imagenet_ssl_epoch.py

import argparse
import os
import random
import shutil
import time
import warnings
import numpy as np
from tqdm import tqdm

import torch
import torch.nn as nn
import torch.nn.parallel
import torch.backends.cudnn as cudnn
import torch.distributed as dist
import torch.optim
import torch.multiprocessing as mp
import torch.utils.data
import torch.utils.data.distributed
import torchvision.transforms as transforms
import torchvision.datasets as datasets
import torchvision.models as models

import wandb
from dataset.folder import ImageFolder
from warmup_scheduler import GradualWarmupScheduler
from taskonomy_encoder import TaskonomyEncoder

model_names = sorted(name for name in models.__dict__
    if name.islower() and not name.startswith("__")
    and callable(models.__dict__[name]))

parser = argparse.ArgumentParser(description='PyTorch ImageNet Training')
parser.add_argument('data', metavar='DIR',
                    help='path to dataset')
parser.add_argument('-a', '--arch', metavar='ARCH', default='resnet50',
                    choices=model_names,
                    help='model architecture: ' +
                        ' | '.join(model_names) +
                        ' (default: resnet50)')
parser.add_argument('-j', '--workers', default=32, type=int, metavar='N',
                    help='number of data loading workers (default: 4)')
parser.add_argument('--epochs', default=100, type=int, metavar='N',
                    help='number of total epochs to run')
parser.add_argument('--start-epoch', default=0, type=int, metavar='N',
                    help='manual epoch number (useful on restarts)')
parser.add_argument('-b', '--batch-size', default=256, type=int,
                    metavar='N',
                    help='mini-batch size (default: 256), this is the total '
                         'batch size of all GPUs on the current node when '
                         'using Data Parallel or Distributed Data Parallel')
parser.add_argument('--lr', '--learning-rate', default=0.1, type=float,
                    metavar='LR', help='initial learning rate', dest='lr')
parser.add_argument('--momentum', default=0.9, type=float, metavar='M',
                    help='momentum')
parser.add_argument('--wd', '--weight-decay', default=1e-4, type=float,
                    metavar='W', help='weight decay (default: 1e-4)',
                    dest='weight_decay')
parser.add_argument('-p', '--print-freq', default=10, type=int,
                    metavar='N', help='print frequency (default: 10)')
parser.add_argument('--resume', default='', type=str, metavar='PATH',
                    help='path to latest checkpoint (default: none)')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
                    help='evaluate model on validation set')
parser.add_argument('--pretrained', dest='pretrained', action='store_true',
                    help='use pre-trained model')
parser.add_argument('--world-size', default=-1, type=int,
                    help='number of nodes for distributed training')
parser.add_argument('--rank', default=-1, type=int,
                    help='node rank for distributed training')
parser.add_argument('--dist-url', default='tcp://224.66.41.62:23456', type=str,
                    help='url used to set up distributed training')
parser.add_argument('--dist-backend', default='nccl', type=str,
                    help='distributed backend')
parser.add_argument('--seed', default=None, type=int,
                    help='seed for initializing training. ')
parser.add_argument('--cache', default=False, action='store_true')
parser.add_argument('--gpu', default=None, type=int,
                    help='GPU id to use.')
parser.add_argument('--multiprocessing-distributed', action='store_true',
                    help='Use multi-processing distributed training to launch '
                         'N processes per node, which has N GPUs. This is the '
                         'fastest way to use PyTorch for either single node or '
                         'multi node data parallel training')

parser.add_argument('--model_name', default='scratch', type=str,
                    help='SSL or baseline model name')

parser.add_argument('--freeze', dest='freeze', action='store_true',
                    help='freeze the pretrained model')
parser.add_argument('--rgb2lab', default=False, action='store_true',
                    help='Convert RGB to Lab. (default: False)')

parser.add_argument('--max_images', type=int, default=None,
                    help='Number of training and validation images (0.8 for training, 0.2 for validation). Uses all if None. (default: None)')
parser.add_argument('--warmup', default=None, type=int, 
                    help='The number of epochs to warm up the lr')
parser.add_argument('--lr_step', default=30, type=int,
                    help='learning rate decay step.')

