Upload folder using huggingface_hub

config.json

@@ -1,5 +1,5 @@
 {
-    "type": "posevla_qwen",
+    "type": "dynamic_pi05_qwen",
     "n_obs_steps": 1,
     "input_features": {
         "observation.state": {
@@ -19,37 +19,27 @@
     },
     "device": "cuda",
     "use_amp": false,
-    "push_to_hub": false,
-    "repo_id": "ShikangWang/posevla_full_delta",
+    "push_to_hub": true,
+    "repo_id": "lerobot/trajvla",
     "private": null,
     "tags": null,
     "license": null,
-    "pretrained_path": null,
+    "pretrained_path": "lerobot/pi05_base",
     "qwen3_vl_variant": "qwen3_vl_28l",
     "action_expert_variant": "qwen3_28l",
     "dtype": "bfloat16",
-    "pretrain": false,
     "chunk_size": 50,
     "n_action_steps": 50,
     "max_state_dim": 32,
     "max_action_dim": 32,
-    "perc_query_num": 20,
     "num_inference_steps": 10,
-    "time_sampling_beta_alpha": 1.5,
+    "time_sampling_beta_alpha": 2.5,
     "time_sampling_beta_beta": 1.0,
     "time_sampling_scale": 0.999,
     "time_sampling_offset": 0.001,
     "min_period": 0.004,
     "max_period": 4.0,
-    "pose_loss_weight": 0.1,
-    "loss_weight_exist": 1.0,
-    "loss_weight_position": 5.0,
-    "loss_weight_direction": 2.0,
-    "loss_weight_vmf": 0.1,
-    "loss_weight_geodesic": 1.0,
-    "match_cost_exist": 1.0,
-    "match_cost_position": 5.0,
-    "match_cost_direction": 0.5,
+    "random_time_prob": 0.2,
     "image_resolution": [
         224,
         224
@@ -67,7 +57,7 @@
     "freeze_vision_encoder": false,
     "train_expert_only": false,
     "freeze_action_expert": false,
-    "optimizer_lr": 5e-05,
+    "optimizer_lr": 2.5e-05,
     "optimizer_betas": [
         0.9,
         0.95
@@ -76,7 +66,7 @@
     "optimizer_weight_decay": 0.01,
     "optimizer_grad_clip_norm": 1.0,
     "vlm_lr_multiplier": 1.0,
-    "scheduler_warmup_steps": 2000,
-    "scheduler_decay_steps": 200000,
-    "scheduler_decay_lr": 5e-06
+    "scheduler_warmup_steps": 1000,
+    "scheduler_decay_steps": 30000,
+    "scheduler_decay_lr": 2.5e-06
 }

model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:5372fa7c1e6bcf0b40c03c3b93e80be0694de813454ea7975caede9d1ddf723b
-size 6030576512
+oid sha256:9612a04319184503a25fdbd64c99b29a179d85bc9072f8d08d6ccf8d6a84d5a6
+size 5157742576

stats.json

@@ -1,21 +1,95 @@
 {
     "aloha": {
-        "depth.stats": {
+        "object.poses": {
             "min": [
-                0.9118702411651611,
-                1.340759038925171
+                0.0,
+                0.0,
+                0.0,
+                0.0,
+                -1.0,
+                -0.9999995827674866,
+                -0.999999463558197,
+                -1.0,
+                -0.9999992847442627,
+                -0.999997615814209,
+                0.0,
+                0.0
             ],
             "max": [
-                153.3542022705078,
-                404.8944091796875
+                1.0,
+                0.9999998211860657,
+                0.9999935030937195,
+                1.0,
+                1.0,
+                0.9999986886978149,
+                0.9999992847442627,
+                1.0,
+                0.9999973773956299,
+                0.9999920129776001,
+                1.0,
+                10.0
             ],
             "mean": [
-                1.0104806423187256,
-                2.3054113388061523
+                0.2970513701438904,
+                0.14921699464321136,
+                0.11780871450901031,
+                0.07064822316169739,
+                0.02209528535604477,
+                -0.03858209401369095,
+                -0.07095351070165634,
+                0.038906268775463104,
+                -0.021387789398431778,
+                0.016884474083781242,
+                0.10209379345178604,
+                2.9705135822296143
             ],
             "std": [
-                0.10744518041610718,
-                0.7919256687164307
+                0.4569593369960785,
+                0.2758782207965851,
+                0.2220076471567154,
+                0.15976020693778992,
+                0.3349374532699585,
+                0.3041366636753082,
+                0.2921609878540039,
+                0.3853122889995575,
+                0.30464646220207214,
+                0.23134422302246094,
+                0.30277159810066223,
+                4.569594860076904
