Resnet 50 and Darknet 53.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "%reload_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[PosixPath('../data/imagenet/train/n01644373'),\n",
       " PosixPath('../data/imagenet/train/n03791053'),\n",
       " PosixPath('../data/imagenet/train/n03109150'),\n",
       " PosixPath('../data/imagenet/train/n03100240'),\n",
       " PosixPath('../data/imagenet/train/n04131690'),\n",
       " PosixPath('../data/imagenet/train/n02951358'),\n",
       " PosixPath('../data/imagenet/train/n04162706'),\n",
       " PosixPath('../data/imagenet/train/n02091831'),\n",
       " PosixPath('../data/imagenet/train/n03954731'),\n",
       " PosixPath('../data/imagenet/train/n02391049'),\n",
       " PosixPath('../data/imagenet/train/n03761084'),\n",
       " PosixPath('../data/imagenet/train/n09835506'),\n",
       " PosixPath('../data/imagenet/train/n04554684'),\n",
       " PosixPath('../data/imagenet/train/n04532106'),\n",
       " PosixPath('../data/imagenet/train/n07693725'),\n",
       " PosixPath('../data/imagenet/train/n03935335'),\n",
       " PosixPath('../data/imagenet/train/n06785654'),\n",
       " PosixPath('../data/imagenet/train/n02859443'),\n",
       " PosixPath('../data/imagenet/train/n03355925'),\n",
       " PosixPath('../data/imagenet/train/n12267677'),\n",
       " PosixPath('../data/imagenet/train/n02870880'),\n",
       " PosixPath('../data/imagenet/train/n04243546'),\n",
       " PosixPath('../data/imagenet/train/n02643566'),\n",
       " PosixPath('../data/imagenet/train/n04228054'),\n",
       " PosixPath('../data/imagenet/train/n04606251'),\n",
       " PosixPath('../data/imagenet/train/n03131574'),\n",
       " PosixPath('../data/imagenet/train/n03345487'),\n",
       " PosixPath('../data/imagenet/train/n03259280'),\n",
       " PosixPath('../data/imagenet/train/n02105251'),\n",
       " PosixPath('../data/imagenet/train/n03787032'),\n",
       " PosixPath('../data/imagenet/train/n02492035'),\n",
       " PosixPath('../data/imagenet/train/n07565083'),\n",
       " PosixPath('../data/imagenet/train/n03840681'),\n",
       " PosixPath('../data/imagenet/train/n02837789'),\n",
       " PosixPath('../data/imagenet/train/n01833805'),\n",
       " PosixPath('../data/imagenet/train/n02117135'),\n",
       " PosixPath('../data/imagenet/train/n04023962'),\n",
       " PosixPath('../data/imagenet/train/n04487394'),\n",
       " PosixPath('../data/imagenet/train/n01986214'),\n",
       " PosixPath('../data/imagenet/train/n03250847'),\n",
       " PosixPath('../data/imagenet/train/n02906734'),\n",
       " PosixPath('../data/imagenet/train/n02769748'),\n",
       " PosixPath('../data/imagenet/train/n02128757'),\n",
       " PosixPath('../data/imagenet/train/n01756291'),\n",
       " PosixPath('../data/imagenet/train/n02111889'),\n",
       " PosixPath('../data/imagenet/train/n02100583'),\n",
       " PosixPath('../data/imagenet/train/n02177972'),\n",
       " PosixPath('../data/imagenet/train/n04033995'),\n",
       " PosixPath('../data/imagenet/train/n03692522'),\n",
       " PosixPath('../data/imagenet/train/n03376595'),\n",
       " PosixPath('../data/imagenet/train/n03786901'),\n",
       " PosixPath('../data/imagenet/train/n04275548'),\n",
       " PosixPath('../data/imagenet/train/n03642806'),\n",
       " PosixPath('../data/imagenet/train/n02092002'),\n",
       " PosixPath('../data/imagenet/train/n03045698'),\n",
       " PosixPath('../data/imagenet/train/n03814906'),\n",
       " PosixPath('../data/imagenet/train/n04335435'),\n",
       " PosixPath('../data/imagenet/train/n02797295'),\n",
       " PosixPath('../data/imagenet/train/n01860187'),\n",
       " PosixPath('../data/imagenet/train/n03063599'),\n",
       " PosixPath('../data/imagenet/train/n07880968'),\n",
       " PosixPath('../data/imagenet/train/n04009552'),\n",
       " PosixPath('../data/imagenet/train/n03670208'),\n",
       " PosixPath('../data/imagenet/train/n02894605'),\n",
       " PosixPath('../data/imagenet/train/n03773504'),\n",
       " PosixPath('../data/imagenet/train/n03447721'),\n",
       " PosixPath('../data/imagenet/train/n03666591'),\n",
       " PosixPath('../data/imagenet/train/n02486261'),\n",
       " PosixPath('../data/imagenet/train/n09332890'),\n",
       " PosixPath('../data/imagenet/train/n04239074'),\n",
       " PosixPath('../data/imagenet/train/n02099601'),\n",
       " PosixPath('../data/imagenet/train/n01632458'),\n",
       " PosixPath('../data/imagenet/train/n04065272'),\n",
       " PosixPath('../data/imagenet/train/n02804414'),\n",
       " PosixPath('../data/imagenet/train/n02174001'),\n",
       " PosixPath('../data/imagenet/train/n03594734'),\n",
       " PosixPath('../data/imagenet/train/n04536866'),\n",
       " PosixPath('../data/imagenet/train/n03498962'),\n",
       " PosixPath('../data/imagenet/train/n02098413'),\n",
       " PosixPath('../data/imagenet/train/n02114855'),\n",
       " PosixPath('../data/imagenet/train/n03838899'),\n",
       " PosixPath('../data/imagenet/train/n07730033'),\n",
       " PosixPath('../data/imagenet/train/n02096177'),\n",
       " PosixPath('../data/imagenet/train/n02749479'),\n",
       " PosixPath('../data/imagenet/train/n04404412'),\n",
       " PosixPath('../data/imagenet/train/n03673027'),\n",
       " PosixPath('../data/imagenet/train/n02086910'),\n",
       " PosixPath('../data/imagenet/train/n04254120'),\n",
       " PosixPath('../data/imagenet/train/n02687172'),\n",
       " PosixPath('../data/imagenet/train/n02233338'),\n",
       " PosixPath('../data/imagenet/train/n04040759'),\n",
       " PosixPath('../data/imagenet/train/n04229816'),\n",
       " PosixPath('../data/imagenet/train/n02480855'),\n",
       " PosixPath('../data/imagenet/train/n04389033'),\n",
       " PosixPath('../data/imagenet/train/n04005630'),\n",
       " PosixPath('../data/imagenet/train/n02422699'),\n",
       " PosixPath('../data/imagenet/train/n04133789'),\n",
       " PosixPath('../data/imagenet/train/n01806567'),\n",
       " PosixPath('../data/imagenet/train/n03877472'),\n",
       " PosixPath('../data/imagenet/train/n02422106'),\n",
       " PosixPath('../data/imagenet/train/n04070727'),\n",
       " PosixPath('../data/imagenet/train/n02793495'),\n",
       " PosixPath('../data/imagenet/train/n02488702'),\n",
       " PosixPath('../data/imagenet/train/n04146614'),\n",
       " PosixPath('../data/imagenet/train/n03459775'),\n",
       " PosixPath('../data/imagenet/train/n01843383'),\n",
       " PosixPath('../data/imagenet/train/n02113799'),\n",
       " PosixPath('../data/imagenet/train/n04252225'),\n",
       " PosixPath('../data/imagenet/train/n07892512'),\n",
       " PosixPath('../data/imagenet/train/n04505470'),\n",
       " PosixPath('../data/imagenet/train/n04597913'),\n",
       " PosixPath('../data/imagenet/train/n02107908'),\n",
       " PosixPath('../data/imagenet/train/n02489166'),\n",
       " PosixPath('../data/imagenet/train/n02097047'),\n",
       " PosixPath('../data/imagenet/train/n03372029'),\n",
       " PosixPath('../data/imagenet/train/n02229544'),\n",
       " PosixPath('../data/imagenet/train/n02966193'),\n",
       " PosixPath('../data/imagenet/train/n04004767'),\n",
       " PosixPath('../data/imagenet/train/n04067472'),\n",
       " PosixPath('../data/imagenet/train/n04347754'),\n",
       " PosixPath('../data/imagenet/train/n03724870'),\n",
       " PosixPath('../data/imagenet/train/n01704323'),\n",
       " PosixPath('../data/imagenet/train/n04336792'),\n",
       " PosixPath('../data/imagenet/train/n04019541'),\n",
       " PosixPath('../data/imagenet/train/n03065424'),\n",
       " PosixPath('../data/imagenet/train/n02444819'),\n",
       " PosixPath('../data/imagenet/train/n02092339'),\n",
       " PosixPath('../data/imagenet/train/n07716358'),\n",
       " PosixPath('../data/imagenet/train/n10148035'),\n",
       " PosixPath('../data/imagenet/train/n02123597'),\n",
       " PosixPath('../data/imagenet/train/n09468604'),\n",
       " PosixPath('../data/imagenet/train/n03770679'),\n",
       " PosixPath('../data/imagenet/train/n02236044'),\n",
       " PosixPath('../data/imagenet/train/n02226429'),\n",
       " PosixPath('../data/imagenet/train/n03297495'),\n",
       " PosixPath('../data/imagenet/train/n04259630'),\n",
       " PosixPath('../data/imagenet/train/n02165105'),\n",
       " PosixPath('../data/imagenet/train/n02950826'),\n",
       " PosixPath('../data/imagenet/train/n03461385'),\n",
       " PosixPath('../data/imagenet/train/n02704792'),\n",
       " PosixPath('../data/imagenet/train/n07734744'),\n",
       " PosixPath('../data/imagenet/train/n03709823'),\n",
       " PosixPath('../data/imagenet/train/n04111531'),\n",
       " PosixPath('../data/imagenet/train/n02106030'),\n",
       " PosixPath('../data/imagenet/train/n02110627'),\n",
       " PosixPath('../data/imagenet/train/n09256479'),\n",
       " PosixPath('../data/imagenet/train/n03841143'),\n",
       " PosixPath('../data/imagenet/train/n03871628'),\n",
       " PosixPath('../data/imagenet/train/n04074963'),\n",
       " PosixPath('../data/imagenet/train/n02111129'),\n",
       " PosixPath('../data/imagenet/train/n03452741'),\n",
       " PosixPath('../data/imagenet/train/n03649909'),\n",
       " PosixPath('../data/imagenet/train/n04154565'),\n",
       " PosixPath('../data/imagenet/train/n02219486'),\n",
       " PosixPath('../data/imagenet/train/n02321529'),\n",
       " PosixPath('../data/imagenet/train/n07932039'),\n",
       " PosixPath('../data/imagenet/train/n02091244'),\n",
       " PosixPath('../data/imagenet/train/n03794056'),\n",
       " PosixPath('../data/imagenet/train/n04273569'),\n",
       " PosixPath('../data/imagenet/train/n01828970'),\n",
