jaehyunkang/pi05-real-workbench-preset-2view-object-classification-60k

Pi0.5 Real Workbench — preset-2view-object-classification

Final checkpoint after 60,000 optimization steps. This is a trained policy, not an evaluation result.

• Dataset: Myungkyu/real_workbench-preset-gemini
• Task scope: object-classification
• Instruction: Per-frame parquet subtask
• Views: 2 (exterior and wrist)
• Global batch: 32; training GPUs: 2; seed: 42
• State: 8 dimensions; action: delta EEF 7 dimensions (6 Cartesian velocity + gripper)
• Image storage: 224×126; policy pads to 224×224
• Action chunk / execution horizon: 50; inference denoising steps: 10
• Training implementation: RLWRLD/hiwrld-ll-policy, vendored LeRobot Pi0.5. Custom input fields may require the matching implementation.

Policy weights, policy configuration, preprocessing/postprocessing and normalization states are at the repository root. Training resume files are in training_state/. Host-specific paths were removed from JSON metadata; supply local dataset/output paths when resuming. The tokenizer reference points to google/paligemma-3b-pt-224 (training revision 35e4f46485b4d07967e7e9935bc3786aad50687c).

For subtask models, supply the corresponding per-frame subtask as the policy task text. For 3-view models also supply the dataset-defined keyframe image. No real-robot evaluation metrics are claimed here.

File sizes and SHA-256 hashes are recorded in artifact_manifest.json. Original training checkpoints were preserved.