
+        0.1692967414855957,
+        0.21809111535549164,
+        0.158147931098938,
+        0.17850598692893982,
+        0.050502315163612366,
+        0.023258700966835022,
+        0.058882467448711395,
+        0.0003625280806534051
+      ],
+      "max": [
+        0.47981399297714233,
+        0.6772664189338684,
+        0.6746510863304138,
+        0.0010172375477850437,
+        0.0007091082516126335,
+        0.001881288131698966,
+        0.0011398319620639086,
+        6.141255539660051e-07,
+        0.3043450713157654,
+        6.343479981296696e-07,
+        0.6933000087738037,
+        1.4612544775009155,
+        1.4651201963424683,
+        1.4609057903289795,
+        0.2809508740901947,
+        0.34028318524360657,
+        0.47627460956573486,
+        0.26476219296455383,
+        0.20825636386871338,
+        0.4566418528556824,
+        0.42864030599594116,
+        0.1656116098165512,
+        -0.1549365073442459,
+        0.5154499411582947,
+        0.4242899715900421,
+        0.8548054695129395,
+        0.8040095567703247,
+        0.9811649322509766,
+        0.136736661195755,
+        0.195722296833992,
+        0.45781663060188293,
+        0.7400000095367432
+      ],
+      "min": [
+        -0.02442001923918724,
+        -0.0517612099647522,
+        -0.0006534014828503132,
+        -0.5095356106758118,
+        -1.323034405708313,
+        -1.3221508264541626,
+        -1.3230019807815552,
+        -0.5770347714424133,
+        -0.4338151812553406,
+        -0.6721642017364502,
+        -0.0017213862156495452,
+        -7.534810038123396e-07,
+        -0.001927333534695208,
+        -1.075333216249419e-06,
+        -0.43650975823402405,
+        0.15721464157104492,
+        -0.5489339232444763,
+        -0.2632291913032532,
+        -0.3508843183517456,
+        -0.23784859478473663,
+        -0.4281824827194214,
+        -0.5803383588790894,
+        -0.7118590474128723,
+        -1.0344431400299072,
+        -0.7932196259498596,
+        -1.0205217599868774,
+        -0.3445618450641632,
+        -0.5986371040344238,
+        -0.13537253439426422,
+        -0.0017330688424408436,
+        -0.1421850621700287,
+        0.7400000095367432
+      ],
+      "q01": [
+        -2.9736981127825855e-06,
+        -0.002589050980750471,
+        -2.1223586691121456e-06,
+        -0.00027220559131819756,
+        -5.2740163209819e-06,
+        -4.3062968397862275e-05,
+        6.949997242600148e-08,
+        -0.5698864543437958,
+        -0.3627366861701012,
+        -0.6697060906887055,
+        -1.2418152664395165e-06,
+        -1.287923055315332e-07,
+        -7.915375590528129e-07,
+        -1.582540755862283e-07,
+        -0.2560530769824982,
+        0.16664464086294173,
+        -0.4414859291911125,
+        -0.14972542390227317,
+        -0.2697399368882179,
+        -0.16884329065680503,
+        -0.18771703973412515,
+        -0.4346155697107315,
+        -0.5672112548351288,
+        -0.739670946598053,
+        -0.49567418187856677,
+        -0.4431849017739296,
+        -0.22463233456015586,
+        -0.3607582712173462,
+        -0.10682432219386101,
+        0.02814220966771245,
+        -0.0969948647916317,
+        0.7400000095367432
+      ],
+      "q99": [
+        3.4415612958582677e-06,
+        0.019452356398105622,
+        0.01667371392250061,
+        0.0005855054722633213,
+        0.0003600666584679857,
+        0.0006024087872356176,
+        0.0003732459741877392,
+        2.2463960860363857e-07,
+        0.15193727359175682,
+        5.1716865669959736e-08,
+        0.5587792527675629,
+        1.4585348367691042,
+        0.5443805891275406,
+        1.370901610851288,
+        0.1576733058691028,
+        0.24011337146163012,
+        0.27520660489797616,
+        0.2312955512106419,
+        0.05981415051967266,
+        0.3088961178064349,
+        0.3407434976100922,
+        0.030774814467877155,
+        -0.17435445189476012,
+        0.3570790392160416,
+        0.2687951233983051,
+        0.6410461419820785,
+        0.5765821474790573,
+        0.49433160990476627,
+        0.09976441204547885,
+        0.14455557838082314,
+        0.21845388084650053,
+        0.7400000095367432
+      ]
+    },
+    "num_transitions": 62764,
+    "num_trajectories": 99
+  }
+}