       " PosixPath('../data/imagenet/train/n01807496'),\n",
       " PosixPath('../data/imagenet/train/n04552348'),\n",
       " PosixPath('../data/imagenet/train/n02484975'),\n",
       " PosixPath('../data/imagenet/train/n01729322'),\n",
       " PosixPath('../data/imagenet/train/n02107312'),\n",
       " PosixPath('../data/imagenet/train/n03888605'),\n",
       " PosixPath('../data/imagenet/train/n01943899'),\n",
       " PosixPath('../data/imagenet/train/n02087394'),\n",
       " PosixPath('../data/imagenet/train/n03743016'),\n",
       " PosixPath('../data/imagenet/train/n03950228'),\n",
       " PosixPath('../data/imagenet/train/n03877845'),\n",
       " PosixPath('../data/imagenet/train/n01592084'),\n",
       " PosixPath('../data/imagenet/train/n04579432'),\n",
       " PosixPath('../data/imagenet/train/n02268853'),\n",
       " PosixPath('../data/imagenet/train/n04612504'),\n",
       " PosixPath('../data/imagenet/train/n01818515'),\n",
       " PosixPath('../data/imagenet/train/n02397096'),\n",
       " PosixPath('../data/imagenet/train/n03803284'),\n",
       " PosixPath('../data/imagenet/train/n03884397'),\n",
       " PosixPath('../data/imagenet/train/n03188531'),\n",
       " PosixPath('../data/imagenet/train/n01518878'),\n",
       " PosixPath('../data/imagenet/train/n04548362'),\n",
       " PosixPath('../data/imagenet/train/n03814639'),\n",
       " PosixPath('../data/imagenet/train/n02099712'),\n",
       " PosixPath('../data/imagenet/train/n02091467'),\n",
       " PosixPath('../data/imagenet/train/n04311174'),\n",
       " PosixPath('../data/imagenet/train/n04254777'),\n",
       " PosixPath('../data/imagenet/train/n03680355'),\n",
       " PosixPath('../data/imagenet/train/n03388043'),\n",
       " PosixPath('../data/imagenet/train/n03467068'),\n",
       " PosixPath('../data/imagenet/train/n01883070'),\n",
       " PosixPath('../data/imagenet/train/n02791124'),\n",
       " PosixPath('../data/imagenet/train/n04235860'),\n",
       " PosixPath('../data/imagenet/train/n01980166'),\n",
       " PosixPath('../data/imagenet/train/n03527444'),\n",
       " PosixPath('../data/imagenet/train/n01665541'),\n",
       " PosixPath('../data/imagenet/train/n02018795'),\n",
       " PosixPath('../data/imagenet/train/n02441942'),\n",
       " PosixPath('../data/imagenet/train/n02101388'),\n",
       " PosixPath('../data/imagenet/train/n02871525'),\n",
       " PosixPath('../data/imagenet/train/n02110341'),\n",
       " PosixPath('../data/imagenet/train/n03141823'),\n",
       " PosixPath('../data/imagenet/train/n02808304'),\n",
       " PosixPath('../data/imagenet/train/n02115913'),\n",
       " PosixPath('../data/imagenet/train/n02094433'),\n",
       " PosixPath('../data/imagenet/train/n02115641'),\n",
       " PosixPath('../data/imagenet/train/n01616318'),\n",
       " PosixPath('../data/imagenet/train/n02699494'),\n",
       " PosixPath('../data/imagenet/train/n02110063'),\n",
       " PosixPath('../data/imagenet/train/n04507155'),\n",
       " PosixPath('../data/imagenet/train/n07875152'),\n",
       " PosixPath('../data/imagenet/train/n03944341'),\n",
       " PosixPath('../data/imagenet/train/n03532672'),\n",
       " PosixPath('../data/imagenet/train/n03710637'),\n",
       " PosixPath('../data/imagenet/train/n01978287'),\n",
       " PosixPath('../data/imagenet/train/n02169497'),\n",
       " PosixPath('../data/imagenet/train/n03447447'),\n",
       " PosixPath('../data/imagenet/train/n02445715'),\n",
       " PosixPath('../data/imagenet/train/n07583066'),\n",
       " PosixPath('../data/imagenet/train/n02783161'),\n",
       " PosixPath('../data/imagenet/train/n02447366'),\n",
       " PosixPath('../data/imagenet/train/n02794156'),\n",
       " PosixPath('../data/imagenet/train/n03249569'),\n",
       " PosixPath('../data/imagenet/train/n09288635'),\n",
       " PosixPath('../data/imagenet/train/n02971356'),\n",
       " PosixPath('../data/imagenet/train/n02088094'),\n",
       " PosixPath('../data/imagenet/train/n04525305'),\n",
       " PosixPath('../data/imagenet/train/n03207743'),\n",
       " PosixPath('../data/imagenet/train/n02799071'),\n",
       " PosixPath('../data/imagenet/train/n01968897'),\n",
       " PosixPath('../data/imagenet/train/n02879718'),\n",
       " PosixPath('../data/imagenet/train/n07613480'),\n",
       " PosixPath('../data/imagenet/train/n02102480'),\n",
       " PosixPath('../data/imagenet/train/n04540053'),\n",
       " PosixPath('../data/imagenet/train/n02483708'),\n",
       " PosixPath('../data/imagenet/train/n02747177'),\n",
       " PosixPath('../data/imagenet/train/n04584207'),\n",
       " PosixPath('../data/imagenet/train/n02692877'),\n",
       " PosixPath('../data/imagenet/train/n02701002'),\n",
       " PosixPath('../data/imagenet/train/n04147183'),\n",
       " PosixPath('../data/imagenet/train/n02395406'),\n",
       " PosixPath('../data/imagenet/train/n04370456'),\n",
       " PosixPath('../data/imagenet/train/n04486054'),\n",
       " PosixPath('../data/imagenet/train/n01667778'),\n",
       " PosixPath('../data/imagenet/train/n04367480'),\n",
       " PosixPath('../data/imagenet/train/n09472597'),\n",
       " PosixPath('../data/imagenet/train/n01530575'),\n",
       " PosixPath('../data/imagenet/train/n02095889'),\n",
       " PosixPath('../data/imagenet/train/n04376876'),\n",
       " PosixPath('../data/imagenet/train/n04350905'),\n",
       " PosixPath('../data/imagenet/train/n02085620'),\n",
       " PosixPath('../data/imagenet/train/n03062245'),\n",
       " PosixPath('../data/imagenet/train/n02412080'),\n",
       " PosixPath('../data/imagenet/train/n03595614'),\n",
       " PosixPath('../data/imagenet/train/n02834397'),\n",
       " PosixPath('../data/imagenet/train/n04456115'),\n",
       " PosixPath('../data/imagenet/train/n02892201'),\n",
       " PosixPath('../data/imagenet/train/n07717410'),\n",
       " PosixPath('../data/imagenet/train/n02500267'),\n",
       " PosixPath('../data/imagenet/train/n01695060'),\n",
       " PosixPath('../data/imagenet/train/n03404251'),\n",
       " PosixPath('../data/imagenet/train/n02101556'),\n",
       " PosixPath('../data/imagenet/train/n02483362'),\n",
       " PosixPath('../data/imagenet/train/n01641577'),\n",
       " PosixPath('../data/imagenet/train/n02125311'),\n",
       " PosixPath('../data/imagenet/train/n02088632'),\n",
       " PosixPath('../data/imagenet/train/n07614500'),\n",
       " PosixPath('../data/imagenet/train/n02106166'),\n",
       " PosixPath('../data/imagenet/train/n02326432'),\n",
       " PosixPath('../data/imagenet/train/n02787622'),\n",
       " PosixPath('../data/imagenet/train/n04039381'),\n",
       " PosixPath('../data/imagenet/train/n02093647'),\n",
       " PosixPath('../data/imagenet/train/n03063689'),\n",
       " PosixPath('../data/imagenet/train/n03481172'),\n",
       " PosixPath('../data/imagenet/train/n03920288'),\n",
       " PosixPath('../data/imagenet/train/n01798484'),\n",
       " PosixPath('../data/imagenet/train/n02317335'),\n",
       " PosixPath('../data/imagenet/train/n04399382'),\n",
       " PosixPath('../data/imagenet/train/n07802026'),\n",
       " PosixPath('../data/imagenet/train/n02007558'),\n",
       " PosixPath('../data/imagenet/train/n04372370'),\n",
       " PosixPath('../data/imagenet/train/n03017168'),\n",
       " PosixPath('../data/imagenet/train/n03476991'),\n",
       " PosixPath('../data/imagenet/train/n03792782'),\n",
       " PosixPath('../data/imagenet/train/n04330267'),\n",
       " PosixPath('../data/imagenet/train/n02097130'),\n",
       " PosixPath('../data/imagenet/train/n07836838'),\n",
       " PosixPath('../data/imagenet/train/n04485082'),\n",
       " PosixPath('../data/imagenet/train/n07584110'),\n",
       " PosixPath('../data/imagenet/train/n07745940'),\n",
       " PosixPath('../data/imagenet/train/n04252077'),\n",
       " PosixPath('../data/imagenet/train/n02096437'),\n",
       " PosixPath('../data/imagenet/train/n02206856'),\n",
       " PosixPath('../data/imagenet/train/n01494475'),\n",
       " PosixPath('../data/imagenet/train/n02107683'),\n",
       " PosixPath('../data/imagenet/train/n04355933'),\n",
       " PosixPath('../data/imagenet/train/n02992529'),\n",
       " PosixPath('../data/imagenet/train/n13133613'),\n",
       " PosixPath('../data/imagenet/train/n03627232'),\n",
       " PosixPath('../data/imagenet/train/n04447861'),\n",
       " PosixPath('../data/imagenet/train/n12144580'),\n",
       " PosixPath('../data/imagenet/train/n02790996'),\n",
       " PosixPath('../data/imagenet/train/n01930112'),\n",
       " PosixPath('../data/imagenet/train/n03594945'),\n",
       " PosixPath('../data/imagenet/train/n01955084'),\n",
       " PosixPath('../data/imagenet/train/n03933933'),\n",
       " PosixPath('../data/imagenet/train/n07579787'),\n",
       " PosixPath('../data/imagenet/train/n01768244'),\n",
       " PosixPath('../data/imagenet/train/n03482405'),\n",
       " PosixPath('../data/imagenet/train/n02128925'),\n",
       " PosixPath('../data/imagenet/train/n02980441'),\n",
       " PosixPath('../data/imagenet/train/n04371774'),\n",
       " PosixPath('../data/imagenet/train/n03924679'),\n",
       " PosixPath('../data/imagenet/train/n09246464'),\n",
       " PosixPath('../data/imagenet/train/n02113186'),\n",
       " PosixPath('../data/imagenet/train/n01514668'),\n",
       " PosixPath('../data/imagenet/train/n07742313'),\n",
       " PosixPath('../data/imagenet/train/n02138441'),\n",
       " PosixPath('../data/imagenet/train/n02398521'),\n",
       " PosixPath('../data/imagenet/train/n07920052'),\n",
       " PosixPath('../data/imagenet/train/n03720891'),\n",
       " PosixPath('../data/imagenet/train/n04493381'),\n",