class KEpochsSampler(object):
    def __init__(self, base_iterator, k=-1):
        self.base_iterator = base_iterator
        self.k = k
    
    def __iter__(self):
        it = 0
        while it < self.k or self.k == -1:
            yield from self.base_iterator
            it += 1

    @staticmethod
    def __len__(self):
        return self.k


model_paths = {
    'moco': '/PATH_TO/moco_v2_800ep_pretrain.pth.tar',
    'swav': '/PATH_TO/swav_800ep_pretrain.pth.tar',
    'simclr': '/PATH_TO/simclr_800ep.torch',
    'barlow': '/PATH_TO/barlow_twins_1000ep.pth',
    'moco_task': '/PATH_TO/MoCov2Taskonomy.pt',
    'swav_task': '/PATH_TO/SWAVTaskonomy.pt',
    'color': '/PATH_TO/vissl_colorization_rn50.torch',
    'color_task': '/PATH_TO/task_colorization_encoder.dat',
    'jigsaw': '/PATH_TO/vissl_jigsaw_in1k_rn50.torch',
    'jigsaw_task': '/PATH_TO/task_jigsaw_encoder.dat',
    'simsiam': '/PATH_TO/simsiam_100ep_256bs.pth.tar',
    'pirl': '/PATH_TO/pirl_800ep_liner.torch'
}


best_acc1 = 0

def main():
    args = parser.parse_args()

    if args.seed is not None:
        random.seed(args.seed)
        torch.manual_seed(args.seed)
        torch.cuda.manual_seed(args.seed)
        np.random.seed(args.seed)

        print("using random seed: ", args.seed)
        # cudnn.deterministic = True
        # warnings.warn('You have chosen to seed training. '
        #               'This will turn on the CUDNN deterministic setting, '
        #               'which can slow down your training considerably! '
        #               'You may see unexpected behavior when restarting '
        #               'from checkpoints.')

    if args.gpu is not None:
        warnings.warn('You have chosen a specific GPU. This will completely '
                      'disable data parallelism.')

    if args.dist_url == "env://" and args.world_size == -1:
        args.world_size = int(os.environ["WORLD_SIZE"])

    args.distributed = args.world_size > 1 or args.multiprocessing_distributed

    ngpus_per_node = torch.cuda.device_count()
    if args.multiprocessing_distributed:
        # Since we have ngpus_per_node processes per node, the total world_size
        # needs to be adjusted accordingly
        args.world_size = ngpus_per_node * args.world_size
        # Use torch.multiprocessing.spawn to launch distributed processes: the
        # main_worker process function
        mp.spawn(main_worker, nprocs=ngpus_per_node, args=(ngpus_per_node, args))
    else:
        # Simply call main_worker function
        main_worker(args.gpu, ngpus_per_node, args)


def main_worker(gpu, ngpus_per_node, args):
    global best_acc1
    args.gpu = gpu

    if args.gpu is not None:
        print("Use GPU: {} for training".format(args.gpu))

    if args.distributed:
        if args.dist_url == "env://" and args.rank == -1:
            args.rank = int(os.environ["RANK"])
        if args.multiprocessing_distributed:
            # For multiprocessing distributed training, rank needs to be the
            # global rank among all the processes
            args.rank = args.rank * ngpus_per_node + gpu
        dist.init_process_group(backend=args.dist_backend, init_method=args.dist_url,
                                world_size=args.world_size, rank=args.rank)

    # create model, do not freeze the feature extractor
    if args.model_name == 'jigsaw_task':
        print("=> creating model TaskonomyEncoder")
        model = TaskonomyEncoder()
    elif args.model_name == 'color_task':
        print("=> creating model TaskonomyEncoder")
        model = TaskonomyEncoder()
        model.conv1 = nn.Conv2d(1, 64, kernel_size=7, stride=2, padding=0, bias=False)
    elif args.model_name == 'color':
        print("=> creating model '{}'".format(args.arch))
        model = models.__dict__[args.arch]()
        model.conv1 = nn.Conv2d(1, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)
    else:
        print("=> creating model '{}'".format(args.arch))
        model = models.__dict__[args.arch]()


    # freeze the pretrained backbone but the last fc layer
    if args.freeze:
        print("=> Freezing all the layers but the fc layer")
        for name, param in model.named_parameters():
            if name not in ['fc.weight', 'fc.bias']:
                param.requires_grad = False

