Datasets:

---

olingoudey

/

8tasks_rollouts

Like 0

Adds pooled/features.npz (per-episode float32 arrays of shape (policy steps, 6144): the mean base-image, wrist-image and non-padded language token embeddings of every policy step), pooled/targets.npz (the failure mode of each failed episode, the timeline phrase of every policy step and CLIP ViT-B/32 text embeddings of the phrases, built from the labels), pooled/README.md (format, keys, a loading example) and pooled/pool_features.py (the code that produced the features; it reproduces the stored array exactly).

Both files are keyed by the dir column of manifest.csv; the 399 arrays match the manifest's n_policy. Nothing in the existing files is changed.