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/final_model/pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1109463a97d234bf908d94e5f4aefc67cc4733b29a65514b4db1ca4e43ff26a6
+size 8604557774

intervla-m1/simple/G1WholebodyLocomotionPickBetweenTablesTeleop/20260409_052110/summary.jsonl

@@ -0,0 +1,8 @@
+{"steps": 5000}
+{"steps": 10000}
+{"steps": 15000}
+{"steps": 20000}
+{"steps": 25000}
+{"steps": 30000}
+{"steps": 35000}
+{"steps": 40000}

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_10000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a4c1423afa3642772d54f313d495edc8b36c26cb7369aa8a28a6efe77388975
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_15000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1e3a356b4abbfa51d31a369075f8634f7136034f1bf6f3364846ec40f39d663
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_20000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc6bb69676e3b2321d66dbc1c2318bb31e9d32036e6d7c1bf2d53862a7ddf355
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_25000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eaf44ddba127eb83c151bb7e347080765d7b6ed3b548ed664471f7b8919fad6
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_30000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b802f0d6833b5a293bf1784129a017652841862c3189cd4860d55bc9e198f52
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_35000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95b62139bca7bb8cdc48c5ce7530154d737292045eeb416d0b9f442a6288e51b
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_40000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56c78ff6e3c141e32bb727c439373da5992fd5524c804611f5b934c6e5f42d5d
+size 8604575530

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/checkpoints/steps_5000_pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01afd6ef0afc4a55e55b0f021b6a9f27b609ab0c526eaab0aff52488d38ba394
+size 8604574397