        # init the fc layer
        model.fc.weight.data.normal_(mean=0.0, std=0.01)
        model.fc.bias.data.zero_()


    # load from pre-trained, before DistributedDataParallel constructor
    if args.pretrained:
        print("Using model: ", args.model_name)
        model_path = model_paths[args.model_name]
        print("Model path: ", model_path)

        if args.model_name == 'moco':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                checkpoint = torch.load(model_path, map_location="cpu")

                # rename moco pre-trained keys
                state_dict = checkpoint['state_dict']
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('module.encoder_q') and not k.startswith('module.encoder_q.fc'):
                        # remove prefix
                        state_dict[k[len("module.encoder_q."):]] = state_dict[k]
                    # delete renamed or unused k
                    del state_dict[k]

                args.start_epoch = 0
                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))
        
        elif args.model_name == 'swav':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                state_dict = torch.load(model_path, map_location="cpu")

                if "state_dict" in state_dict:
                    state_dict = state_dict["state_dict"]

                # remove prefixe "module."
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('module.') and not k.startswith('module.fc'):
                        # remove prefix
                        state_dict[k[len("module."):]] = state_dict[k]
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))
        
        elif args.model_name == 'simclr':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                state_dict = torch.load(model_path, map_location="cpu")

                state_dict = state_dict['classy_state_dict']['base_model']['model']['trunk']
            
                # remove prefixe "_feature_blocks."
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('_feature_blocks.'):
                        # remove prefix
                        state_dict[k[len("_feature_blocks."):]] = state_dict[k]
    
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'barlow':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                state_dict = torch.load(model_path, map_location="cpu")

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'simsiam':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                
                checkpoint = torch.load(model_path, map_location="cpu")
                state_dict = checkpoint['state_dict']

                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('module.encoder.'):
                        # remove prefix
                        state_dict[k[len("module.encoder."):]] = state_dict[k]
    
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'pirl':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))

                checkpoint = torch.load(model_path, map_location="cpu")
                state_dict = checkpoint['classy_state_dict']['base_model']['model']['trunk']

                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if 'num_tracked_batches' not in k:
                        # remove prefix
                        state_dict[k[len("_features_block."):]] = state_dict[k]
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'swav_task':
            if os.path.isfile(model_path):
                checkpoint = torch.load(model_path, map_location="cpu")

                # remove prefixe "module."
                state_dict = checkpoint
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('model.'):
                        # remove prefix
                        state_dict[k[len("model."):]] = state_dict[k]

                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))
        
        elif args.model_name == 'color':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                state_dict = torch.load(model_path, map_location=torch.device('cpu'))    
                state_dict = state_dict['model_state_dict']

                # remove prefixe "_feature_blocks."
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('_feature_blocks.'):
                        # remove prefix
                        state_dict[k[len("_feature_blocks."):]] = state_dict[k]
    
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'jigsaw':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))
                state_dict = torch.load(model_path, map_location=torch.device('cpu'))    
                state_dict = state_dict['model_state_dict']

                # remove prefixe "_feature_blocks."
                for k in list(state_dict.keys()):
                    # retain only encoder_q up to before the embedding layer
                    if k.startswith('_feature_blocks.'):
                        # remove prefix
                        state_dict[k[len("_feature_blocks."):]] = state_dict[k]
    
                    # delete renamed or unused k
                    del state_dict[k]

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))

        elif args.model_name == 'jigsaw_task':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))

                checkpoint = torch.load(model_path, map_location=torch.device('cpu'))
                state_dict = checkpoint['state_dict']

