jgrusewski
3e8718468f
test: fix dqn_action_collapse_fix_test for GPU composite reward
- Replace test_hold_penalty_weight_used_directly with
test_idle_penalty_in_gpu_composite_reward (verifies w_idle > 0)
- Update hyperopt dimension assertions from 31D to 38D
(7 composite reward weight dimensions added in prior commit)
All integration tests pass:
- dqn_training_smoke_test: 1/1
- hyperopt lib tests: 134/134
- dqn_action_collapse_fix_test: 10/10
- dqn_early_stopping_termination_test: 4/4
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>