       " PosixPath('../data/imagenet/train/n04346328'),\n",
       " PosixPath('../data/imagenet/train/n03937543'),\n",
       " PosixPath('../data/imagenet/train/n03032252'),\n",
       " PosixPath('../data/imagenet/train/n02074367'),\n",
       " PosixPath('../data/imagenet/train/n02051845'),\n",
       " PosixPath('../data/imagenet/train/n02815834'),\n",
       " PosixPath('../data/imagenet/train/n01560419'),\n",
       " PosixPath('../data/imagenet/train/n03218198'),\n",
       " PosixPath('../data/imagenet/train/n02415577'),\n",
       " PosixPath('../data/imagenet/train/n03775071'),\n",
       " PosixPath('../data/imagenet/train/n04328186'),\n",
       " PosixPath('../data/imagenet/train/n02088466'),\n",
       " PosixPath('../data/imagenet/train/n01981276'),\n",
       " PosixPath('../data/imagenet/train/n02093428'),\n",
       " PosixPath('../data/imagenet/train/n07871810'),\n",
       " PosixPath('../data/imagenet/train/n02676566'),\n",
       " PosixPath('../data/imagenet/train/n01945685'),\n",
       " PosixPath('../data/imagenet/train/n03956157'),\n",
       " PosixPath('../data/imagenet/train/n02487347'),\n",
       " PosixPath('../data/imagenet/train/n04086273'),\n",
       " PosixPath('../data/imagenet/train/n01580077'),\n",
       " PosixPath('../data/imagenet/train/n03891332'),\n",
       " PosixPath('../data/imagenet/train/n03662601'),\n",
       " PosixPath('../data/imagenet/train/n02106662'),\n",
       " PosixPath('../data/imagenet/train/n04461696'),\n",
       " PosixPath('../data/imagenet/train/n02641379'),\n",
       " PosixPath('../data/imagenet/train/n03793489'),\n",
       " PosixPath('../data/imagenet/train/n03124043'),\n",
       " PosixPath('../data/imagenet/train/n03729826'),\n",
       " PosixPath('../data/imagenet/train/n02417914'),\n",
       " PosixPath('../data/imagenet/train/n03041632'),\n",
       " PosixPath('../data/imagenet/train/n02490219'),\n",
       " PosixPath('../data/imagenet/train/n02028035'),\n",
       " PosixPath('../data/imagenet/train/n02492660'),\n",
       " PosixPath('../data/imagenet/train/n02795169'),\n",
       " PosixPath('../data/imagenet/train/n01796340'),\n",
       " PosixPath('../data/imagenet/train/n01749939'),\n",
       " PosixPath('../data/imagenet/train/n02782093'),\n",
       " PosixPath('../data/imagenet/train/n04136333'),\n",
       " PosixPath('../data/imagenet/train/n03903868'),\n",
       " PosixPath('../data/imagenet/train/n04026417'),\n",
       " PosixPath('../data/imagenet/train/n07711569'),\n",
       " PosixPath('../data/imagenet/train/n02112018'),\n",
       " PosixPath('../data/imagenet/train/n02130308'),\n",
       " PosixPath('../data/imagenet/train/n04179913'),\n",
       " PosixPath('../data/imagenet/train/n11879895'),\n",
       " PosixPath('../data/imagenet/train/n02114367'),\n",
       " PosixPath('../data/imagenet/train/n03388183'),\n",
       " PosixPath('../data/imagenet/train/n03916031'),\n",
       " PosixPath('../data/imagenet/train/n02108089'),\n",
       " PosixPath('../data/imagenet/train/n01614925'),\n",
       " PosixPath('../data/imagenet/train/n03124170'),\n",
       " PosixPath('../data/imagenet/train/n03617480'),\n",
       " PosixPath('../data/imagenet/train/n03018349'),\n",
       " PosixPath('../data/imagenet/train/n03272010'),\n",
       " PosixPath('../data/imagenet/train/n03207941'),\n",
       " PosixPath('../data/imagenet/train/n03785016'),\n",
       " PosixPath('../data/imagenet/train/n02356798'),\n",
       " PosixPath('../data/imagenet/train/n01776313'),\n",
       " PosixPath('../data/imagenet/train/n03676483'),\n",
       " PosixPath('../data/imagenet/train/n02917067'),\n",
       " PosixPath('../data/imagenet/train/n03733281'),\n",
       " PosixPath('../data/imagenet/train/n07873807'),\n",
       " PosixPath('../data/imagenet/train/n02437616'),\n",
       " PosixPath('../data/imagenet/train/n04325704'),\n",
       " PosixPath('../data/imagenet/train/n04371430'),\n",
       " PosixPath('../data/imagenet/train/n04201297'),\n",
       " PosixPath('../data/imagenet/train/n04008634'),\n",
       " PosixPath('../data/imagenet/train/n04429376'),\n",
       " PosixPath('../data/imagenet/train/n04033901'),\n",
       " PosixPath('../data/imagenet/train/n01984695'),\n",
       " PosixPath('../data/imagenet/train/n02979186'),\n",
       " PosixPath('../data/imagenet/train/n01847000'),\n",
       " PosixPath('../data/imagenet/train/n03887697'),\n",
       " PosixPath('../data/imagenet/train/n02137549'),\n",
       " PosixPath('../data/imagenet/train/n01692333'),\n",
       " PosixPath('../data/imagenet/train/n01582220'),\n",
       " PosixPath('../data/imagenet/train/n09229709'),\n",
       " PosixPath('../data/imagenet/train/n01755581'),\n",
       " PosixPath('../data/imagenet/train/n02939185'),\n",
       " PosixPath('../data/imagenet/train/n07831146'),\n",
       " PosixPath('../data/imagenet/train/n04285008'),\n",
       " PosixPath('../data/imagenet/train/n03110669'),\n",
       " PosixPath('../data/imagenet/train/n02077923'),\n",
       " PosixPath('../data/imagenet/train/n04332243'),\n",
       " PosixPath('../data/imagenet/train/n09421951'),\n",
       " PosixPath('../data/imagenet/train/n01806143'),\n",
       " PosixPath('../data/imagenet/train/n10565667'),\n",
       " PosixPath('../data/imagenet/train/n02823428'),\n",
       " PosixPath('../data/imagenet/train/n01685808'),\n",
       " PosixPath('../data/imagenet/train/n04515003'),\n",
       " PosixPath('../data/imagenet/train/n04517823'),\n",
       " PosixPath('../data/imagenet/train/n02190166'),\n",
       " PosixPath('../data/imagenet/train/n02493509'),\n",
       " PosixPath('../data/imagenet/train/n02988304'),\n",
       " PosixPath('../data/imagenet/train/n02111277'),\n",
       " PosixPath('../data/imagenet/train/n04204347'),\n",
       " PosixPath('../data/imagenet/train/n02396427'),\n",
       " PosixPath('../data/imagenet/train/n01873310'),\n",
       " PosixPath('../data/imagenet/train/n03958227'),\n",
       " PosixPath('../data/imagenet/train/n02916936'),\n",
       " PosixPath('../data/imagenet/train/n02840245'),\n",
       " PosixPath('../data/imagenet/train/n02127052'),\n",
       " PosixPath('../data/imagenet/train/n03272562'),\n",
       " PosixPath('../data/imagenet/train/n04081281'),\n",
       " PosixPath('../data/imagenet/train/n02804610'),\n",
       " PosixPath('../data/imagenet/train/n03127925'),\n",
       " PosixPath('../data/imagenet/train/n03788195'),\n",
       " PosixPath('../data/imagenet/train/n04501370'),\n",
       " PosixPath('../data/imagenet/train/n03804744'),\n",
       " PosixPath('../data/imagenet/train/n03769881'),\n",
       " PosixPath('../data/imagenet/train/n01872401'),\n",
       " PosixPath('../data/imagenet/train/n07753592'),\n",
       " PosixPath('../data/imagenet/train/n03690938'),\n",
       " PosixPath('../data/imagenet/train/n04099969'),\n",
       " PosixPath('../data/imagenet/train/n02992211'),\n",
       " PosixPath('../data/imagenet/train/n04522168'),\n",
       " PosixPath('../data/imagenet/train/n07930864'),\n",
       " PosixPath('../data/imagenet/train/n02389026'),\n",
       " PosixPath('../data/imagenet/train/n02058221'),\n",
       " PosixPath('../data/imagenet/train/n01742172'),\n",
       " PosixPath('../data/imagenet/train/n04476259'),\n",
       " PosixPath('../data/imagenet/train/n03530642'),\n",
       " PosixPath('../data/imagenet/train/n02106382'),\n",
       " PosixPath('../data/imagenet/train/n04465501'),\n",
       " PosixPath('../data/imagenet/train/n03980874'),\n",
       " PosixPath('../data/imagenet/train/n01820546'),\n",
       " PosixPath('../data/imagenet/train/n01855672'),\n",
       " PosixPath('../data/imagenet/train/n03976657'),\n",
       " PosixPath('../data/imagenet/train/n04479046'),\n",
       " PosixPath('../data/imagenet/train/n01664065'),\n",
       " PosixPath('../data/imagenet/train/n02264363'),\n",
       " PosixPath('../data/imagenet/train/n03450230'),\n",
       " PosixPath('../data/imagenet/train/n04200800'),\n",
       " PosixPath('../data/imagenet/train/n03485794'),\n",
       " PosixPath('../data/imagenet/train/n02090622'),\n",
       " PosixPath('../data/imagenet/train/n02526121'),\n",
       " PosixPath('../data/imagenet/train/n02319095'),\n",
       " PosixPath('../data/imagenet/train/n03854065'),\n",
       " PosixPath('../data/imagenet/train/n02088238'),\n",
       " PosixPath('../data/imagenet/train/n02256656'),\n",
       " PosixPath('../data/imagenet/train/n02869837'),\n",
       " PosixPath('../data/imagenet/train/n03337140'),\n",
       " PosixPath('../data/imagenet/train/n02814860'),\n",
       " PosixPath('../data/imagenet/train/n07697313'),\n",
       " PosixPath('../data/imagenet/train/n03982430'),\n",
       " PosixPath('../data/imagenet/train/n03976467'),\n",
       " PosixPath('../data/imagenet/train/n02788148'),\n",
       " PosixPath('../data/imagenet/train/n03908618'),\n",
       " PosixPath('../data/imagenet/train/n04462240'),\n",
       " PosixPath('../data/imagenet/train/n03457902'),\n",
       " PosixPath('../data/imagenet/train/n03584254'),\n",
       " PosixPath('../data/imagenet/train/n07749582'),\n",
       " PosixPath('../data/imagenet/train/n04483307'),\n",
       " PosixPath('../data/imagenet/train/n03930313'),\n",
       " PosixPath('../data/imagenet/train/n02814533'),\n",
       " PosixPath('../data/imagenet/train/n04553703'),\n",
       " PosixPath('../data/imagenet/train/n03496892'),\n",
       " PosixPath('../data/imagenet/train/n02808440'),\n",
       " PosixPath('../data/imagenet/train/n02442845'),\n",
       " PosixPath('../data/imagenet/train/n03291819'),\n",
       " PosixPath('../data/imagenet/train/n04209239'),\n",
