Add files using upload-large-folder tool

CHECKPOINTS.txt

@@ -0,0 +1,12 @@
+Included checkpoint steps for `pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage2_a100_2gpu_b16`:
+
+- 20000
+- 25000
+- 30000
+
+Each checkpoint directory contains:
+
+- model.safetensors
+- metadata.pt
+- optimizer.pt
+- assets/ur5_lab_test_tube_camera_shifts/norm_stats.json

README.md

@@ -1,3 +1,42 @@
----
-license: mit
----
+# GeoPi0-new
+
+PyTorch `pi0` policy fine-tuned on the local UR5 real-robot LeRobot-format dataset with the GT stage-2 foreground cross-view distillation setup:
+
+- dataset: `ur5_lab_test_tube_camera_shifts`
+- training config name: `pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard`
+- experiment name: `pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage2_a100_2gpu_b16`
+- source Slurm job: `9555584`
+- base camera view: `observation.images.context_left_rgb`
+- wrist camera view: `observation.images.wrist_right_rgb`
+- base model init: `/scratch/yz11445/pi0_base`
+- stage-1 init weights: `/scratch/yz11445/tmp/openpi_cam/checkpoints/pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage1_hard/pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage1_a100_2gpu_b16/5000`
+- currently included checkpoint steps: `20000`, `25000`, `30000`
+
+This run completed through `30000` steps.
+
+## Included Files
+
+- `config.json`: base Pi0 model config copied from the initialization checkpoint
+- `model_architecture_config.json`: fine-tuned architecture settings used by this run
+- `training_config_summary.json`: training/data/run summary for this release
+- `assets/ur5_lab_test_tube_camera_shifts/norm_stats.json`: normalization statistics
+- `checkpoints/<step>/`: checkpoint snapshots with `model.safetensors`, `metadata.pt`, `optimizer.pt`, and copied assets
+
+## Inference
+
+Serve one of the included checkpoints with:
+
+```bash
+uv run scripts/serve_policy.py policy:checkpoint \
+  --policy.config=pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard \
+  --policy.dir=/path/to/GeoPi0-new/checkpoints/30000
+```
+
+Replace `30000` with one of `20000`, `25000`, or `30000`.
+
+## Notes
+
+- The policy loader uses the code-defined training config `pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard` from `src/openpi/training/config.py`.
+- This GT stage-2 variant uses PRoPE ray encoding, foreground cross-view fusion, and hard-confidence auxiliary point supervision from the grid-224 GT target cache.
+- Normalization stats are loaded from `assets/ur5_lab_test_tube_camera_shifts/norm_stats.json` inside each checkpoint directory.
+- Tokenizer assets are not bundled in this release directory. In this codebase, the Pi0 tokenizer is loaded at runtime from external sources referenced in `src/openpi/models/tokenizer.py`.

assets/ur5_lab_test_tube_camera_shifts/norm_stats.json

@@ -0,0 +1,80 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -3.340229034423828,
+        -0.993720531463623,
+        1.4748752117156982,
+        -2.027818441390991,
+        -1.5979795455932617,
+        4.608609199523926,
+        0.5474765300750732
+      ],
+      "std": [
+        0.19318173825740814,
+        0.14630725979804993,
+        0.2504045069217682,
+        0.11725258827209473,
+        0.0358712300658226,
+        0.20255139470100403,
+        0.49774089455604553
+      ],
+      "q01": [
+        -3.7030182856559755,
+        -1.2474105840682983,
+        0.7992883544683457,
+        -2.24708842587471,
+        -1.7682025527954102,
+        4.231052602767944,
+        0.0
+      ],
+      "q99": [
+        -3.0268489632606506,
+        -0.5849727642059327,
+        1.865938720214367,
+        -1.7351905375003815,
+        -1.5186957674026489,
+        4.9647599239349365,
+        0.9998
+      ]
+    },
+    "actions": {
+      "mean": [
+        0.016937920823693275,
+        -0.004969933535903692,
+        0.009962701238691807,
+        -0.001688019372522831,
+        0.0012489908840507269,
+        0.018919125199317932,
+        0.5474765300750732
+      ],
+      "std": [
+        0.06978615373373032,
+        0.06991613656282425,
+        0.09595422446727753,
+        0.059662021696567535,
+        0.010708698071539402,
+        0.06966347992420197,
+        0.49774089455604553
+      ],
+      "q01": [
+        -0.1645936336517334,
+        -0.21226087988615036,
+        -0.2664332183122635,
+        -0.1527675679922104,
+        -0.02410040438175201,
+        -0.154551212310791,
+        0.0
+      ],
+      "q99": [
+        0.21269720268249515,
+        0.21668597954511637,
+        0.3531587926626205,
+        0.18083614184856417,
+        0.03492184810638427,
+        0.21823061800003052,
+        0.9998
+      ]
+    }
+  }
+}

