Update train.py

train.py

@@ -56,8 +56,8 @@ except Exception as e:
 MAX_SEQ_LEN = 1024
 SFT_STEPS = 80  # Increased warmup for JSON format - key fix!
 GRPO_STEPS = 250
-GRPO_K = 2
-GRPO_LR = 2e-5  # Slightly higher LR for faster initial learning
+GRPO_K = 4
+GRPO_LR = 5e-6  # Slightly higher LR for faster initial learning
 CURRICULUM_SWITCH = 0  # Start with Level 1, advance early
 GRAD_CLIP = 1.0
 SAVE_EVERY = 50