       " PosixPath('../data/imagenet/train/n02328150'),\n",
       " PosixPath('../data/imagenet/train/n02655020'),\n",
       " PosixPath('../data/imagenet/train/n02111500'),\n",
       " PosixPath('../data/imagenet/train/n03000134'),\n",
       " PosixPath('../data/imagenet/train/n03825788'),\n",
       " PosixPath('../data/imagenet/train/n03126707'),\n",
       " PosixPath('../data/imagenet/train/n03028079'),\n",
       " PosixPath('../data/imagenet/train/n03759954'),\n",
       " PosixPath('../data/imagenet/train/n04423845'),\n",
       " PosixPath('../data/imagenet/train/n01689811'),\n",
       " PosixPath('../data/imagenet/train/n04149813'),\n",
       " PosixPath('../data/imagenet/train/n03866082'),\n",
       " PosixPath('../data/imagenet/train/n03216828'),\n",
       " PosixPath('../data/imagenet/train/n04599235'),\n",
       " PosixPath('../data/imagenet/train/n02129165'),\n",
       " PosixPath('../data/imagenet/train/n04592741'),\n",
       " PosixPath('../data/imagenet/train/n04317175'),\n",
       " PosixPath('../data/imagenet/train/n01877812'),\n",
       " PosixPath('../data/imagenet/train/n02105855'),\n",
       " PosixPath('../data/imagenet/train/n07753113'),\n",
       " PosixPath('../data/imagenet/train/n03837869'),\n",
       " PosixPath('../data/imagenet/train/n02480495'),\n",
       " PosixPath('../data/imagenet/train/n03717622'),\n",
       " PosixPath('../data/imagenet/train/n13054560'),\n",
       " PosixPath('../data/imagenet/train/n07697537'),\n",
       " PosixPath('../data/imagenet/train/n03417042'),\n",
       " PosixPath('../data/imagenet/train/n02841315'),\n",
       " PosixPath('../data/imagenet/train/n07716906'),\n",
       " PosixPath('../data/imagenet/train/n03888257'),\n",
       " PosixPath('../data/imagenet/train/n02165456'),\n",
       " PosixPath('../data/imagenet/train/n04310018'),\n",
       " PosixPath('../data/imagenet/train/n03733131'),\n",
       " PosixPath('../data/imagenet/train/n02361337'),\n",
       " PosixPath('../data/imagenet/train/n04141327'),\n",
       " PosixPath('../data/imagenet/train/n01871265'),\n",
       " PosixPath('../data/imagenet/train/n02276258'),\n",
       " PosixPath('../data/imagenet/train/n04277352'),\n",
       " PosixPath('../data/imagenet/train/n03868863'),\n",
       " PosixPath('../data/imagenet/train/n03495258'),\n",
       " PosixPath('../data/imagenet/train/n03602883'),\n",
       " PosixPath('../data/imagenet/train/n02112137'),\n",
       " PosixPath('../data/imagenet/train/n04208210'),\n",
       " PosixPath('../data/imagenet/train/n04049303'),\n",
       " PosixPath('../data/imagenet/train/n03721384'),\n",
       " PosixPath('../data/imagenet/train/n02927161'),\n",
       " PosixPath('../data/imagenet/train/n01914609'),\n",
       " PosixPath('../data/imagenet/train/n02883205'),\n",
       " PosixPath('../data/imagenet/train/n03895866'),\n",
       " PosixPath('../data/imagenet/train/n01829413'),\n",
       " PosixPath('../data/imagenet/train/n04116512'),\n",
       " PosixPath('../data/imagenet/train/n02172182'),\n",
       " PosixPath('../data/imagenet/train/n02403003'),\n",
       " PosixPath('../data/imagenet/train/n03529860'),\n",
       " PosixPath('../data/imagenet/train/n02110958'),\n",
       " PosixPath('../data/imagenet/train/n04428191'),\n",
       " PosixPath('../data/imagenet/train/n06359193'),\n",
       " PosixPath('../data/imagenet/train/n02948072'),\n",
       " PosixPath('../data/imagenet/train/n04270147'),\n",
       " PosixPath('../data/imagenet/train/n03133878'),\n",
       " PosixPath('../data/imagenet/train/n04487081'),\n",
       " PosixPath('../data/imagenet/train/n07760859'),\n",
       " PosixPath('../data/imagenet/train/n09428293'),\n",
       " PosixPath('../data/imagenet/train/n12768682'),\n",
       " PosixPath('../data/imagenet/train/n02510455'),\n",
       " PosixPath('../data/imagenet/train/n02493793'),\n",
       " PosixPath('../data/imagenet/train/n02093754'),\n",
       " PosixPath('../data/imagenet/train/n03991062'),\n",
       " PosixPath('../data/imagenet/train/n03544143'),\n",
       " PosixPath('../data/imagenet/train/n03930630'),\n",
       " PosixPath('../data/imagenet/train/n03832673'),\n",
       " PosixPath('../data/imagenet/train/n02892767'),\n",
       " PosixPath('../data/imagenet/train/n04509417'),\n",
       " PosixPath('../data/imagenet/train/n03857828'),\n",
       " PosixPath('../data/imagenet/train/n07695742'),\n",
       " PosixPath('../data/imagenet/train/n03782006'),\n",
       " PosixPath('../data/imagenet/train/n03899768'),\n",
       " PosixPath('../data/imagenet/train/n04141076'),\n",
       " PosixPath('../data/imagenet/train/n02978881'),\n",
       " PosixPath('../data/imagenet/train/n01498041'),\n",
       " PosixPath('../data/imagenet/train/n02105162'),\n",
       " PosixPath('../data/imagenet/train/n04523525'),\n",
       " PosixPath('../data/imagenet/train/n03443371'),\n",
       " PosixPath('../data/imagenet/train/n02363005'),\n",
       " PosixPath('../data/imagenet/train/n01534433'),\n",
       " PosixPath('../data/imagenet/train/n02999410'),\n",
       " PosixPath('../data/imagenet/train/n04596742'),\n",
       " PosixPath('../data/imagenet/train/n02259212'),\n",
       " PosixPath('../data/imagenet/train/n07718472'),\n",
       " PosixPath('../data/imagenet/train/n02100877'),\n",
       " PosixPath('../data/imagenet/train/n02364673'),\n",
       " PosixPath('../data/imagenet/train/n02277742'),\n",
       " PosixPath('../data/imagenet/train/n02120079'),\n",
       " PosixPath('../data/imagenet/train/n01882714'),\n",
       " PosixPath('../data/imagenet/train/n01990800'),\n",
       " PosixPath('../data/imagenet/train/n02113023'),\n",
       " PosixPath('../data/imagenet/train/n02006656'),\n",
       " PosixPath('../data/imagenet/train/n02108422'),\n",
       " PosixPath('../data/imagenet/train/n02128385'),\n",
       " PosixPath('../data/imagenet/train/n04482393'),\n",
       " PosixPath('../data/imagenet/train/n02860847'),\n",
       " PosixPath('../data/imagenet/train/n09399592'),\n",
       " PosixPath('../data/imagenet/train/n02776631'),\n",
       " PosixPath('../data/imagenet/train/n07717556'),\n",
       " PosixPath('../data/imagenet/train/n04037443'),\n",
       " PosixPath('../data/imagenet/train/n03947888'),\n",
       " PosixPath('../data/imagenet/train/n01985128'),\n",
       " PosixPath('../data/imagenet/train/n03445777'),\n",
       " PosixPath('../data/imagenet/train/n03697007'),\n",
       " PosixPath('../data/imagenet/train/n02667093'),\n",
       " PosixPath('../data/imagenet/train/n03874599'),\n",
       " PosixPath('../data/imagenet/train/n02094114'),\n",
       " PosixPath('../data/imagenet/train/n02895154'),\n",
       " PosixPath('../data/imagenet/train/n03998194'),\n",
       " PosixPath('../data/imagenet/train/n02025239'),\n",
       " PosixPath('../data/imagenet/train/n03776460'),\n",
       " PosixPath('../data/imagenet/train/n04591713'),\n",
       " PosixPath('../data/imagenet/train/n03938244'),\n",
       " PosixPath('../data/imagenet/train/n03775546'),\n",
       " PosixPath('../data/imagenet/train/n02105505'),\n",
       " PosixPath('../data/imagenet/train/n03598930'),\n",
       " PosixPath('../data/imagenet/train/n03584829'),\n",
       " PosixPath('../data/imagenet/train/n02011460'),\n",
       " PosixPath('../data/imagenet/train/n02108915'),\n",
       " PosixPath('../data/imagenet/train/n07718747'),\n",
       " PosixPath('../data/imagenet/train/n04357314'),\n",
       " PosixPath('../data/imagenet/train/n03344393'),\n",
       " PosixPath('../data/imagenet/train/n03637318'),\n",
       " PosixPath('../data/imagenet/train/n04356056'),\n",
       " PosixPath('../data/imagenet/train/n07753275'),\n",
       " PosixPath('../data/imagenet/train/n04118776'),\n",
       " PosixPath('../data/imagenet/train/n03220513'),\n",
       " PosixPath('../data/imagenet/train/n02454379'),\n",
       " PosixPath('../data/imagenet/train/n03908714'),\n",
       " PosixPath('../data/imagenet/train/n02071294'),\n",
       " PosixPath('../data/imagenet/train/n03089624'),\n",
       " PosixPath('../data/imagenet/train/n02129604'),\n",
       " PosixPath('../data/imagenet/train/n02488291'),\n",
       " PosixPath('../data/imagenet/train/n02536864'),\n",
       " PosixPath('../data/imagenet/train/n04442312'),\n",
       " PosixPath('../data/imagenet/train/n02877765'),\n",
       " PosixPath('../data/imagenet/train/n01440764'),\n",
       " PosixPath('../data/imagenet/train/n01735189'),\n",
       " PosixPath('../data/imagenet/train/n03223299'),\n",
       " PosixPath('../data/imagenet/train/n13040303'),\n",
       " PosixPath('../data/imagenet/train/n02106550'),\n",
       " PosixPath('../data/imagenet/train/n03384352'),\n",
       " PosixPath('../data/imagenet/train/n02107142'),\n",
       " PosixPath('../data/imagenet/train/n02865351'),\n",
       " PosixPath('../data/imagenet/train/n02123045'),\n",
       " PosixPath('../data/imagenet/train/n02097474'),\n",
       " PosixPath('../data/imagenet/train/n01514859'),\n",
       " PosixPath('../data/imagenet/train/n04090263'),\n",
       " PosixPath('../data/imagenet/train/n02443114'),\n",
       " PosixPath('../data/imagenet/train/n01631663'),\n",
       " PosixPath('../data/imagenet/train/n01944390'),\n",
       " PosixPath('../data/imagenet/train/n09193705'),\n",
       " PosixPath('../data/imagenet/train/n04296562'),\n",
       " PosixPath('../data/imagenet/train/n02342885'),\n",
       " PosixPath('../data/imagenet/train/n07754684'),\n",
       " PosixPath('../data/imagenet/train/n04550184'),\n",
       " PosixPath('../data/imagenet/train/n04355338'),\n",
       " PosixPath('../data/imagenet/train/n03742115'),\n",
       " PosixPath('../data/imagenet/train/n06596364'),\n",
