feat(kaggle): add clean_launch.py + shrink budget to 20/25/30 = 75 eps cd923aa Uddiii commited on 29 days ago
feat(kaggle): default to fixed-budget curriculum 20/30/50 episodes 69f89ec Uddiii commited on 29 days ago
fix(grpo): skip reference model when kl_beta=0 to save 5GB VRAM on T4 0566783 Uddiii commited on 29 days ago
fix(kaggle): align pip-managed numpy with kernel's loaded numpy 27cf9cd Uddiii commited on 29 days ago
fix(kaggle): escape backslash-n in REPAIR cell separator print 04688c1 Uddiii commited on 29 days ago