+            ],
+            "count": [
+                123800120
+            ]
+        },
+        "frame_index": {
+            "min": [
+                0.0
+            ],
+            "max": [
+                1642.0
+            ],
+            "mean": [
+                160.4863739013672
+            ],
+            "std": [
+                162.0838623046875
+            ],
+            "count": [
+                6190006
+            ]
+        },
+        "index": {
+            "min": [
+                0.0
+            ],
+            "max": [
+                393156.0
+            ],
+            "mean": [
+                71589.4609375
+            ],
+            "std": [
+                70381.171875
             ],
             "count": [
                 6190006
@@ -179,137 +253,24 @@
                 ]
             ]
         },
-        "observation.state": {
+        "episode_index": {
             "min": [
-                -7.323299407958984,
-                -0.4973906874656677,
-                -0.18824216723442078,
-                -1.9566115140914917,
-                -1.5014400482177734,
-                -8.142146110534668,
-                0.0,
-                -6.009664535522461,
-                -0.6168152093887329,
-                -0.0855255201458931,
-                -2.203854560852051,
-                -2.1122918128967285,
-                -8.424712181091309,
                 0.0
             ],
             "max": [
-                6.436553001403809,
-                4.218555450439453,
-                5.788050651550293,
-                1.7917098999023438,
-                2.1023752689361572,
-                8.890085220336914,
-                1.0,
-                6.736955165863037,
-                4.175413608551025,
-                5.656769752502441,
-                1.9575397968292236,
-                1.5057553052902222,
-                7.0688958168029785,
-                1.0
+                499.0
             ],
             "mean": [
-                -0.24772560596466064,
-                1.1504002809524536,
-                0.8188694715499878,
-                -0.3086509108543396,
-                0.05627923831343651,
-                -0.01663512922823429,
-                0.6496613621711731,
-                0.2113080769777298,
-                1.0522962808609009,
-                0.7641482949256897,
-                -0.31528306007385254,
-                -0.020076364278793335,
-                0.03699800744652748,
-                0.6886420249938965
+                229.02072143554688
             ],
             "std": [
-                0.3350416421890259,
-                0.999181866645813,
-                0.7699509263038635,
-                0.6670005321502686,
-                0.27609556913375854,
-                0.5944604277610779,
-                0.46040260791778564,
-                0.30005842447280884,
-                1.0245376825332642,
-                0.7941036224365234,
-                0.6766195297241211,
-                0.23668619990348816,
-                0.5875358581542969,
-                0.44642767310142517
+                151.9712677001953
             ],
             "count": [
                 6190006
             ]
         },
-        "object.poses": {
-            "min": [
-                0.0,
-                0.0,
-                0.0,
-                0.0,
-                -1.0,
-                -0.9999995827674866,
-                -0.999999463558197,
-                -1.0,
-                -0.9999992847442627,
-                -0.999997615814209,
-                0.0,
-                0.0
-            ],
-            "max": [
-                1.0,
-                0.9999998211860657,
-                0.9999935030937195,
-                1.0,
-                1.0,
-                0.9999986886978149,
-                0.9999992847442627,
-                1.0,
-                0.9999973773956299,
-                0.9999920129776001,
-                1.0,
-                10.0
-            ],
-            "mean": [
-                0.2970513701438904,
-                0.14921699464321136,
-                0.11780871450901031,
-                0.07064822316169739,
-                0.02209528535604477,
-                -0.03858209401369095,
-                -0.07095351070165634,
-                0.038906268775463104,
-                -0.021387789398431778,
