# Download checkpoint from HF Hub
huggingface-cli download ankile/openpi-pi05-droid-pretrained --local-dir <local_path>

# Run inference server
cd deps/openpi
uv run python scripts/serve_policy.py pi05_droid \
  --checkpoint-dir <local_path>

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_droid")
policy = openpi_policy_config.create_trained_policy(
    train_config, "<local_path>"
)
result = policy.infer(obs_dict)
actions = result["actions"]

Checkpoint Format

Orbax format, all parameters in bfloat16.

├── _CHECKPOINT_METADATA
├── assets/
│   └── (normalization stats)
└── params/
    └── (orbax checkpoint files)

License

Apache 2.0