openpi-pi05-franka-insert-marker-v2-dagger-r5-action-expert-only-ft

A Pi0.5 model fine-tuned using the OpenPI framework.

Model Details

Property	Value
OpenPI config	pi05_sir_droid_finetune_action_expert_only
Checkpoint step	4999
Training data	N/A
Precision	bfloat16
Parameter size	~6.7 GB
Source checkpoint	/iris/u/ankile/self-improving-robots-workspace/real-world-worktree/deps/openpi/checkpoints/pi05_sir_droid_finetune_action_expert_only/sir_r5_action_expert_only_5k_bs32_iris_hi/4999
Hugging Face repo	ankile/openpi-pi05-franka-insert-marker-v2-dagger-r5-action-expert-only-ft
W&B run	link

Usage

Download and run inference

# Download checkpoint from HF Hub
huggingface-cli download ankile/openpi-pi05-franka-insert-marker-v2-dagger-r5-action-expert-only-ft --local-dir <local_path>

# Run inference server
cd deps/openpi
uv run python scripts/serve_policy.py pi05_sir_droid_finetune_action_expert_only \
  --checkpoint-dir <local_path>

In-process inference (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_action_expert_only")
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
├── checkpoint_provenance.json
├── openpi_config.json
├── assets/
│   └── (normalization stats)
└── params/
    └── (orbax checkpoint files)

License

Apache 2.0