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))
        
        elif args.model_name == 'color_task':
            if os.path.isfile(model_path):
                print("=> loading checkpoint '{}'".format(model_path))

                checkpoint = torch.load(model_path, map_location=torch.device('cpu'))
                state_dict = checkpoint['state_dict']

                msg = model.load_state_dict(state_dict, strict=False)
                assert set(msg.missing_keys) == {"fc.weight", "fc.bias"}

                print("=> loaded pre-trained model '{}'".format(model_path))
            else:
                print("=> no checkpoint found at '{}'".format(model_path))


    if not torch.cuda.is_available():
        print('using CPU, this will be slow')

    elif args.distributed:
        # For multiprocessing distributed, DistributedDataParallel constructor
        # should always set the single device scope, otherwise,
        # DistributedDataParallel will use all available devices.
        if args.gpu is not None:
            torch.cuda.set_device(args.gpu)
            model.cuda(args.gpu)
            # When using a single GPU per process and per
            # DistributedDataParallel, we need to divide the batch size
            # ourselves based on the total number of GPUs we have
            args.batch_size = int(args.batch_size / ngpus_per_node)
            args.workers = int((args.workers + ngpus_per_node - 1) / ngpus_per_node)
            model = torch.nn.parallel.DistributedDataParallel(model, device_ids=[args.gpu])
        else:
            model.cuda()
            # DistributedDataParallel will divide and allocate batch_size to all
            # available GPUs if device_ids are not set
            model = torch.nn.parallel.DistributedDataParallel(model)
    elif args.gpu is not None:
        torch.cuda.set_device(args.gpu)
        model = model.cuda(args.gpu)
    else:
        # DataParallel will divide and allocate batch_size to all available GPUs
        if args.arch.startswith('alexnet') or args.arch.startswith('vgg'):
            model.features = torch.nn.DataParallel(model.features)
            model.cuda()
        else:
            model = torch.nn.DataParallel(model).cuda()

    # define loss function (criterion) and optimizer
    criterion = nn.CrossEntropyLoss().cuda(args.gpu)

    if args.freeze:
        parameters = list(filter(lambda p: p.requires_grad, model.parameters()))
        assert len(parameters) == 2  # fc.weight, fc.bias
        optimizer = torch.optim.SGD(parameters, args.lr,
                                momentum=args.momentum,
                                weight_decay=args.weight_decay)
    else:
        optimizer = torch.optim.SGD(model.parameters(), args.lr,
                                momentum=args.momentum,
                                weight_decay=args.weight_decay)
    
    scheduler_lr = torch.optim.lr_scheduler.StepLR(optimizer, step_size=args.lr_step, gamma=0.1)
    
    # set lr warmup scheduler 
    scheduler_warmup = None
    if args.warmup:
        scheduler_warmup = GradualWarmupScheduler(optimizer, multiplier=1, total_epoch=args.warmup, after_scheduler=scheduler_lr)

    # optionally resume from a checkpoint
    if args.resume:
        if os.path.isfile(args.resume):
            print("=> loading checkpoint '{}'".format(args.resume))
            if args.gpu is None:
                checkpoint = torch.load(args.resume)
            else:
                # Map model to be loaded to specified single gpu.
                loc = 'cuda:{}'.format(args.gpu)
                checkpoint = torch.load(args.resume, map_location=loc)
            args.start_epoch = checkpoint['epoch']
            best_acc1 = checkpoint['best_acc1']
            if args.gpu is not None:
                # best_acc1 may be from a checkpoint from a different GPU
                best_acc1 = best_acc1.to(args.gpu)
            model.load_state_dict(checkpoint['state_dict'])
            optimizer.load_state_dict(checkpoint['optimizer'])
            print("=> loaded checkpoint '{}' (epoch {})"
                  .format(args.resume, checkpoint['epoch']))
        else:
            print("=> no checkpoint found at '{}'".format(args.resume))

    cudnn.benchmark = True

    # Data loading code
    traindir = os.path.join(args.data, 'train')
    valdir = os.path.join(args.data, 'val')
    normalize = transforms.Normalize(mean=[0.485, 0.456, 0.406],
                                     std=[0.229, 0.224, 0.225])

    if args.rgb2lab:
        train_transform = transforms.Compose([
            transforms.RandomResizedCrop(224),
            transforms.RandomHorizontalFlip(),
        ])
    else:
        train_transform = transforms.Compose([
            transforms.RandomResizedCrop(224),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor(),
            normalize,
        ])

    train_dataset = ImageFolder(
        traindir,
        int(args.max_images*0.8),                   # 0.8 percent images for training
        'train',
        transform=train_transform,
        seed=args.seed,
        cache=args.cache,
        rgb2lab=args.rgb2lab,
    )

    if args.distributed:
        train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
    else:
        train_sampler = None