-                0.016884474083781242,
-                0.10209379345178604,
-                2.9705135822296143
-            ],
-            "std": [
-                0.4569593369960785,
-                0.2758782207965851,
-                0.2220076471567154,
-                0.15976020693778992,
-                0.3349374532699585,
-                0.3041366636753082,
-                0.2921609878540039,
-                0.3853122889995575,
-                0.30464646220207214,
-                0.23134422302246094,
-                0.30277159810066223,
-                4.569594860076904
-            ],
-            "count": [
-                123800120
-            ]
-        },
-        "observation.images.cam_right_wrist": {
+        "observation.images.cam_high": {
             "min": [
                 [
                     [
@@ -484,41 +445,7 @@
                 6190006
             ]
         },
-        "episode_index": {
-            "min": [
-                0.0
-            ],
-            "max": [
-                499.0
-            ],
-            "mean": [
-                229.02072143554688
-            ],
-            "std": [
-                151.9712677001953
-            ],
-            "count": [
-                6190006
-            ]
-        },
-        "timestamp": {
-            "min": [
-                0.0
-            ],
-            "max": [
-                32.84000015258789
-            ],
-            "mean": [
-                3.209726572036743
-            ],
-            "std": [
-                3.2416794300079346
-            ],
-            "count": [
-                6190006
-            ]
-        },
-        "observation.images.cam_high": {
+        "observation.images.cam_right_wrist": {
             "min": [
                 [
                     [
@@ -676,18 +603,87 @@
                 ]
             ]
         },
-        "index": {
+        "observation.state": {
             "min": [
+                -7.323299407958984,
+                -0.4973906874656677,
+                -0.18824216723442078,
+                -1.9566115140914917,
+                -1.5014400482177734,
+                -8.142146110534668,
+                0.0,
+                -6.009664535522461,
+                -0.6168152093887329,
+                -0.0855255201458931,
+                -2.203854560852051,
+                -2.1122918128967285,
+                -8.424712181091309,
                 0.0
             ],
             "max": [
-                393156.0
+                6.436553001403809,
+                4.218555450439453,
+                5.788050651550293,
+                1.7917098999023438,
+                2.1023752689361572,
+                8.890085220336914,
+                1.0,
+                6.736955165863037,
+                4.175413608551025,
+                5.656769752502441,
+                1.9575397968292236,
+                1.5057553052902222,
+                7.0688958168029785,
+                1.0
             ],
             "mean": [
-                71589.4609375
+                -0.24772560596466064,
+                1.1504002809524536,
+                0.8188694715499878,
+                -0.3086509108543396,
+                0.05627923831343651,
+                -0.01663512922823429,
+                0.6496613621711731,
+                0.2113080769777298,
+                1.0522962808609009,
+                0.7641482949256897,
+                -0.31528306007385254,
+                -0.020076364278793335,
+                0.03699800744652748,
+                0.6886420249938965
             ],
             "std": [
-                70381.171875
+                0.3350416421890259,
+                0.999181866645813,
+                0.7699509263038635,
+                0.6670005321502686,
+                0.27609556913375854,
+                0.5944604277610779,
+                0.46040260791778564,
+                0.30005842447280884,
+                1.0245376825332642,
+                0.7941036224365234,
+                0.6766195297241211,
+                0.23668619990348816,
+                0.5875358581542969,
+                0.44642767310142517
+            ],
+            "count": [
+                6190006
+            ]
+        },
+        "timestamp": {
+            "min": [
+                0.0
+            ],
+            "max": [