       " PosixPath('../data/imagenet/train/n01774750'),\n",
       " PosixPath('../data/imagenet/train/n03691459'),\n",
       " PosixPath('../data/imagenet/train/n07768694'),\n",
       " PosixPath('../data/imagenet/train/n02817516'),\n",
       " PosixPath('../data/imagenet/train/n03201208'),\n",
       " PosixPath('../data/imagenet/train/n04532670'),\n",
       " PosixPath('../data/imagenet/train/n07714990'),\n",
       " PosixPath('../data/imagenet/train/n04265275'),\n",
       " PosixPath('../data/imagenet/train/n02486410'),\n",
       " PosixPath('../data/imagenet/train/n03179701'),\n",
       " PosixPath('../data/imagenet/train/n02494079'),\n",
       " PosixPath('../data/imagenet/train/n03710721'),\n",
       " PosixPath('../data/imagenet/train/n04152593'),\n",
       " PosixPath('../data/imagenet/train/n11939491'),\n",
       " PosixPath('../data/imagenet/train/n02090379'),\n",
       " PosixPath('../data/imagenet/train/n03394916'),\n",
       " PosixPath('../data/imagenet/train/n02786058'),\n",
       " PosixPath('../data/imagenet/train/n02231487'),\n",
       " PosixPath('../data/imagenet/train/n03874293'),\n",
       " PosixPath('../data/imagenet/train/n02909870'),\n",
       " PosixPath('../data/imagenet/train/n02481823'),\n",
       " PosixPath('../data/imagenet/train/n04041544'),\n",
       " PosixPath('../data/imagenet/train/n03325584'),\n",
       " PosixPath('../data/imagenet/train/n02112350'),\n",
       " PosixPath('../data/imagenet/train/n01608432'),\n",
       " PosixPath('../data/imagenet/train/n03868242'),\n",
       " PosixPath('../data/imagenet/train/n02963159'),\n",
       " PosixPath('../data/imagenet/train/n02037110'),\n",
       " PosixPath('../data/imagenet/train/n01770393'),\n",
       " PosixPath('../data/imagenet/train/n03929660'),\n",
       " PosixPath('../data/imagenet/train/n02791270'),\n",
       " PosixPath('../data/imagenet/train/n04192698'),\n",
       " PosixPath('../data/imagenet/train/n03633091'),\n",
       " PosixPath('../data/imagenet/train/n03710193'),\n",
       " PosixPath('../data/imagenet/train/n07590611'),\n",
       " PosixPath('../data/imagenet/train/n02096051'),\n",
       " PosixPath('../data/imagenet/train/n04251144'),\n",
       " PosixPath('../data/imagenet/train/n04141975'),\n",
       " PosixPath('../data/imagenet/train/n01910747'),\n",
       " PosixPath('../data/imagenet/train/n03770439'),\n",
       " PosixPath('../data/imagenet/train/n02423022'),\n",
       " PosixPath('../data/imagenet/train/n04380533'),\n",
       " PosixPath('../data/imagenet/train/n03424325'),\n",
       " PosixPath('../data/imagenet/train/n02102177'),\n",
       " PosixPath('../data/imagenet/train/n04525038'),\n",
       " PosixPath('../data/imagenet/train/n03970156'),\n",
       " PosixPath('../data/imagenet/train/n03125729'),\n",
       " PosixPath('../data/imagenet/train/n01924916'),\n",
       " PosixPath('../data/imagenet/train/n02910353'),\n",
       " PosixPath('../data/imagenet/train/n01687978'),\n",
       " PosixPath('../data/imagenet/train/n07860988'),\n",
       " PosixPath('../data/imagenet/train/n02930766'),\n",
       " PosixPath('../data/imagenet/train/n02666196'),\n",
       " PosixPath('../data/imagenet/train/n15075141'),\n",
       " PosixPath('../data/imagenet/train/n02104365'),\n",
       " PosixPath('../data/imagenet/train/n02105412'),\n",
       " PosixPath('../data/imagenet/train/n03014705'),\n",
       " PosixPath('../data/imagenet/train/n04458633'),\n",
       " PosixPath('../data/imagenet/train/n02097658'),\n",
       " PosixPath('../data/imagenet/train/n02091134'),\n",
       " PosixPath('../data/imagenet/train/n02108551'),\n",
       " PosixPath('../data/imagenet/train/n02730930'),\n",
       " PosixPath('../data/imagenet/train/n02102318'),\n",
       " PosixPath('../data/imagenet/train/n03494278'),\n",
       " PosixPath('../data/imagenet/train/n01491361'),\n",
       " PosixPath('../data/imagenet/train/n07615774'),\n",
       " PosixPath('../data/imagenet/train/n01558993'),\n",
       " PosixPath('../data/imagenet/train/n03706229'),\n",
       " PosixPath('../data/imagenet/train/n02690373'),\n",
       " PosixPath('../data/imagenet/train/n04125021'),\n",
       " PosixPath('../data/imagenet/train/n02727426'),\n",
       " PosixPath('../data/imagenet/train/n04398044'),\n",
       " PosixPath('../data/imagenet/train/n06874185'),\n",
       " PosixPath('../data/imagenet/train/n04613696'),\n",
       " PosixPath('../data/imagenet/train/n03781244'),\n",
       " PosixPath('../data/imagenet/train/n02802426'),\n",
       " PosixPath('../data/imagenet/train/n04366367'),\n",
       " PosixPath('../data/imagenet/train/n03876231'),\n",
       " PosixPath('../data/imagenet/train/n03967562'),\n",
       " PosixPath('../data/imagenet/train/n02346627'),\n",
       " PosixPath('../data/imagenet/train/n01784675'),\n",
       " PosixPath('../data/imagenet/train/n07720875'),\n",
       " PosixPath('../data/imagenet/train/n02113712'),\n",
       " PosixPath('../data/imagenet/train/n03255030'),\n",
       " PosixPath('../data/imagenet/train/n04557648'),\n",
       " PosixPath('../data/imagenet/train/n02279972'),\n",
       " PosixPath('../data/imagenet/train/n01531178'),\n",
       " PosixPath('../data/imagenet/train/n02640242'),\n",
       " PosixPath('../data/imagenet/train/n01775062'),\n",
       " PosixPath('../data/imagenet/train/n03000684'),\n",
       " PosixPath('../data/imagenet/train/n02672831'),\n",
       " PosixPath('../data/imagenet/train/n04286575'),\n",
       " PosixPath('../data/imagenet/train/n04417672'),\n",
       " PosixPath('../data/imagenet/train/n02113624'),\n",
       " PosixPath('../data/imagenet/train/n01774384'),\n",
       " PosixPath('../data/imagenet/train/n02110806'),\n",
       " PosixPath('../data/imagenet/train/n07747607'),\n",
       " PosixPath('../data/imagenet/train/n02090721'),\n",
       " PosixPath('../data/imagenet/train/n02981792'),\n",
       " PosixPath('../data/imagenet/train/n02114548'),\n",
       " PosixPath('../data/imagenet/train/n03095699'),\n",
       " PosixPath('../data/imagenet/train/n02843684'),\n",
       " PosixPath('../data/imagenet/train/n03400231'),\n",
       " PosixPath('../data/imagenet/train/n01443537'),\n",
       " PosixPath('../data/imagenet/train/n02012849'),\n",
       " PosixPath('../data/imagenet/train/n02410509'),\n",
       " PosixPath('../data/imagenet/train/n01797886'),\n",
       " PosixPath('../data/imagenet/train/n02109047'),\n",
       " PosixPath('../data/imagenet/train/n04266014'),\n",
       " PosixPath('../data/imagenet/train/n02086079'),\n",
       " PosixPath('../data/imagenet/train/n01537544'),\n",
       " PosixPath('../data/imagenet/train/n13044778'),\n",
       " PosixPath('../data/imagenet/train/n03160309'),\n",
       " PosixPath('../data/imagenet/train/n04542943'),\n",
       " PosixPath('../data/imagenet/train/n03047690'),\n",
       " PosixPath('../data/imagenet/train/n02777292'),\n",
       " PosixPath('../data/imagenet/train/n12620546'),\n",
       " PosixPath('../data/imagenet/train/n03661043'),\n",
       " PosixPath('../data/imagenet/train/n03777754'),\n",
       " PosixPath('../data/imagenet/train/n02134418'),\n",
       " PosixPath('../data/imagenet/train/n02110185'),\n",
       " PosixPath('../data/imagenet/train/n01632777'),\n",
       " PosixPath('../data/imagenet/train/n07715103'),\n",
       " PosixPath('../data/imagenet/train/n02088364'),\n",
       " PosixPath('../data/imagenet/train/n03873416'),\n",
       " PosixPath('../data/imagenet/train/n04153751'),\n",
       " PosixPath('../data/imagenet/train/n02268443')]"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from fastai.conv_learner import *\n",
    "PATH = Path(\"../data/imagenet/\")\n",
    "list((PATH/'train').iterdir())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "L = list((PATH/'train').iterdir())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "L1 = L[1].glob('*')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[PosixPath('../data/imagenet/train/n03791053/n03791053_924.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_5620.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_29469.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_22130.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_29063.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_12416.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_6960.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_6634.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_26811.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_7250.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_57293.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_29727.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_21340.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_34832.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_3601.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_10181.JPEG'),\n",
       " PosixPath('../data/imagenet/train/n03791053/n03791053_15494.JPEG')]"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(L1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "filenames, classes = [],[]\n",
    "TRN_PATH = (PATH/'train')\n",
    "for directory in TRN_PATH.iterdir():\n",
    "    for fn in directory.glob('*'):\n",
    "        filenames.append(str(fn)[len(str(TRN_PATH))+1:])\n",
    "        classes.append(str(directory)[len(str(TRN_PATH))+1:])\n",
    "class_names = list(set(classes))\n",
    "class2idx = {c:i for i,c in enumerate(class_names)}\n",
    "labels = [class2idx[c] for c in classes]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.DataFrame({'filenames':filenames, 'cats':labels}, columns=['filenames', 'cats'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>filenames</th>\n",
