ThomasWalker1/paarbench-data

Optional training dataset

The PAARBench evaluation protocol does not require training trajectories. Its goal segments are tracked under data/pushobj_eval/, and evaluation reads fixed normalization metadata without opening the training dataset.

The training release is provided for researchers who want to reproduce base-model or adapter training, or develop methods that explicitly use offline trajectories:

.venv/bin/python scripts/download_data.py

The downloader fetches the immutable benchmark release from the ThomasWalker1/paarbench-data dataset repository into data/pushobj_multishape/ and verifies it against DATASET_MANIFEST.json.

Contents

The release contains only the nonredundant splits used by the training pipeline:

split	episodes	files	bytes
train	7,200	7,206	864,106,305
val	800	806	97,366,292

Each split contains states.pth, velocities.pth, rel_actions.pth, abs_actions.pth, seq_lengths.pkl, shapes.pkl, and one MP4 observation stream per episode under obses/. Redundant root-level copies and intermediate shards are not part of the release.

The trajectories were generated in the PushObj simulator across the T, L, Z, and + shapes. No license is asserted here beyond the rights of the repository owner; users should review the dataset repository card and applicable upstream asset licenses before redistribution.