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/config.json

@@ -0,0 +1,151 @@
+{
+  "run_id": "G1WholebodyXMoveBendPickTeleop",
+  "run_root_dir": "runs/InternVLA/Checkpoints",
+  "seed": 42,
+  "trackers": [
+    "jsonl",
+    "wandb"
+  ],
+  "wandb_entity": "jliu530-soochow-university",
+  "wandb_project": "psi",
+  "is_debug": false,
+  "framework": {
+    "framework_py": "InternVLA-M1",
+    "qwenvl": {
+      "base_vlm": "Qwen/Qwen2.5-VL-3B-Instruct",
+      "attn_implementation": "flash_attention_2",
+      "vl_hidden_dim": 2048
+    },
+    "dino": {
+      "dino_backbone": "dinov2_vits14"
+    },
+    "layer_qformer": {
+      "qformer_end_layer": 37,
+      "qformer_start_layer": 36,
+      "num_query_tokens": 64,
+      "input_dim": 2048,
+      "ouptput_dim": 768,
+      "grad_scale": 0.5
+    },
+    "action_model": {
+      "action_model_type": "DiT-B",
+      "action_hidden_dim": 768,
+      "action_dim": 36,
+      "use_ema": false,
+      "future_action_window_size": 15,
+      "past_action_window_size": 0,
+      "repeated_diffusion_steps": 8
+    },
+    "fm_head_config": {
+      "input_embedding_dim": 1536,
+      "hidden_size": 1024,
+      "add_pos_embed": true,
+      "max_seq_len": 1024,
+      "action_dim": 36,
+      "future_action_window_size": 15,
+      "action_horizon": 16,
+      "past_action_window_size": 0,
+      "noise_beta_alpha": 1.5,
+      "noise_beta_beta": 1.0,
+      "noise_s": 0.999,
+      "num_timestep_buckets": 1000,
+      "num_inference_timesteps": 4,
+      "num_target_vision_tokens": 32,
+      "diffusion_model_cfg": {
+        "attention_head_dim": 48,
+        "cross_attention_dim": 2048,
+        "dropout": 0.2,
+        "final_dropout": true,
+        "interleave_self_attention": true,
+        "norm_type": "ada_norm",
+        "num_attention_heads": 32,
+        "num_layers": 16,
+        "output_dim": 1024,
+        "positional_embeddings": null
+      }
+    }
+  },
+  "datasets": {
+    "vlm_data": {
+      "dataset_py": "vlm_datasets",
+      "dataformat": "llava_json",
+      "dataset_use": "asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en",
+      "eval_dataset": "aokvqa_cauldron_llava_format",
+      "data_flatten": false,
+      "base_interval": 2,
+      "max_pixels": 50176,
+      "min_pixels": 784,
+      "model_max_length": 2048,
+      "model_type": "qwen2.5vl",
+      "per_device_batch_size": 4
+    },
+    "vla_data": {
+      "dataset_py": "lerobot_datasets",
+      "data_root_dir": "/data/jliu/data/G1WholebodyXMoveBendPickTeleop-v0",
+      "data_mix": "humanoid_",
+      "action_type": "abs_joints",
+      "CoT_prompt": "Your task is {instruction}. To identify the key objects for your task. Locate their bounding boxes in [x1,y1,x2,y2] format.",
+      "CoT_answer": "bbox",
+      "default_image_resolution": [
+        3,
+        224,
+        224
+      ],
+      "per_device_batch_size": 64,
+      "preload_all": true,
+      "load_all_data_for_training": true,
+      "obs": [
+        "image_0"
+      ],
+      "image_size": [
+        224,
+        224
+      ]
+    }
+  },
+  "trainer": {
+    "epochs": 100,
+    "max_train_steps": 40000,
+    "num_warmup_steps": 0,
+    "save_interval": 5000,
+    "eval_interval": 100,
+    "learning_rate": {
+      "base": 5e-05,
+      "qwen_vl_interface": 1e-05,
+      "action_model": 0.0001
+    },
+    "lr_scheduler_type": "cosine_with_min_lr",
+    "scheduler_specific_kwargs": {
+      "min_lr": 5e-07
+    },
+    "freeze_modules": "qwen_vl_interface,layer_qformer,dino_encoder,dino_pro",
+    "loss_scale": {
+      "vla": 1.0,
+      "vlm": 0.1
+    },
+    "pretrained_checkpoint": "/hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt",
+    "skip_reload_modules": "action_model",
+    "repeated_diffusion_steps": 4,
+    "max_grad_norm": 1.0,
+    "warmup_ratio": 0.1,
+    "weight_decay": 0.0,
+    "logging_frequency": 10,
+    "gradient_clipping": 1.0,
+    "gradient_accumulation_steps": 1,
+    "optimizer": {
+      "name": "AdamW",
+      "betas": [
+        0.9,
+        0.95
+      ],
+      "eps": 1e-08,
+      "weight_decay": 1e-08
+    },
+    "is_resume": false,
+    "resume_epoch": null,
+    "resume_step": null,
+    "enable_gradient_checkpointing": true,
+    "enable_mixed_precision_training": true
+  },
+  "output_dir": "runs/InternVLA/Checkpoints/G1WholebodyXMoveBendPickTeleop/20260403_151218"
+}