       "      <th>cats</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>n01644373/n01644373_8692.JPEG</td>\n",
       "      <td>114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>n01644373/n01644373_7872.JPEG</td>\n",
       "      <td>114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>n01644373/n01644373_3042.JPEG</td>\n",
       "      <td>114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>n01644373/n01644373_1037.JPEG</td>\n",
       "      <td>114</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>n01644373/n01644373_7109.JPEG</td>\n",
       "      <td>114</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                       filenames  cats\n",
       "0  n01644373/n01644373_8692.JPEG   114\n",
       "1  n01644373/n01644373_7872.JPEG   114\n",
       "2  n01644373/n01644373_3042.JPEG   114\n",
       "3  n01644373/n01644373_1037.JPEG   114\n",
       "4  n01644373/n01644373_7109.JPEG   114"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv(PATH/'train.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "stats = (np.array([0.485, 0.456, 0.406]), np.array([ 0.229, 0.224, 0.225]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_data(sz,bs):\n",
    "    tfms = tfms_from_model(resnet50,sz)\n",
    "    return ImageClassifierData.from_csv(PATH,'train',PATH/'train.csv',bs=bs, tfms=tfms)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "size = 256\n",
    "batch_size = 32\n",
    "data = get_data(size,batch_size)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Resnet 50"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Even if I'll use the one from pytorch to be **really** sure it's the good one, it doesn't harm to rewrite it."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Bottleneck(nn.Module):\n",
    "    \n",
    "    def __init__(self, ch_in, ch_out, stride=1):\n",
    "        super().__init__()\n",
    "        ch_hid = ch_out//4\n",
    "        self.bn1 = nn.BatchNorm2d(ch_hid)\n",
    "        self.conv1 = nn.Conv2d(ch_in, ch_hid, kernel_size=1, stride=stride, padding=0, bias=False)\n",
    "        self.bn2 = nn.BatchNorm2d(ch_hid)\n",
    "        self.conv2 = nn.Conv2d(ch_hid, ch_hid, kernel_size=3, stride=1, padding=1, bias=False)\n",
    "        self.bn3 = nn.BatchNorm2d(ch_out)\n",
    "        self.conv3 = nn.Conv2d(ch_hid, ch_out, kernel_size=1, stride=1, padding=0, bias=False)\n",
    "\n",
    "        if stride != 1 or ch_in != ch_out:\n",
    "            self.shortcut = nn.Sequential(\n",
    "                nn.Conv2d(ch_in, ch_out, kernel_size=1, stride=stride, bias=False),\n",
    "                nn.BatchNorm2d(ch_out)\n",
    "            )\n",
    "\n",
    "    def forward(self, x):\n",
    "        shortcut = self.shortcut(x) if hasattr(self, 'shortcut') else x\n",
    "        out = F.relu(self.bn1(self.conv1(x)))\n",
    "        out = F.relu(self.bn2(self.conv2(out)))\n",
    "        out = self.bn3(self.conv3(out))\n",
    "        out += shortcut\n",
    "        return F.relu(out)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ResNet(nn.Module):\n",
    "    def __init__(self, num_blocks, num_classes=1000):\n",
    "        super().__init__()\n",
    "        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)\n",
    "        self.bn1 = nn.BatchNorm2d(64)\n",
    "        self.layer1 = self.make_group_layer(64, 256, num_blocks[0], stride=1)\n",
    "        self.layer2 = self.make_group_layer(256, 512, num_blocks[1], stride=2)\n",
    "        self.layer3 = self.make_group_layer(512, 1024, num_blocks[2], stride=2)\n",
    "        self.layer4 = self.make_group_layer(1024, 2048, num_blocks[3], stride=2)\n",
    "        self.linear = nn.Linear(2048, num_classes)\n",
    "\n",
    "    def make_group_layer(self,ch_in, ch_out, num_blocks, stride):\n",
    "        layers = [Bottleneck(ch_in, ch_out, stride)]\n",
    "        for i in range(num_blocks-1):\n",
    "            layers.append(Bottleneck(ch_out, ch_out, stride=1))\n",
    "        return nn.Sequential(*layers)\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = F.relu(self.bn1(self.conv1(x)))\n",
    "        out = F.max_pool2d(out, kernel_size=3,stride=2,padding=1)\n",
    "        out = self.layer1(out)\n",
    "        out = self.layer2(out)\n",
    "        out = self.layer3(out)\n",
    "        out = self.layer4(out)\n",
    "        out = F.adaptive_avg_pool2d(out, 1)\n",
    "        out = out.view(out.size(0), -1)\n",
    "        return F.log_softmax(self.linear(out))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "resnet50a = ResNet([3,4,6,3])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now the one from pytorch."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "from torchvision.models import resnet50"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "class addLast(nn.Module):\n",
    "    def __init__(self,model):\n",
    "        super().__init__()\n",
    "        self.model = model\n",
    "        self.linear = nn.Linear(8192,1000)\n",
    "    \n",
    "    def forward(self,x):\n",
    "        out = self.model(x)\n",
    "        out = out.view(out.size(0),-1)\n",
    "        return F.log_softmax(self.linear(out))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "resnet50py = resnet50()\n",
    "removed = list(resnet50py.children())[:-1]\n",
    "resnet50py = addLast(nn.Sequential(*removed))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn1 = ConvLearner.from_model_data(resnet50py, data)\n",
    "learn1.crit = F.nll_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a2cc610879174380a618a8d0c0d7ac7a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                   \n",
      "    0      6.896336   9.951899   0.000772  \n",
      "\n"
     ]
    }
   ],
   "source": [
    "learn1.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn1.sched.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "abbecd7303774bd28df698fb5bc1220c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=3), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                   \n",
      "    0      6.75804    9.671463   0.002058  \n",
      "    1      6.723974   9.446687   0.001029                   \n",
      "    2      6.713316   9.101116   0.002573                   \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[9.101116424542882, 0.002572678157962439]"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn1.fit(0.5,3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# PreAct Resnet"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Won't be used, but useful to warm up."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [],
   "source": [
    "class PreActBottleneck(nn.Module):\n",
    "    \n",
    "    def __init__(self, ch_in, ch_out, stride=1):\n",
    "        super().__init__()\n",
    "        ch_hid = ch_out//4\n",
    "        self.bn1 = nn.BatchNorm2d(ch_in)\n",
    "        self.conv1 = nn.Conv2d(ch_in, ch_hid, kernel_size=1, stride=stride, padding=0, bias=False)\n",
    "        self.bn2 = nn.BatchNorm2d(ch_hid)\n",
    "        self.conv2 = nn.Conv2d(ch_hid, ch_hid, kernel_size=3, stride=1, padding=1, bias=False)\n",
    "        self.bn3 = nn.BatchNorm2d(ch_hid)\n",
    "        self.conv3 = nn.Conv2d(ch_hid, ch_out, kernel_size=1, stride=1, padding=0, bias=False)\n",
    "\n",
    "        if stride != 1 or ch_in != ch_out:\n",
    "            self.shortcut = nn.Conv2d(ch_in, ch_out, kernel_size=1, stride=stride, bias=False)\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = F.relu(self.bn1(x))\n",
    "        shortcut = self.shortcut(out) if hasattr(self, 'shortcut') else out\n",
    "        out = F.relu(self.bn2(self.conv1(x)))\n",
    "        out = F.relu(self.bn3(self.conv2(out)))\n",
    "        out = self.conv3(out)\n",
    "        out += shortcut\n",
    "        return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "class PreActResNet(nn.Module):\n",
    "    def __init__(self, num_blocks, num_classes=1000):\n",
    "        super().__init__()\n",
    "        self.conv1 = nn.Conv2d(3, 64, kernel_size=7, stride=2, padding=3, bias=False)\n",
    "        self.bn1 = nn.BatchNorm2d(64)\n",
    "        self.layer1 = self.make_group_layer(64, 256, num_blocks[0], stride=1)\n",
    "        self.layer2 = self.make_group_layer(256, 512, num_blocks[1], stride=2)\n",
    "        self.layer3 = self.make_group_layer(512, 1024, num_blocks[2], stride=2)\n",
    "        self.layer4 = self.make_group_layer(1024, 2048, num_blocks[3], stride=2)\n",
    "        self.bn2 = nn.BatchNorm2d(2048)\n",
    "        self.linear = nn.Linear(2048, num_classes)\n",
    "\n",
    "    def make_group_layer(self,ch_in, ch_out, num_blocks, stride):\n",
    "        layers = [PreActBottleneck(ch_in, ch_out, stride)]\n",
    "        for i in range(num_blocks-1):\n",
    "            layers.append(Bottleneck(ch_out, ch_out, stride=1))\n",
    "        return nn.Sequential(*layers)\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = F.relu(self.bn1(self.conv1(x)))\n",
    "        out = F.max_pool2d(out, kernel_size=3,stride=2,padding=1)\n",
    "        out = self.layer1(out)\n",
    "        out = self.layer2(out)\n",
    "        out = self.layer3(out)\n",
    "        out = self.layer4(out)\n",
    "        out = F.relu(self.bn2(out))\n",
    "        out = F.adaptive_avg_pool2d(out, 1)\n",
    "        out = out.view(out.size(0), -1)\n",
    "        return F.log_softmax(self.linear(out))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 64,
   "metadata": {},
   "outputs": [],
   "source": [
