fix GRPOConfig: remove invalid max_length

train_grpo.py

@@ -167,9 +167,9 @@ def main() -> None:
             save_strategy="steps",
             save_steps=50,
             save_total_limit=1,
-            max_length=1024,
             temperature=0.7,
             bf16=True,
+            remove_unused_columns=False,
             push_to_hub=False,
             hub_model_id=HUB_MODEL_ID,
             report_to="none",