value_model can't be None, so it shouldn't be Optional or have default None

Author: AMindToThink

trl/trainer/ppo_trainer.py

@@ -107,7 +107,7 @@ def __init__(
         ref_model: Optional[nn.Module],
         reward_model: nn.Module,
         train_dataset: Dataset,
-        value_model: Optional[nn.Module] = None,
+        value_model: nn.Module,
         data_collator: Optional[DataCollatorWithPadding] = None,
         eval_dataset: Optional[Union[Dataset, dict[str, Dataset]]] = None,
         # less commonly used