YAML Metadata Warning:empty or missing yaml metadata in repo card

Check out the documentation for more information.

openpi_cache

Mirror of the OpenPI model cache directory (~/.cache/openpi) used by VLAStudio.

OpenPI upstream hosts checkpoints on Google Cloud Storage (gs://openpi-assets, gs://big_vision), which is unreachable from many networks. This repository mirrors the default OpenPI data-home layout so it can be restored from Hugging Face instead.

Layout

openpi-assets/checkpoints/pi0_base/         # JAX ?0 base (fine-tune source)
openpi-assets/checkpoints/pi0_base_torch/   # PyTorch ?0 base (model.safetensors + config.json)
openpi-assets/checkpoints/pi05_base/        # JAX ?0.5 base (fine-tune source)
openpi-assets/checkpoints/pi05_base_torch/  # PyTorch ?0.5 base (model.safetensors + config.json)
big_vision/paligemma_tokenizer.model        # PaliGemma sentencepiece tokenizer

The JAX checkpoints also carry assets/<robot>/norm_stats.json normalization statistics.

Usage with VLAStudio

VLAStudio's OpenPI policy auto-downloads from this repository when a configured cache path is missing at startup (see src/vlastudio/policy/openpi/cache_utils.py):

  • Mirror repo id: WWZzz/openpi_cache (override with VLASTUDIO_OPENPI_CACHE_REPO)
  • Disable with VLASTUDIO_OPENPI_AUTO_DOWNLOAD=0

Manual restore:

hf download WWZzz/openpi_cache --local-dir ~/.cache/openpi

Note: hf download skips .lock metadata files, which is fine; they are transient GCS download locks.

Downloads last month

-

Downloads are not tracked for this model. How to track
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support