    "preactresnet50 = PreActResNet([3,4,6,3])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Darknet"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "class ConvBN(nn.Module):\n",
    "    #convolutional layer then Batchnorm\n",
    "    def __init__(self, ch_in, ch_out, kernel_size = 3, stride=1, padding=0):\n",
    "        super().__init__()\n",
    "        self.conv = nn.Conv2d(ch_in, ch_out, kernel_size=kernel_size, stride=stride, padding=padding, bias=False)\n",
    "        self.bn = nn.BatchNorm2d(ch_out, momentum=0.01)\n",
    "\n",
    "    def forward(self, x):\n",
    "        return F.leaky_relu(self.bn(self.conv(x)), negative_slope=0.1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "class DarknetBlock(nn.Module):\n",
    "    #The basic blocs.   \n",
    "    def __init__(self, ch_in):\n",
    "        super().__init__()\n",
    "        ch_hid = ch_in//2\n",
    "        self.conv1 = ConvBN(ch_in, ch_hid, kernel_size=1, stride=1, padding=0)\n",
    "        self.conv2 = ConvBN(ch_hid, ch_in, kernel_size=3, stride=1, padding=1)\n",
    "        \n",
    "    def forward(self, x):\n",
    "        out = self.conv1(x)\n",
    "        out = self.conv2(out)\n",
    "        return out + x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Darknet(nn.Module):\n",
    "    #Replicates the table 1 from the YOLOv3 paper\n",
    "    def __init__(self, num_blocks, num_classes=1000):\n",
    "        super().__init__()\n",
    "        self.conv = ConvBN(3, 32, kernel_size=3, stride=1, padding=1)\n",
    "        self.layer1 = self.make_group_layer(32, num_blocks[0])\n",
    "        self.layer2 = self.make_group_layer(64, num_blocks[1], stride=2)\n",
    "        self.layer3 = self.make_group_layer(128, num_blocks[2], stride=2)\n",
    "        self.layer4 = self.make_group_layer(256, num_blocks[3], stride=2)\n",
    "        self.layer5 = self.make_group_layer(512, num_blocks[4], stride=2)\n",
    "        self.linear = nn.Linear(1024, num_classes)\n",
    "\n",
    "    def make_group_layer(self, ch_in, num_blocks, stride=1):\n",
    "        layers = [ConvBN(ch_in,ch_in*2,stride=stride)]\n",
    "        for i in range(num_blocks):\n",
    "            layers.append(DarknetBlock(ch_in*2))\n",
    "        return nn.Sequential(*layers)\n",
    "\n",
    "    def forward(self, x):\n",
    "        out = self.conv(x)\n",
    "        out = self.layer1(out)\n",
    "        out = self.layer2(out)\n",
    "        out = self.layer3(out)\n",
    "        out = self.layer4(out)\n",
    "        out = self.layer5(out)\n",
    "        out = F.adaptive_avg_pool2d(out, 1)\n",
    "        out = out.view(out.size(0), -1)\n",
    "        return F.log_softmax(self.linear(out))   "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "darknet53 = Darknet([1,2,8,8,4])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = ConvLearner.from_model_data(darknet53, data)\n",
    "learn.crit = F.nll_loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "OrderedDict([('Conv2d-1',\n",
       "              OrderedDict([('input_shape', [-1, 3, 256, 256]),\n",
       "                           ('output_shape', [-1, 32, 256, 256]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 864)])),\n",
       "             ('BatchNorm2d-2',\n",
       "              OrderedDict([('input_shape', [-1, 32, 256, 256]),\n",
       "                           ('output_shape', [-1, 32, 256, 256]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 64)])),\n",
       "             ('ConvBN-3',\n",
       "              OrderedDict([('input_shape', [-1, 3, 256, 256]),\n",
       "                           ('output_shape', [-1, 32, 256, 256]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-4',\n",
       "              OrderedDict([('input_shape', [-1, 32, 256, 256]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 18432)])),\n",
       "             ('BatchNorm2d-5',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 128)])),\n",
       "             ('ConvBN-6',\n",
       "              OrderedDict([('input_shape', [-1, 32, 256, 256]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-7',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 32, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('BatchNorm2d-8',\n",
       "              OrderedDict([('input_shape', [-1, 32, 254, 254]),\n",
       "                           ('output_shape', [-1, 32, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 64)])),\n",
       "             ('ConvBN-9',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 32, 254, 254]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-10',\n",
       "              OrderedDict([('input_shape', [-1, 32, 254, 254]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 18432)])),\n",
       "             ('BatchNorm2d-11',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 128)])),\n",
       "             ('ConvBN-12',\n",
       "              OrderedDict([('input_shape', [-1, 32, 254, 254]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-13',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 64, 254, 254]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-14',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 73728)])),\n",
       "             ('BatchNorm2d-15',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-16',\n",
       "              OrderedDict([('input_shape', [-1, 64, 254, 254]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-17',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 8192)])),\n",
       "             ('BatchNorm2d-18',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 128)])),\n",
       "             ('ConvBN-19',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-20',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 73728)])),\n",
       "             ('BatchNorm2d-21',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-22',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-23',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-24',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 8192)])),\n",
       "             ('BatchNorm2d-25',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 128)])),\n",
       "             ('ConvBN-26',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 64, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-27',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 73728)])),\n",
       "             ('BatchNorm2d-28',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-29',\n",
       "              OrderedDict([('input_shape', [-1, 64, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-30',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 128, 126, 126]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-31',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-32',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-33',\n",
       "              OrderedDict([('input_shape', [-1, 128, 126, 126]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-34',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-35',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-36',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-37',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-38',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-39',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-40',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-41',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-42',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-43',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-44',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-45',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-46',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-47',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-48',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-49',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-50',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-51',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-52',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-53',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-54',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-55',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-56',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-57',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-58',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-59',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-60',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-61',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-62',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-63',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-64',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-65',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-66',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-67',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-68',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-69',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-70',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-71',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-72',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-73',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-74',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-75',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-76',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-77',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-78',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-79',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-80',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-81',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-82',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-83',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 32768)])),\n",
