Multiple labels inference from single label training?

I have started with the course. Reproduced first 3 lessons on my Laptop. All codes in the notebook are running properly.

So, now I decided to try something on my own.

I tried to train a model using photo of each of 3000 employee in our organization. I could gather only 1 reference image for each employee ID. So I trained it with 1 employee id label for each photo.

Next I collected some more photos and tried to predict the employee ID. However all the inference results are coming wrong. When I tried fit_one_cycle(4), I could correctly predict just 1 out of 40 photos. Train_loss was around 8.xyz. Than I tried fit_one_cycle(20) and Train_loss dropped to 0.2. Now I got 0 corecct prediction out of 40.

I am guessing, that inference outputs a probability value for each of the labels. How to get the top 10 predictions ? I hope I get correct answer in at lest top 10 prediction.

pred_class,pred_idx,outputs = learn.predict(img)

here pred_class is just single label. How to get top 10 probable label.