+                32.84000015258789
+            ],
+            "mean": [
+                3.209726572036743
+            ],
+            "std": [
+                3.2416794300079346
             ],
             "count": [
                 6190006
@@ -762,18 +758,22 @@
                 242125300
             ]
         },
-        "frame_index": {
+        "depth.stats": {
             "min": [
-                0.0
+                0.9118702411651611,
+                1.340759038925171
             ],
             "max": [
-                1642.0
+                153.3542022705078,
+                404.8944091796875
             ],
             "mean": [
-                160.4863739013672
+                1.0104806423187256,
+                2.3054113388061523
             ],
             "std": [
-                162.0838623046875
+                0.10744518041610718,
+                0.7919256687164307
             ],
             "count": [
                 6190006

train_config.json

@@ -1,7 +1,7 @@
 {
     "dataset": {
-        "type": "posevla_qwen",
-        "repo_id": "adjust_bottle/demo_clean adjust_bottle/demo_randomized beat_block_hammer/demo_clean beat_block_hammer/demo_randomized blocks_ranking_rgb/demo_clean blocks_ranking_rgb/demo_randomized blocks_ranking_size/demo_clean blocks_ranking_size/demo_randomized click_alarmclock/demo_clean click_alarmclock/demo_randomized click_bell/demo_clean click_bell/demo_randomized dump_bin_bigbin/demo_clean dump_bin_bigbin/demo_randomized grab_roller/demo_clean grab_roller/demo_randomized handover_block/demo_clean handover_block/demo_randomized handover_mic/demo_clean handover_mic/demo_randomized hanging_mug/demo_clean hanging_mug/demo_randomized lift_pot/demo_clean lift_pot/demo_randomized move_can_pot/demo_clean move_can_pot/demo_randomized move_pillbottle_pad/demo_clean move_pillbottle_pad/demo_randomized move_playingcard_away/demo_clean move_playingcard_away/demo_randomized move_stapler_pad/demo_clean move_stapler_pad/demo_randomized open_laptop/demo_clean open_laptop/demo_randomized open_microwave/demo_clean open_microwave/demo_randomized pick_diverse_bottles/demo_clean pick_diverse_bottles/demo_randomized pick_dual_bottles/demo_clean pick_dual_bottles/demo_randomized place_a2b_left/demo_clean place_a2b_left/demo_randomized place_a2b_right/demo_clean place_a2b_right/demo_randomized place_bread_basket/demo_clean place_bread_basket/demo_randomized place_bread_skillet/demo_clean place_bread_skillet/demo_randomized place_burger_fries/demo_clean place_burger_fries/demo_randomized place_can_basket/demo_clean place_can_basket/demo_randomized place_cans_plasticbox/demo_clean place_cans_plasticbox/demo_randomized place_container_plate/demo_clean place_container_plate/demo_randomized place_dual_shoes/demo_clean place_dual_shoes/demo_randomized place_empty_cup/demo_clean place_empty_cup/demo_randomized place_fan/demo_clean place_fan/demo_randomized place_mouse_pad/demo_clean place_mouse_pad/demo_randomized place_object_basket/demo_clean place_object_basket/demo_randomized place_object_scale/demo_clean place_object_scale/demo_randomized place_object_stand/demo_clean place_object_stand/demo_randomized place_phone_stand/demo_clean place_phone_stand/demo_randomized place_shoe/demo_clean place_shoe/demo_randomized press_stapler/demo_clean press_stapler/demo_randomized put_bottles_dustbin/demo_clean put_bottles_dustbin/demo_randomized put_object_cabinet/demo_clean put_object_cabinet/demo_randomized rotate_qrcode/demo_clean rotate_qrcode/demo_randomized scan_object/demo_clean scan_object/demo_randomized shake_bottle/demo_clean shake_bottle/demo_randomized shake_bottle_horizontally/demo_clean shake_bottle_horizontally/demo_randomized stack_blocks_three/demo_clean stack_blocks_three/demo_randomized stack_blocks_two/demo_clean stack_blocks_two/demo_randomized stack_bowls_three/demo_clean stack_bowls_three/demo_randomized stack_bowls_two/demo_clean stack_bowls_two/demo_randomized stamp_seal/demo_clean stamp_seal/demo_randomized turn_switch/demo_clean turn_switch/demo_randomized",
+        "type": "dynamic_pi05_qwen",
+        "repo_id": "adjust_bottle/demo_clean",
         "root": "/mnt/data_pfs/robinwang/lerobot_datasets/robotwin_pose_v2",
         "episodes": null,
         "image_transforms": {
@@ -135,7 +135,7 @@
                     "max_action_dim": 32
                 },
