Do I still need data set even at the time of prediction?

I have gone through upto 3 lessons from part 1 and so I have pretty good trained model (Dog vs cat classifier) with me.
Now I’m running same on different machine but this time I’m reusing trained model basically learned weights.
So why I also need Data set for dog and cats at time of prediction.

data = ImageClassifierData.from_paths(PATH, tfms=tfms_from_model(arch, sz))
learn = ConvLearner.pretrained(arch,data, precompute=False)

What I want is to run this without copying same huge data into my new machine because I already have trained model so I should be able to predict from it without loading data set again.
Appreciate your help.

1 Like