s015-scan-object-ctr50-nomask-pi05-step10000

E748 Scan Object PI0.5 checkpoint — step 10000

Inference checkpoint from the S015 CTR50 without IdleMask arm. It contains the OpenPI parameter tree and normalization assets. Optimizer and training loader state are excluded.

Training provenance

• Experiment / Run: E748 / E748-R001
• Policy: PI0.5 JAX LoRA; checkpoint step 10000 of 30,000
• Initialization: XinY0201/openpi-pi05-base-jax at 5e62884fcf8cb8f9fc693c9163ea18d3e3739658
• Seed: 87431; batch size 16; gradient accumulation 1
• Dataset: Shiki42/ctr-scan-object-uniform50-20260917 at revision a34f2a32d3738e3607f226a777b41d42d0522908
• Data: Uniform50, 50 episodes / 14,163 frames
• IdleMask: disabled; standard temporal padding mask retained
• Normalization: exact global NumPy quantiles at 1% and 99%, linear interpolation; artifact SHA-256 e08ec02f286016017abf90e5be9f11e7507bd4b93a1b68473ac22dd7c3fb62b8
• CTR commit: a58fedcff4270d60c52a965fd58e7ce6a62d91d1
• OpenPI commit: e9ba7b7732a3e66e4bd87d6d3429f7cef6352ead
• Runtime: 2026-08-22.2; manifest SHA-256 06e035fbc62dd7f32f01a2f5ad536f01bd5af6caae3239698dbb8c07ed503332
• Checkpoint verification: fresh-process reload and finite-parameter checks passed for steps 10k, 20k, and 30k
• Evaluation: pending; this card makes no success-rate claim

Files

• params/: OpenPI model parameters
• assets/: normalization assets
• resolved_config.json: resolved inference configuration with machine-local paths removed
• training-provenance.json: pinned training, dataset, normalization, and runtime identities

Archival identity and evaluation version mapping (2026-10-05)

Scientific weights/configuration/actual normalization unchanged. E607 original central Experiment record unchanged. The historical inference revision below remains the weights identity. Only evaluation bindings for this exact step/revision are listed; unresolved receipt references stay explicit.

