oat-libero-policy-early-exit
Policy checkpoint from the oat-early-exit fork (OAT + optional early-exit decode).
Files
oat_policy_latest.ckpt — OAT policy weights (latest.ckpt from training).eval_log.json — simulator eval summary (if provided).logs.json — training logs.json (if provided).
Repro
Source git revision (when uploaded): ecf66fb.
Citation
If you use OAT, cite the original OAT paper/repo; add your own citation for this fork if needed.