checkpoints/20000/assets/ur5_lab_test_tube_camera_shifts/norm_stats.json

@@ -0,0 +1,80 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -3.340229034423828,
+        -0.993720531463623,
+        1.4748752117156982,
+        -2.027818441390991,
+        -1.5979795455932617,
+        4.608609199523926,
+        0.5474765300750732
+      ],
+      "std": [
+        0.19318173825740814,
+        0.14630725979804993,
+        0.2504045069217682,
+        0.11725258827209473,
+        0.0358712300658226,
+        0.20255139470100403,
+        0.49774089455604553
+      ],
+      "q01": [
+        -3.7030182856559755,
+        -1.2474105840682983,
+        0.7992883544683457,
+        -2.24708842587471,
+        -1.7682025527954102,
+        4.231052602767944,
+        0.0
+      ],
+      "q99": [
+        -3.0268489632606506,
+        -0.5849727642059327,
+        1.865938720214367,
+        -1.7351905375003815,
+        -1.5186957674026489,
+        4.9647599239349365,
+        0.9998
+      ]
+    },
+    "actions": {
+      "mean": [
+        0.016937920823693275,
+        -0.004969933535903692,
+        0.009962701238691807,
+        -0.001688019372522831,
+        0.0012489908840507269,
+        0.018919125199317932,
+        0.5474765300750732
+      ],
+      "std": [
+        0.06978615373373032,
+        0.06991613656282425,
+        0.09595422446727753,
+        0.059662021696567535,
+        0.010708698071539402,
+        0.06966347992420197,
+        0.49774089455604553
+      ],
+      "q01": [
+        -0.1645936336517334,
+        -0.21226087988615036,
+        -0.2664332183122635,
+        -0.1527675679922104,
+        -0.02410040438175201,
+        -0.154551212310791,
+        0.0
+      ],
+      "q99": [
+        0.21269720268249515,
+        0.21668597954511637,
+        0.3531587926626205,
+        0.18083614184856417,
+        0.03492184810638427,
+        0.21823061800003052,
+        0.9998
+      ]
+    }
+  }
+}

checkpoints/20000/metadata.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c896057f44b5d30f50f678f3c9c5e5f0be89005d07511dc4cbc032c37813a34
+size 4787

checkpoints/20000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f369071a6817ffd1cead58eaebd7262c5d0cf933e326daae4cb088dcdaeb866
+size 7141718132

checkpoints/20000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed09e458297e75aec54c1e1d47d0d2877299ca6a211872c19b65cc09e27aa365
+size 12809338147

checkpoints/25000/assets/ur5_lab_test_tube_camera_shifts/norm_stats.json

@@ -0,0 +1,80 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -3.340229034423828,
+        -0.993720531463623,
+        1.4748752117156982,
+        -2.027818441390991,
+        -1.5979795455932617,
+        4.608609199523926,
+        0.5474765300750732
+      ],
+      "std": [
+        0.19318173825740814,
+        0.14630725979804993,
+        0.2504045069217682,
+        0.11725258827209473,
+        0.0358712300658226,
+        0.20255139470100403,
+        0.49774089455604553
+      ],
+      "q01": [
+        -3.7030182856559755,
+        -1.2474105840682983,
+        0.7992883544683457,
+        -2.24708842587471,
+        -1.7682025527954102,
+        4.231052602767944,
+        0.0
+      ],
+      "q99": [
+        -3.0268489632606506,
+        -0.5849727642059327,
+        1.865938720214367,
+        -1.7351905375003815,
+        -1.5186957674026489,
+        4.9647599239349365,
+        0.9998
+      ]
+    },
+    "actions": {
+      "mean": [
+        0.016937920823693275,
+        -0.004969933535903692,
+        0.009962701238691807,
+        -0.001688019372522831,
+        0.0012489908840507269,
+        0.018919125199317932,
+        0.5474765300750732
+      ],
+      "std": [
+        0.06978615373373032,
+        0.06991613656282425,
+        0.09595422446727753,
+        0.059662021696567535,
+        0.010708698071539402,
+        0.06966347992420197,
+        0.49774089455604553
+      ],
+      "q01": [
+        -0.1645936336517334,
+        -0.21226087988615036,
+        -0.2664332183122635,
+        -0.1527675679922104,
+        -0.02410040438175201,
+        -0.154551212310791,
+        0.0
+      ],
+      "q99": [
+        0.21269720268249515,
+        0.21668597954511637,
+        0.3531587926626205,
+        0.18083614184856417,
+        0.03492184810638427,
+        0.21823061800003052,
+        0.9998
+      ]
+    }
+  }
+}

