# Image Prediction Randomness

The model is giving a different prediction for the same image. So what I did is that after creating Image databunch (data_p) I added that to the model (saved model (.pkl )). When is predict that using

phy.predict(phy.data.train_ds[0][0]) --> it gave 170 as category .

but when I tried phy.predict(data_p.train_ds[0][0]) --> it gave 160 as category .
Two different predictions for the same image. You can check it in the snapshot.
How this is working I want an explanation. @rachel @jeremy @sgugger