Hi Yacine,
You pass the number of epochs you want the model to train, aka the number of times the models gets to improve using the entire dataset.
For more details, have a look at my answer here: "fine_tune" vs. "fit_one_cycle" - #7 by zerotosingularity