checkpoints/25000/metadata.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54efb129055f709d05ae2b71832d1b482ac285cb928e49a6f3e984ea673138b9
+size 4787

checkpoints/25000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:632788e304af673507df80d544545af531c7cab333c89c8f2c29ac7bc685d304
+size 7141718132

checkpoints/25000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:208e094178e01cfd15a26ae3b17d13fd69743639e59ec877c34cb7f8252d9c40
+size 12809338147

checkpoints/30000/assets/ur5_lab_test_tube_camera_shifts/norm_stats.json

@@ -0,0 +1,80 @@
+{
+  "norm_stats": {
+    "state": {
+      "mean": [
+        -3.340229034423828,
+        -0.993720531463623,
+        1.4748752117156982,
+        -2.027818441390991,
+        -1.5979795455932617,
+        4.608609199523926,
+        0.5474765300750732
+      ],
+      "std": [
+        0.19318173825740814,
+        0.14630725979804993,
+        0.2504045069217682,
+        0.11725258827209473,
+        0.0358712300658226,
+        0.20255139470100403,
+        0.49774089455604553
+      ],
+      "q01": [
+        -3.7030182856559755,
+        -1.2474105840682983,
+        0.7992883544683457,
+        -2.24708842587471,
+        -1.7682025527954102,
+        4.231052602767944,
+        0.0
+      ],
+      "q99": [
+        -3.0268489632606506,
+        -0.5849727642059327,
+        1.865938720214367,
+        -1.7351905375003815,
+        -1.5186957674026489,
+        4.9647599239349365,
+        0.9998
+      ]
+    },
+    "actions": {
+      "mean": [
+        0.016937920823693275,
+        -0.004969933535903692,
+        0.009962701238691807,
+        -0.001688019372522831,
+        0.0012489908840507269,
+        0.018919125199317932,
+        0.5474765300750732
+      ],
+      "std": [
+        0.06978615373373032,
+        0.06991613656282425,
+        0.09595422446727753,
+        0.059662021696567535,
+        0.010708698071539402,
+        0.06966347992420197,
+        0.49774089455604553
+      ],
+      "q01": [
+        -0.1645936336517334,
+        -0.21226087988615036,
+        -0.2664332183122635,
+        -0.1527675679922104,
+        -0.02410040438175201,
+        -0.154551212310791,
+        0.0
+      ],
+      "q99": [
+        0.21269720268249515,
+        0.21668597954511637,
+        0.3531587926626205,
+        0.18083614184856417,
+        0.03492184810638427,
+        0.21823061800003052,
+        0.9998
+      ]
+    }
+  }
+}

checkpoints/30000/metadata.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b183989355a8574ea019c1b6462e3a92adc962a336d0d4dc794611646a090ee
+size 4787

checkpoints/30000/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a203044bce2adead66adf580747822509675194d99bf34f97a04bf709d65c0a0
+size 7141718132

checkpoints/30000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a84f3bd8a26745b0b41ef476249963f7bc2d7015563d5258c1b56822d7e4ec6
+size 12809338147

config.json

@@ -0,0 +1,7 @@
+{
+  "action_dim": 32,
+  "action_horizon": 50,
+  "paligemma_variant": "gemma_2b",
+  "action_expert_variant": "gemma_300m",
+  "precision": "bfloat16"
+}

