%reload_ext autoreload
%autoreload 2
%matplotlib inline

#export
from fastai import *
from fastai.vision import *
import fastai

tfms = get_transforms()
bs = 64
path = untar_data(URLs.PETS)/'images'

data = (ImageFileList.from_folder(path)
        .label_from_re(r'^(.*)_\d+.jpg$')
        .random_split_by_pct(0.2)
        .datasets(ImageClassificationDataset)
        .transform(tfms, size=224)
        .databunch(bs=bs)
        .normalize(imagenet_stats))

img = open_image(get_image_files(path)[0])
img.show()

data.show_batch(3)

learn = create_cnn(data, models.resnet34, metrics=accuracy)

# Run this once to create your model
# learn.fit_one_cycle(1)
# learn.save('one-epoch')

learn.load('one-epoch')

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

fastai.defaults.device = torch.device('cpu')

# You wouldn't use data.classes really
# - instead load your classes list from somewhere else
data2 = ImageDataBunch.single_from_classes(
    path, data.classes, tfms=tfms, size=224).normalize(imagenet_stats)
learn = create_cnn(data2, models.resnet34)
learn.load('one-epoch')

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