Commit History
-
pushblock/simple-v0: re-record with the action-label gap fix (46.6% of rows held a placeholder 0); band U(3.25,4.75) unchanged
Verified signature08db4f4 -
pushblock/medium-v0: re-record with the action-label gap fix (61.0% of rows held a placeholder 0); band U(2.1,3.2) unchanged
Verified signature23025f5 -
pushblock/expert-v0: fix action labels — the collector recorded a fabricated 0 for every step an agent spent inside ML-Agents' action-repeat gap (72.16% of rows); it now carries the last real decision forward, so the action column matches the policy the trajectories were actually driven by
Verified signature364d655
-
pushblock/simple-v0: fixed-T -> per-episode softmax-temperature-range tier (T~U(3.25,4.75) per episode, step-reward norm 0.6025)
Verified signature60ecdbf -
pushblock/medium-v0: fixed-T -> per-episode softmax-temperature-range tier (T~U(2.1,3.2) per episode, step-reward norm 0.7984)
Verified signature666f067
-
pushblock/simple-v0: replace epsilon tier with softmax-temperature tier (T=4.15, step-reward calibrated)
Verified signaturef5fa0b7 -
pushblock/medium-v0: replace epsilon tier with softmax-temperature tier (T=2.8, step-reward calibrated)
Verified signature8f94074
-
feat(pushblock): simple tier, per-episode epsilon pool
Verified signature2721f9c -
feat(pushblock): medium tier, per-episode epsilon pool
Verified signaturef5460d9 -
Replace pushblock expert-v0 with release-23 trajectories
Verified signature3f95d27
-
Replace PushBlock sample with 1M real rollouts
Verified signatureReplaces the 2,557-transition demo conversion with 1,000,000 transitions collected from the stock release_20 PushBlock.nn policy running in Unity Barracuda. 52,034 episodes; Tuple(Box(105), Box(105)); Discrete(7).3c2c889 -
Add discrete PushBlock expert Minari dataset
Verified signatureConverted from the ML-Agents release_20 ExpertPushBlock.demo. Adds 96 episodes / 2,557 transitions with Tuple(Box(105), Box(105)) observations and Discrete(7) actions.7deb23f