model_architecture_config.json

@@ -0,0 +1,48 @@
+{
+  "model_family": "pi0",
+  "framework": "pytorch",
+  "dtype": "bfloat16",
+  "paligemma_variant": "gemma_2b",
+  "action_expert_variant": "gemma_300m",
+  "action_dim": 32,
+  "action_horizon": 50,
+  "max_token_len": 48,
+  "pi05": false,
+  "discrete_state_input": false,
+  "pose_enc_type": "prope",
+  "ray_enc_type": true,
+  "view_enc_type": false,
+  "cross_view": {
+    "type": "standard",
+    "aa_order": "fg",
+    "prope_layer_idx": [
+      0
+    ]
+  },
+  "disable_geometric_augs": true,
+  "action_loss_weight": 1.0,
+  "aux_point_head": {
+    "enabled": true,
+    "loss_weight": 0.05,
+    "loss_type": "legacy_conf_mse",
+    "conf_weight_mode": "hard",
+    "conf_threshold": 0.1,
+    "output_resolution": 224
+  },
+  "point_targets": {
+    "type": "gt",
+    "gt_ratio": 1.0,
+    "cams": [
+      [
+        "base",
+        "base"
+      ],
+      [
+        "left_wrist",
+        "left_wrist"
+      ]
+    ]
+  },
+  "ray_embed_pi3x_init_path": null,
+  "ray_embed_pi3x_init_scale": 1.0
+}

training_config_summary.json

@@ -0,0 +1,68 @@
+{
+  "repo_name": "GeoPi0-new",
+  "source_slurm_job_id": 9555584,
+  "train_config_name": "pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard",
+  "exp_name": "pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage2_a100_2gpu_b16",
+  "stage1_train_config_name": "pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage1_hard",
+  "stage1_exp_name": "pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage1_a100_2gpu_b16",
+  "dataset_path": "/scratch/yz11445/real_robot_data/ur5_lab_test_tube_camera_shifts",
+  "dataset_repo_id": "ur5_lab_test_tube_camera_shifts",
+  "norm_stats_source": "/scratch/yz11445/pi0_ur5_real_robot/ur5_lab_test_tube_camera_shifts",
+  "gt_point_targets_root": "/scratch/yz11445/gt_point_targets_grid224/ur5_lab_test_tube_camera_shifts",
+  "checkpoint_base_dir": "/scratch/yz11445/tmp/openpi_cam/checkpoints",
+  "release_checkpoint_source_dir": "/scratch/yz11445/tmp/openpi_cam/checkpoints/pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard/pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage2_a100_2gpu_b16",
+  "base_model_path": "/scratch/yz11445/pi0_base",
+  "init_weights_path": "/scratch/yz11445/tmp/openpi_cam/checkpoints/pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage1_hard/pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage1_a100_2gpu_b16/5000",
+  "num_gpus": 2,
+  "batch_size": 16,
+  "num_train_steps": 30000,
+  "save_interval": 5000,
+  "keep_period": 5000,
+  "completed_checkpoint_steps": [
+    20000,
+    25000,
+    30000
+  ],
+  "available_source_checkpoint_steps_at_packaging_time": [
+    5000,
+    10000,
+    15000,
+    20000,
+    25000,
+    30000
+  ],
+  "base_camera_key": "observation.images.context_left_rgb",
+  "wrist_camera_key": "observation.images.wrist_right_rgb",
+  "prompt_from_task": true,
+  "delta_actions_first_6_joints_only": true,
+  "pose_enc_type": "prope",
+  "ray_enc_type": true,
+  "view_enc_type": false,
+  "cross_view_type": "standard",
+  "cross_view_order": "fg",
+  "prope_layer_idx": [
+    0
+  ],
+  "uses_gt_point_targets": true,
+  "point_target_gt_ratio": 1.0,
+  "point_target_cams": [
+    [
+      "base",
+      "base"
+    ],
+    [
+      "left_wrist",
+      "left_wrist"
+    ]
+  ],
+  "aux_point_head_loss_type": "legacy_conf_mse",
+  "aux_point_head_loss_weight": 0.05,
+  "aux_point_head_conf_threshold": 0.1,
+  "point_target_resolution": 224,
+  "wandb_project": "openpi_cam_real_robot",
+  "wandb_entity": "NYU-robotics",
+  "serve_command_example": "uv run scripts/serve_policy.py policy:checkpoint --policy.config=pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard --policy.dir=/path/to/GeoPi0-new/checkpoints/30000",
+  "train_command_reference": "uv run python scripts/train_pytorch.py pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard --exp_name pi0_ur5_real_robot_cross_attn_fg_distill_gt_hard_stage2_a100_2gpu_b16",
+  "norm_stats_command_reference": "uv run scripts/compute_norm_stats.py --config-name pi0_ur5_real_robot_pytorch_cross_attn_fg_distill_gt_stage2_hard",
+  "run_status": "completed"
+}

wandb_id.txt

@@ -0,0 +1 @@
+unknown