%matplotlib inline
%reload_ext autoreload
%autoreload 2

import os
os.chdir('..')

from nb_007 import *
from models import wrn_22

torch.backends.cudnn.benchmark = True

cifar_norm,cifar_denorm = normalize_funcs(*cifar_stats)

tfms = ([pad(padding=4), crop(size=32, row_pct=(0,1), col_pct=(0,1)), flip_lr(p=0.5)], [])
data = data_from_imagefolder(Path('data/cifar10'), valid='test', ds_tfms=tfms, tfms=cifar_norm)

learn = Learner(data, wrn_22(), metrics=accuracy).to_fp16()

learn.fit_one_cycle(25, 3e-3, wd=0.4, pct_start=0.45)

%time learn.fit_one_cycle(30, 3e-3, wd=0.4)

learn = Learner(data, wrn_22(), metrics=accuracy)

learn.fit_one_cycle(1, 3e-3, wd=0.4)