from fastai.vision.all import *
path = untar_data(URLs.PETS)/'images'
def is_cat(x): return x[0].isupper() 
dls = ImageDataLoaders.from_name_func('.',
    get_image_files(path), valid_pct=0.2, seed=42,
    label_func=is_cat,
    item_tfms=Resize(192))
learn = vision_learner(dls, resnet18, metrics=error_rate)
learn.fine_tune(3)
learn.export('model.pkl')