--- license: apache-2.0 library_name: openpi tags: - robotics - manipulation - pi0.5 - openpi - jax - orbax pipeline_tag: robotics --- # openpi-pi05-routing-d1-baseline-uniform-r0-3cam-crop-20k-idle-ft-delta A [Pi0.5](https://www.physicalintelligence.company/blog/pi0-5) model fine-tuned using the [OpenPI](https://github.com/Physical-Intelligence/openpi) framework. ## Model Details | Property | Value | |---|---| | OpenPI config | `pi05_sir_droid_finetune_routing_3cam_crop` | | Checkpoint step | 24 | | Training data | N/A | | Precision | bfloat16 | | Parameter size | ~6.7 GB | | Source checkpoint | `/work/hdd/bgdg/lankile/sir-workspace/self-improving-robots/deps/openpi/checkpoints/pi05_sir_droid_finetune_routing_3cam_crop/sir_01b_routing_d1_r0_baseline_uniform_pi05_3cam_crop_20k_bs32_idle_delta/24` | | Hugging Face repo | `ankile/openpi-pi05-routing-d1-baseline-uniform-r0-3cam-crop-20k-idle-ft-delta` | | W&B run | [link](https://wandb.ai/self-improving/real-dagger-mining-01b/runs/37kea88a) | | SLURM job ID | `19976891` | ## Usage ### Download and run inference ```bash # Download checkpoint from HF Hub huggingface-cli download ankile/openpi-pi05-routing-d1-baseline-uniform-r0-3cam-crop-20k-idle-ft-delta --local-dir # Run inference server cd deps/openpi uv run python scripts/serve_policy.py pi05_sir_droid_finetune_routing_3cam_crop \ --checkpoint-dir ``` ### In-process inference (Python) ```python from openpi.training import config as openpi_config from openpi.policies import policy_config as openpi_policy_config train_config = openpi_config.get_config("pi05_sir_droid_finetune_routing_3cam_crop") policy = openpi_policy_config.create_trained_policy( train_config, "" ) result = policy.infer(obs_dict) actions = result["actions"] ``` ## Checkpoint Format [Orbax](https://github.com/google/orbax) format, all parameters in bfloat16. ``` ├── _CHECKPOINT_METADATA ├── checkpoint_provenance.json ├── openpi_config.json ├── assets/ │ └── (normalization stats) └── params/ └── (orbax checkpoint files) ``` ## License Apache 2.0