       "             ('BatchNorm2d-84',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 256)])),\n",
       "             ('ConvBN-85',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 128, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-86',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 294912)])),\n",
       "             ('BatchNorm2d-87',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-88',\n",
       "              OrderedDict([('input_shape', [-1, 128, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-89',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 256, 62, 62]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-90',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-91',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-92',\n",
       "              OrderedDict([('input_shape', [-1, 256, 62, 62]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-93',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-94',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-95',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-96',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-97',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-98',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-99',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-100',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-101',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-102',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-103',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-104',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-105',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-106',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-107',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-108',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-109',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-110',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-111',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-112',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-113',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-114',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-115',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-116',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-117',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-118',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-119',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-120',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-121',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-122',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-123',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-124',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-125',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-126',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-127',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-128',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-129',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-130',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-131',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-132',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-133',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-134',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-135',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-136',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-137',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-138',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-139',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-140',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-141',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-142',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 131072)])),\n",
       "             ('BatchNorm2d-143',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 512)])),\n",
       "             ('ConvBN-144',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 256, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-145',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1179648)])),\n",
       "             ('BatchNorm2d-146',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-147',\n",
       "              OrderedDict([('input_shape', [-1, 256, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-148',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 512, 30, 30]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-149',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 4718592)])),\n",
       "             ('BatchNorm2d-150',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('ConvBN-151',\n",
       "              OrderedDict([('input_shape', [-1, 512, 30, 30]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-152',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 524288)])),\n",
       "             ('BatchNorm2d-153',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-154',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-155',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 4718592)])),\n",
       "             ('BatchNorm2d-156',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('ConvBN-157',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-158',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-159',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 524288)])),\n",
       "             ('BatchNorm2d-160',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-161',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-162',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 4718592)])),\n",
       "             ('BatchNorm2d-163',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('ConvBN-164',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-165',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-166',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 524288)])),\n",
       "             ('BatchNorm2d-167',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-168',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-169',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 4718592)])),\n",
       "             ('BatchNorm2d-170',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('ConvBN-171',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-172',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-173',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 524288)])),\n",
       "             ('BatchNorm2d-174',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1024)])),\n",
       "             ('ConvBN-175',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 512, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Conv2d-176',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 4718592)])),\n",
       "             ('BatchNorm2d-177',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 2048)])),\n",
       "             ('ConvBN-178',\n",
       "              OrderedDict([('input_shape', [-1, 512, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('DarknetBlock-179',\n",
       "              OrderedDict([('input_shape', [-1, 1024, 14, 14]),\n",
       "                           ('output_shape', [-1, 1024, 14, 14]),\n",
       "                           ('nb_params', 0)])),\n",
       "             ('Linear-180',\n",
       "              OrderedDict([('input_shape', [-1, 1024]),\n",
       "                           ('output_shape', [-1, 1000]),\n",
       "                           ('trainable', True),\n",
       "                           ('nb_params', 1025000)]))])"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "180d511469c74fb6b50603d2c355f1f2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=1), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 90%|█████████ | 439/486 [07:14<00:46,  1.01it/s, loss=71.3]"
     ]
    }
   ],
   "source": [
    "learn.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.sched.plot(10,5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dab703e6d478412487bb404c0ec5e55a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(IntProgress(value=0, description='Epoch', max=3), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 9/486 [00:10<08:54,  1.12s/it, loss=8.01]  \n",
      "  2%|▏         | 10/486 [00:10<07:59,  1.01s/it, loss=8.01]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Exception in thread Thread-4:\n",
      "Traceback (most recent call last):\n",
      "  File \"/home/paperspace/anaconda3/envs/fastai/lib/python3.6/threading.py\", line 916, in _bootstrap_inner\n",
      "    self.run()\n",
      "  File \"/home/paperspace/anaconda3/envs/fastai/lib/python3.6/site-packages/tqdm/_monitor.py\", line 62, in run\n",
      "    for instance in self.tqdm_cls._instances:\n",
      "  File \"/home/paperspace/anaconda3/envs/fastai/lib/python3.6/_weakrefset.py\", line 60, in __iter__\n",
      "    for itemref in self.data:\n",
      "RuntimeError: Set changed size during iteration\n",
      "\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch      trn_loss   val_loss   accuracy                   \n",
      "    0      6.707158   6.916332   0.001801  \n",
      "    1      6.564623   6.658093   0.00283                    \n",
      "    2      6.469157   6.532844   0.005145                   \n",
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "[6.532844449120068, 0.005145356315924878]"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.fit(0.1,3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}