{
  "experiment": "E748",
  "training_run": "E748-R001",
  "step": 10000,
  "task": "scan-object",
  "algorithm": "pi05",
  "idle_mask": false,
  "data_condition": "ctr_without_idlemask",
  "revision": "e337f497a91b8730b7b2c9f654fc5bdb7c5ecac5",
  "package_manifest_sha256": "6d5164a46645f4cc0174aa0b8aa1b2dad2ad83466b9cdb7b36c64c9952b03ab2",
  "training_configuration": {
    "task": "Scan Object",
    "policy": "PI0.5 Base",
    "training_variant": "ctr_without_idlemask",
    "schedule": "uniform50",
    "idle_mask": false,
    "idle_mask_policy": "do not load or consume observation.arm_active_mask; ordinary temporal padding mask retained",
    "optimizer_updates": 30000,
    "retained_checkpoint_steps": [
      10000,
      20000,
      30000
    ],
    "batch_size": 16,
    "gradient_accumulation_steps": 1,
    "seed": 87431,
    "backend": "jax",
    "normalization_algorithm": "exact_global_native_v1",
    "reference_config": "pi05_putcab_athenb_fullhorizon_mb16_ga1_lora3ep",
    "reference_openpi_commit": "e9ba7b7732a3e66e4bd87d6d3429f7cef6352ead",
    "initialization": {
      "optimizer_step": 0,
      "repo_id": "XinY0201/openpi-pi05-base-jax",
      "revision": "5e62884fcf8cb8f9fc693c9163ea18d3e3739658",
      "state": "fresh model LoRA, optimizer, scheduler and RNG"
    },
    "dataset": {
      "dataset_status": "formally_usable_ctr_dataset",
      "publication_experiment": "E519",
      "repo_id": "Shiki42/ctr-scan-object-uniform50-20260917",
      "revision": "a34f2a32d3738e3607f226a777b41d42d0522908",
      "manifest_sha256": "e321f63b608403ac36c657cca1aaee7584452d0f8ff47ae3898f4ab18fa60e70",
      "episodes": 50,
      "frames": 14163,
      "fps": 25,
      "sample_range": "all 50 episodes",
      "schedule": "uniform50",
      "method": "uniform50",
      "idle_mask": false,
      "camera_preset": "centered_fovy90",
      "codebase_version": "v3.0",
      "require_verified_gripper_targets": true,
      "source_revision": "c46453109ca9e99b4561faf34a73770644917a2b"
    },
    "loader": {
      "adapter": "ctr.lerobot_v3",
      "entry": "ctr.openpi_entry",
      "format": "LeRobot v3.0",
      "gripper_actions": "absolute command",
      "joint_actions": "delta from current measured state once",
      "v3_video_offsets": true,
      "legacy_source_parent_commit": "ed2396899f14925191325a07a455681122f4e0ca"
    },
    "ablation": "S015 mask-only ablation: use the exact same CTR50 revision and samples as E746 while disabling observation.arm_active_mask consumption."
  },
  "dataset": null,
  "actual_normalization": null,
  "eligibility": null,
  "evaluation_bindings": [
    {
      "experiment": "E979",
      "title": "S015 Scan Object / CTR50 without IdleMask / PI0.5 LoRA 10k：100场评测",
      "checkpoint_step": 10000,
      "protocol": {
        "policy": "π0.5",
        "task": "Scan Object",
        "training_variant": "CTR50 without IdleMask",
        "checkpoint_step": 10000,
        "checkpoint_selection": "fixed a priori S015 retained step; all of 10k/20k/30k evaluated",
        "optimizer_updates": 0,
        "metrics": "success rate, Wilson95, Tcap, collisions",
        "scope": "S015 side study; not a Q4-A result cell",
        "protocol": {
          "suite": "paired-uniform-native100-20260917/scan_object",
          "manifest": "eval/robotwin/suites/paired-uniform-native100-20260917/scan_object.json",
          "sha": "056d214f65912164583b4bde806e46d9f2c97fe99d9eef17788dfad2fff6874c",
          "episodes": 100,
          "prompt": "Scan the object.",
          "step_limit": 500,
          "denoising_steps": 10,
          "action_chunk": 50,
          "camera_preset": "centered_fovy90",
          "scene_replacement": false,
          "retries": false
        },
        "performance_method": "retained E608-R002 K8 persistent-environment / fixed-B4 / renderer reuse / 250Hz physics configuration (same task, model family and adapter)",
        "checkpoint": {
          "repo_id": "Shiki42/s015-scan-object-ctr50-nomask-pi05-step10000",
          "revision": "e337f497a91b8730b7b2c9f654fc5bdb7c5ecac5",
          "file_count": 152,
          "manifest_sha256": "6d5164a46645f4cc0174aa0b8aa1b2dad2ad83466b9cdb7b36c64c9952b03ab2",
          "local_package": "/root/autodl-tmp/ctr-s015-eval-20260924/E979/inputs/checkpoint",
          "transport": "target-side HF mirror download at pinned revision"
        },
        "training_dataset_seed_receipt": "/root/autodl-tmp/ctr-q4a-20260921/E607/run-artifacts/dataset-qualified.json"
      },
      "runs": [
        {
          "id": "E979-R001",
          "status": "completed",
          "code_commit": "671a0a101876b035fa5dc2fd02e74d3a4f1f1437",
          "result": null,
          "result_uri": null
        }
      ],
      "step_binding": "explicit"
    }
  ],
  "preservation": "Keep every historical version. Deprecation flags remain visible and do not constitute restored qualification. Missing exact evaluation step must be resolved from raw receipt, not inferred."
}