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/config.yaml

@@ -0,0 +1,130 @@
+run_id: G1WholebodyXMoveBendPickTeleop
+run_root_dir: runs/InternVLA/Checkpoints
+seed: 42
+trackers:
+- jsonl
+- wandb
+wandb_entity: jliu530-soochow-university
+wandb_project: psi
+is_debug: false
+framework:
+  framework_py: InternVLA-M1
+  qwenvl:
+    base_vlm: Qwen/Qwen2.5-VL-3B-Instruct
+    attn_implementation: flash_attention_2
+    vl_hidden_dim: 2048
+  dino:
+    dino_backbone: dinov2_vits14
+  layer_qformer:
+    qformer_end_layer: 37
+    qformer_start_layer: 36
+    num_query_tokens: 64
+    input_dim: 2048
+    ouptput_dim: 768
+    grad_scale: 0.5
+  action_model:
+    action_model_type: DiT-B
+    action_hidden_dim: 768
+    action_dim: 36
+    use_ema: false
+    future_action_window_size: 15
+    past_action_window_size: 0
+    repeated_diffusion_steps: 8
+  fm_head_config:
+    input_embedding_dim: 1536
+    hidden_size: 1024
+    add_pos_embed: true
+    max_seq_len: 1024
+    action_dim: 36
+    future_action_window_size: 15
+    action_horizon: 16
+    past_action_window_size: 0
+    noise_beta_alpha: 1.5
+    noise_beta_beta: 1.0
+    noise_s: 0.999
+    num_timestep_buckets: 1000
+    num_inference_timesteps: 4
+    num_target_vision_tokens: 32
+    diffusion_model_cfg:
+      attention_head_dim: 48
+      cross_attention_dim: 2048
+      dropout: 0.2
+      final_dropout: true
+      interleave_self_attention: true
+      norm_type: ada_norm
+      num_attention_heads: 32
+      num_layers: 16
+      output_dim: 1024
+      positional_embeddings: null
+datasets:
+  vlm_data:
+    dataset_py: vlm_datasets
+    dataformat: llava_json
+    dataset_use: asv2_conversation_en,asv2_detailed_description_en,asv2_region_captioning_en,coco_internvl_longcap_en,coco_karpathy_train_567_en,coco_negative_gpt4o_en,coco_poetry_zh,coco_rem_en_zh,cocorem_exist_yorn_en,cocotextv2_en,cocotextv2_gpt4o_en,okvqa_en,refcoco_grounding_aug_en,refcoco_grounding_en,tallyqa_coco_en,toloka_grounding_aug_en,vqav2_en,vsr_en
+    eval_dataset: aokvqa_cauldron_llava_format
+    data_flatten: false
+    base_interval: 2
+    max_pixels: 50176
+    min_pixels: 784
+    model_max_length: 2048
+    model_type: qwen2.5vl
+    per_device_batch_size: 4
+  vla_data:
+    dataset_py: lerobot_datasets
+    data_root_dir: /data/jliu/data/G1WholebodyXMoveBendPickTeleop-v0
+    data_mix: humanoid_
+    action_type: abs_joints
+    CoT_prompt: Your task is {instruction}. To identify the key objects for your task.
+      Locate their bounding boxes in [x1,y1,x2,y2] format.
+    CoT_answer: bbox
+    default_image_resolution:
+    - 3
+    - 224
+    - 224
+    per_device_batch_size: 64
+    preload_all: true
+    load_all_data_for_training: true
+    obs:
+    - image_0
+    image_size:
+    - 224
+    - 224
+trainer:
+  epochs: 100
+  max_train_steps: 40000
+  num_warmup_steps: 0
+  save_interval: 5000
+  eval_interval: 100
+  learning_rate:
+    base: 5.0e-05
+    qwen_vl_interface: 1.0e-05
+    action_model: 0.0001
+  lr_scheduler_type: cosine_with_min_lr
+  scheduler_specific_kwargs:
+    min_lr: 5.0e-07
+  freeze_modules: qwen_vl_interface,layer_qformer,dino_encoder,dino_pro
+  loss_scale:
+    vla: 1.0
+    vlm: 0.1
+  pretrained_checkpoint: /hfm/cache/checkpoints/InternVLA-M1-Pretrain-RT-1-Bridge/checkpoints/steps_50000_pytorch_model.pt
+  skip_reload_modules: action_model
+  repeated_diffusion_steps: 4
+  max_grad_norm: 1.0
+  warmup_ratio: 0.1
+  weight_decay: 0.0
+  logging_frequency: 10
+  gradient_clipping: 1.0
+  gradient_accumulation_steps: 1
+  optimizer:
+    name: AdamW
+    betas:
+    - 0.9
+    - 0.95
+    eps: 1.0e-08
+    weight_decay: 1.0e-08
+  is_resume: false
+  resume_epoch: null
+  resume_step: null
+  enable_gradient_checkpointing: true
+  enable_mixed_precision_training: true
+output_dir: runs/InternVLA/Checkpoints/G1WholebodyXMoveBendPickTeleop/20260403_151218

