In the paper, are the reported evaluation metrics (e.g., AP, CAP, PCE, CLIP-SCORE) averaged over multiple batches?

I used the code you gave me and the trained model to test on humanart and got the results for each batch. I would like to know how you got the results (e.g., AP) mentioned in your paper? Do you take the average?