How do I get a final accuracy for my model if my data is split into train/validation/test

So I have my data partitioned into train, validation, and test. I used the Trainer class in the Transformers library to make a model for text classification. So I have my model with the eval accuracy and f1 at each validation step, but now how do I run my model against my test set? Do I have to do inference in a separate script and calculate accuracy myself? Or can I perform it at the end of training automatically.