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/dataset_statistics.json

@@ -0,0 +1,480 @@
+{
+  "new_embodiment": {
+    "action": {
+      "mean": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.06578648090362549,
+        -0.0920981913805008,
+        -0.0920981913805008,
+        0.19735944271087646,
+        0.19735944271087646,
+        0.07894159853458405,
+        0.19735944271087646,
+        -0.08300793915987015,
+        0.19217251241207123,
+        -0.07867422699928284,
+        0.051928430795669556,
+        -0.20575666427612305,
+        0.026639869436621666,
+        0.2688467502593994,
+        -0.07158240675926208,
+        -0.1959005445241928,
+        -0.044735364615917206,
+        0.09095965325832367,
+        0.14965403079986572,
+        -0.03834051638841629,
+        -0.059758421033620834,
+        -0.0003535658761393279,
+        0.02151256427168846,
+        0.008055430836975574,
+        0.6770350337028503,
+        0.09653493016958237,
+        -0.0004846698429901153,
+        0.014787894673645496,
+        0.0
+      ],
+      "std": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.16896027326583862,
+        0.2365109622478485,
+        0.2365109622478485,
+        0.5068987011909485,
+        0.5068987011909485,
+        0.2027282416820526,
+        0.5068987011909485,
+        0.04981342703104019,
+        0.009523554705083325,
+        0.17949867248535156,
+        0.055632513016462326,
+        0.057273078709840775,
+        0.058521512895822525,
+        0.11143220216035843,
+        0.07724548876285553,
+        0.023301944136619568,
+        0.10750990360975266,
+        0.06850353628396988,
+        0.07305468618869781,
+        0.06640303879976273,
+        0.16941164433956146,
+        0.041646938771009445,
+        0.0699257031083107,
+        0.027388552203774452,
+        0.08091399818658829,
+        0.11833422631025314,
+        0.017973335459828377,
+        0.03434111177921295,
+        0.0
+      ],
+      "max": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.5,
+        1.5,
+        0.6000000238418579,
+        1.5,
+        0.06430592387914658,
+        0.2996276319026947,
+        0.5128592252731323,
+        0.3017215132713318,
+        -0.038698144257068634,
+        0.21968720853328705,
+        0.827497124671936,
+        0.044904597103595734,
+        -0.1900009959936142,
+        0.413065105676651,
+        0.41873428225517273,
+        0.6618388891220093,
+        0.4026392698287964,
+        0.8194853663444519,
+        0.12383800745010376,
+        0.16346246004104614,
+        0.15494900941848755,
+        0.7400000095367432,
+        0.5,
+        0.21786384284496307,
+        0.1749052256345749,
+        0.0
+      ],
+      "min": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.3314070701599121,
+        0.1900009959936142,
+        -0.8766500353813171,
+        -0.12303244322538376,
+        -0.4908517599105835,
+        -0.2786784768104553,
+        -0.022629141807556152,
+        -0.6784858703613281,
+        -0.5865002870559692,
+        -0.645729660987854,
+        -0.3608185946941376,
+        -0.15172408521175385,
+        -0.4648345112800598,
+        -0.2964947521686554,
+        -0.10700750350952148,
+        -0.21067920327186584,
+        -0.08102670311927795,
+        0.44999998807907104,
+        -0.5,
+        -0.26561295986175537,
+        -0.11697302013635635,
+        0.0
+      ],
+      "q01": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.5,
+        -0.699999988079071,
+        -0.699999988079071,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        -0.2336725726723671,
+        0.1900009959936142,
+        -0.628720715045929,
+        -0.08062581032514572,
+        -0.3772744107246399,
+        -0.14941381871700288,
+        0.023908816780894994,
+        -0.3680631712079048,
+        -0.3140790224075317,
+        -0.33936198383569716,
