chore(training): add air_touch_bonus_weight and restart stage-5 intercepts

Browse Source

air_approach_weight alone didn't move productive_air_touch_fraction after a
further 180M steps (360M cumulative across all six Stage-5 attempts): an
unredirected air-intercept ball falls short of the goal from gravity and
just lands on the floor, so the already-solved ground game collects the
same episode reward whether or not anything touched the ball in the air.
air_touch_bonus_weight adds a conjunctive event bonus on top of
ball_touch_reward for a touch that's both genuinely aerial and
goal-directed, targeting the actual measured behaviour instead of only the
approach to it.

...

This commit is contained in:

Josh Creek

2026-08-19 22:46:04 +01:00

parent 03f49e59c8

commit 602fa297d0

15 changed files with 100 additions and 260 deletions

Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL

Download Patch File Download Diff File Expand all files Collapse all files

training/checkpoints/20260818-1903-gen5-s5-intercepts/final.zip

BIN

Binary file not shown.

training/checkpoints/20260819-0412-gen5-s5-intercepts-retry1/final.zip

BIN

Binary file not shown.

training/checkpoints/20260819-1321-gen5-s5-intercepts-retry2/final.zip

BIN

Binary file not shown.