                 {
-                    "type": "unify_posevla_qwen_inputs"
+                    "type": "unify_dynamic_pi05_qwen_inputs"
                 }
             ],
             "outputs": []
@@ -147,12 +147,11 @@
         "height": 224,
         "width": 224,
         "max_state_dim": 32,
-        "max_action_dim": 32,
-        "pretrain": false
+        "max_action_dim": 32
     },
     "env": null,
     "policy": {
-        "type": "posevla_qwen",
+        "type": "dynamic_pi05_qwen",
         "n_obs_steps": 1,
         "input_features": {
             "observation.state": {
@@ -172,37 +171,27 @@
         },
         "device": "cuda",
         "use_amp": false,
-        "push_to_hub": false,
-        "repo_id": "ShikangWang/posevla_full_delta",
+        "push_to_hub": true,
+        "repo_id": "lerobot/trajvla",
         "private": null,
         "tags": null,
         "license": null,
-        "pretrained_path": null,
+        "pretrained_path": "lerobot/pi05_base",
         "qwen3_vl_variant": "qwen3_vl_28l",
         "action_expert_variant": "qwen3_28l",
         "dtype": "bfloat16",
-        "pretrain": false,
         "chunk_size": 50,
         "n_action_steps": 50,
         "max_state_dim": 32,
         "max_action_dim": 32,
-        "perc_query_num": 20,
         "num_inference_steps": 10,
-        "time_sampling_beta_alpha": 1.5,
+        "time_sampling_beta_alpha": 2.5,
         "time_sampling_beta_beta": 1.0,
         "time_sampling_scale": 0.999,
         "time_sampling_offset": 0.001,
         "min_period": 0.004,
         "max_period": 4.0,
-        "pose_loss_weight": 0.1,
-        "loss_weight_exist": 1.0,
-        "loss_weight_position": 5.0,
-        "loss_weight_direction": 2.0,
-        "loss_weight_vmf": 0.1,
-        "loss_weight_geodesic": 1.0,
-        "match_cost_exist": 1.0,
-        "match_cost_position": 5.0,
-        "match_cost_direction": 0.5,
+        "random_time_prob": 0.2,
         "image_resolution": [
             224,
             224
@@ -220,7 +209,7 @@
         "freeze_vision_encoder": false,
         "train_expert_only": false,
         "freeze_action_expert": false,
-        "optimizer_lr": 5e-05,
+        "optimizer_lr": 2.5e-05,
         "optimizer_betas": [
             0.9,
             0.95
@@ -229,25 +218,25 @@
         "optimizer_weight_decay": 0.01,
         "optimizer_grad_clip_norm": 1.0,
         "vlm_lr_multiplier": 1.0,
-        "scheduler_warmup_steps": 2000,
-        "scheduler_decay_steps": 200000,
-        "scheduler_decay_lr": 5e-06
+        "scheduler_warmup_steps": 1000,
+        "scheduler_decay_steps": 30000,
+        "scheduler_decay_lr": 2.5e-06
     },
-    "output_dir": "outputs/posevla_qwen/2026_06_01_15_02_50-posevla_qwen-delta-posevla_qwen_robotwin_clean-finetune",
-    "job_name": "2026_06_01_15_02_50-posevla_qwen-delta-posevla_qwen_robotwin_clean-finetune",
+    "output_dir": "outputs/train/2026-06-04/16-25-26_dynamic_pi05_qwen",
+    "job_name": "dynamic_pi05_qwen",
     "resume": false,
-    "seed": 42,
-    "num_workers": 8,
-    "batch_size": 8,
-    "steps": 200000,
+    "seed": 1000,
+    "num_workers": 0,
+    "batch_size": 2,
+    "steps": 100000,
     "eval_freq": 20000,
-    "log_freq": 100,
+    "log_freq": 200,
     "save_checkpoint": true,
-    "save_freq": 10000,
+    "save_freq": 100,
     "use_policy_training_preset": true,
     "optimizer": {
         "type": "adamw",
-        "lr": 5e-05,
+        "lr": 2.5e-05,
         "weight_decay": 0.01,
         "grad_clip_norm": 1.0,
         "betas": [
@@ -258,10 +247,10 @@
     },
     "scheduler": {
         "type": "cosine_decay_with_warmup",
-        "num_warmup_steps": 2000,
-        "num_decay_steps": 200000,
-        "peak_lr": 5e-05,
-        "decay_lr": 5e-06
+        "num_warmup_steps": 1000,
+        "num_decay_steps": 30000,
+        "peak_lr": 2.5e-05,
+        "decay_lr": 2.5e-06
     },
     "eval": {
         "n_episodes": 50,
@@ -269,12 +258,12 @@
         "use_async_envs": false
     },
     "wandb": {
-        "enable": true,
-        "disable_artifact": true,
-        "project": "lerobot_lab_posevla_robotwin",
+        "enable": false,
+        "disable_artifact": false,
+        "project": "lerobot",
         "entity": null,
         "notes": null,
-        "run_id": "epwaxfvw",
+        "run_id": null,
         "mode": null
     },
     "checkpoint_path": null,