+        -0.1444373431801796,
+        -0.024716479536145926,
+        -0.25120449274778367,
+        -0.24746618106961252,
+        -0.0921607768535614,
+        -0.1331048083305359,
+        -0.055714426785707476,
+        0.5099999904632568,
+        0.0,
+        -0.08334636241197586,
+        -0.058656642064452175,
+        0.0
+      ],
+      "q99": [
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        0.0,
+        1.5,
+        1.5,
+        0.6000000238418579,
+        1.5,
+        0.005833799573592792,
+        0.24212055698037147,
+        0.31104624718427587,
+        0.2240664350986478,
+        -0.07993344962596893,
+        0.1509539039433002,
+        0.6122507166862484,
+        0.020442928690463276,
+        -0.1900009959936142,
+        0.20298720359802205,
+        0.26972131878137506,
+        0.33771990299224813,
+        0.14230648443102825,
+        0.5476555949449537,
+        0.10829514846205711,
+        0.10879010632634158,
+        0.09789865501224988,
+        0.7400000095367432,
+        0.4088541567325592,
+        0.057279629707336424,
+        0.11845016352832313,
+        0.0
+      ],
+      "mask": [
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true,
+        true
+      ]
+    },
+    "state": {
+      "mean": [
+        1.6478608813486062e-05,
+        -4.823089329875074e-05,
+        -1.5274658835551236e-06,
+        2.0739900719490834e-05,
+        -4.8923579015536234e-05,
+        1.184017673949711e-05,
+        1.8281939446751494e-06,
+        -0.04951467365026474,
+        -0.03387488052248955,
+        -0.05985373631119728,
+        0.034982677549123764,
+        0.09988676756620407,
+        0.05889609828591347,
+        0.09540130198001862,
+        -0.04761885479092598,
+        0.18126359581947327,
+        -0.08981631696224213,
+        0.1303543597459793,
+        -0.20870409905910492,
+        0.08360962569713593,
+        0.26360899209976196,
+        -0.033612482249736786,
+        -0.1811068058013916,
+        -0.027028528973460197,
+        0.1747075915336609,
+        0.1506245732307434,
+        0.02000368759036064,
+        -0.07187763601541519,
+        0.0014201127924025059,
+        0.06093015894293785,
+        0.007754191290587187,
+        0.6778029799461365
+      ],
+      "std": [
+        0.0003944706986658275,
+        0.0007630966720171273,
+        2.2961552531342022e-05,
+        0.000179155234945938,
+        0.0010678438702598214,
+        4.7558256483171135e-05,
+        1.1183346941834316e-05,
+        0.12474565207958221,
+        0.07307292520999908,
+        0.14443156123161316,
+        0.09493549168109894,
+        0.2512502074241638,
+        0.1241452693939209,
+        0.23682793974876404,
+        0.049122974276542664,
+        0.0104843201115727,
+        0.1711176037788391,
+        0.05156445503234863,
+        0.05493027716875076,
+        0.05776740238070488,
+        0.10714928060770035,
+        0.06816332787275314,
+        0.01825851947069168,
+        0.10530710965394974,
+        0.059736523777246475,
+        0.06796342134475708,
+        0.05937612056732178,
+        0.1552959680557251,
+        0.03811460733413696,
+        0.06448719650506973,
+        0.028375638648867607,
+        0.08062339574098587
+      ],
+      "max": [
+        0.013749510049819946,
+        0.0003444451722316444,
+        5.732499630539678e-06,
+        0.0019246992887929082,
+        0.0014607172925025225,
+        0.0007710650679655373,
+        0.0006001993897370994,
+        4.888642592959513e-07,
+        0.06670719385147095,
+        1.4086220971876173e-06,
+        0.43387407064437866,
+        1.2414171695709229,
+        0.6964682936668396,
+        1.2072811126708984,
+        0.0866343304514885,
+        0.2651435434818268,
+        0.49075624346733093,
+        0.34916067123413086,
+        -0.06531530618667603,
+        0.2507650554180145,
