jgrusewski 86863d72e6 plan: GPU composite reward implementation — 7 tasks ...

Task 1: CUDA kernel (8-component formula, PORTFOLIO_STRIDE=12)
Task 2: Config pipeline (DQNHyperparameters + TOML profiles)
Task 3: Experience collector (buffer alloc + kernel launch)
Task 4: Hyperopt adapter (31D→38D search space)
Task 5: Remove dead CPU reward code
Task 6: Integration test (smoke + hyperopt + GPU suite)
Task 7: H100 validation

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

2026-03-22 20:35:57 +01:00

..

plans

plan: GPU composite reward implementation — 7 tasks

2026-03-22 20:35:57 +01:00

specs

spec: address review — DSR formula, div-by-zero guards, stride constant

2026-03-22 20:30:35 +01:00