+        0.9099032282829285,
+        0.07794909924268723,
+        -0.15903376042842865,
+        0.29115578532218933,
+        0.48632845282554626,
+        0.4680853486061096,
+        0.40000519156455994,
+        0.7901750206947327,
+        0.11165501922369003,
+        0.1871986985206604,
+        0.15685616433620453,
+        0.7400000095367432
+      ],
+      "min": [
+        -0.00044060105574317276,
+        -0.029227260500192642,
+        -0.0007062808726914227,
+        -0.006396367214620113,
+        -0.034731876105070114,
+        -0.00020073111227247864,
+        -8.215621392082539e-07,
+        -0.5499086976051331,
+        -0.5100165009498596,
+        -0.613179087638855,
+        -0.0030598489101976156,
+        -0.0002515389060135931,
+        -0.00361030176281929,
+        -0.003131122561171651,
+        -0.30267173051834106,
+        0.162300705909729,
+        -0.8084174394607544,
+        -0.053157128393650055,
+        -0.48188674449920654,
+        -0.28324440121650696,
+        -0.02153456024825573,
+        -0.559512734413147,
+        -0.4063037037849426,
+        -0.625334620475769,
+        -0.17857033014297485,
+        -0.14080968499183655,
+        -0.3861367404460907,
+        -0.2920348048210144,
+        -0.0902835875749588,
+        -0.1666938215494156,
+        -0.07615894079208374,
+        0.44999998807907104
+      ],
+      "q01": [
+        7.74661926357112e-07,
+        -6.3755543715160465e-06,
+        -7.83351255222442e-07,
+        3.4136806561946284e-07,
+        -0.0008449232077691706,
+        4.5431972239384775e-06,
+        7.721260197968149e-07,
+        -0.5028422969579697,
+        -0.3325865414738655,
+        -0.5741579407453536,
+        -1.5824165325284411e-06,
+        -1.323924946916577e-07,
+        -9.707011122372932e-07,
+        -1.9067205457190538e-07,
+        -0.1996450574696064,
+        0.1655060650408268,
+        -0.6192439311742782,
+        0.0036195464059710497,
+        -0.37595251262187956,
+        -0.10034843616187572,
+        0.029302983712404963,
+        -0.29443797826766965,
+        -0.26330254584550855,
+        -0.3154676526784897,
+        0.021579700019210574,
+        -0.02289357639849186,
+        -0.1849268364906311,
+        -0.24821986511349678,
+        -0.07978948682546616,
+        -0.08723165072500706,
+        -0.05445469941943884,
+        0.5099999904632568
+      ],
+      "q99": [
+        2.1163743895158388e-06,
+        3.5946895195593186e-06,
+        9.480705858777559e-07,
+        0.0007039297890150918,
+        0.0005248256213963012,
+        2.909682405515922e-05,
+        3.943643955608417e-06,
+        2.646454402110975e-07,
+        2.4528015819669183e-06,
+        3.882593460957627e-07,
+        0.39650109171867354,
+        1.1101934683322905,
+        0.5173005294799803,
+        1.034569376707077,
+        0.03528875216841695,
+        0.22509524688124644,
+        0.29769810587167667,
+        0.28808553427457806,
+        -0.08342873558402061,
+        0.20056841030716893,
+        0.6333562320470806,
+        0.05092002365738146,
+        -0.1633606669306755,
+        0.21988036155700677,
+        0.343479991853237,
+        0.2987542548775673,
+        0.12391192510724047,
+        0.48378621041774733,
+        0.09747317329049104,
+        0.14307723090052604,
+        0.09777400560677017,
+        0.7400000095367432
+      ]
+    },
+    "num_transitions": 23664,
+    "num_trajectories": 100
+  }
+}

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/final_model/pytorch_model.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58eaff18dd7c1a3d468256151321af5e980b6fe219ac8291f98484134d166afd
+size 8604557774

intervla-m1/simple/G1WholebodyXMoveBendPickTeleop/20260403_151218/summary.jsonl

@@ -0,0 +1,8 @@
+{"steps": 5000}
+{"steps": 10000}
+{"steps": 15000}
+{"steps": 20000}
+{"steps": 25000}
+{"steps": 30000}
+{"steps": 35000}
+{"steps": 40000}