    train_loader = torch.utils.data.DataLoader(
        train_dataset, batch_size=args.batch_size, shuffle=(train_sampler is None),
        num_workers=args.workers, pin_memory=True, sampler=train_sampler)

    if args.rgb2lab:
        val_transform = transforms.Compose([
            transforms.Resize(256),
            transforms.CenterCrop(224),
        ])
    else:
        val_transform = transforms.Compose([
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            normalize,
        ])

    val_loader = torch.utils.data.DataLoader(
        ImageFolder(
            traindir,                           # split validation set from training data
            int(args.max_images*0.2),           # 0.2 percent images for validation
            'val',
            transform=val_transform,
            seed=args.seed,
            cache=args.cache,
            rgb2lab=args.rgb2lab,
            ),
        batch_size=args.batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True
    )

    print("=> Data loaded.")


    if args.evaluate:
        validate(val_loader, model, criterion, args)
        return


    experiment_name = args.model_name + '_' + str(args.max_images//1000) + 'k_' + str(args.seed)
    if args.freeze:
        experiment_name = experiment_name + '_freeze'

    wandb.init(
        name=experiment_name, 
        project='PROJECT_NAME',
        entity='YOUR_ENTITY'
    )
    wandb.watch(model)

    # Training and validation part

    # this zero gradient update is needed to avoid a warning message, issue #8.
    optimizer.zero_grad()
    optimizer.step()

    for epoch in range(args.start_epoch, args.epochs):
        if args.distributed:
            train_sampler.set_epoch(epoch)

        # train for one epoch
        if scheduler_warmup is not None:
            scheduler_warmup.step(epoch)
        
        wandb.log({
            "epoch": epoch,
            "lr": optimizer.param_groups[0]['lr']
        })

        loss, train_acc1, train_acc5 = train(train_loader, model, criterion, optimizer, epoch, args)

        # evaluate on validation set
        acc1, acc5, val_loss = validate(val_loader, model, criterion, args)
        wandb.log({
            "epoch": epoch,
            "Train Loss": loss,
            "Train Acc1": train_acc1,
            "Train Acc5": train_acc5,
            "Valid Acc1": acc1,
            "Valid Acc5": acc5,
            "Valid Loss": val_loss
        })

        # remember best acc@1 and save checkpoint
        is_best = acc1 > best_acc1
        best_acc1 = max(acc1, best_acc1)

        if not args.multiprocessing_distributed or (args.multiprocessing_distributed
                and args.rank % ngpus_per_node == 0):
            save_checkpoint({
                'epoch': epoch + 1,
                'arch': args.arch,
                'state_dict': model.state_dict(),
                'best_acc1': best_acc1,
                'optimizer' : optimizer.state_dict(),
            }, is_best, args.max_images//1000, args.model_name, args.seed, args.freeze)


    # Load the best model, test on validation set
    print("=> Ready for testing on validation split")
    if args.rgb2lab:
        test_transform = transforms.Compose([
            transforms.Resize(256),
            transforms.CenterCrop(224),
        ])
    else:
        test_transform = transforms.Compose([
            transforms.Resize(256),
            transforms.CenterCrop(224),
            transforms.ToTensor(),
            normalize,
        ])

    if args.rgb2lab:
        test_loader = torch.utils.data.DataLoader(
            ImageFolder(valdir, 
            split='test',
            transform=test_transform,
            rgb2lab=args.rgb2lab,
            ),
            batch_size=args.batch_size, shuffle=False,
            num_workers=args.workers, pin_memory=True)
    else:
        test_loader = torch.utils.data.DataLoader(
            datasets.ImageFolder(valdir, 
            transform=test_transform,
            ),
            batch_size=args.batch_size, shuffle=False,
            num_workers=args.workers, pin_memory=True)

    best_name = 'model_best_' + args.model_name + '_' + str(args.max_images//1000) + 'k_'+ str(args.seed) 
    if args.freeze:
        best_name = best_name + '_freeze'
    best_name = best_name + '.pth.tar'
    
    if not os.path.isfile(best_name):
        # Model training failed, no best model stored
        best_name = 'checkpoint_' + args.model_name + '_' + str(args.max_images//1000) + 'k_' + str(args.seed)
        if args.freeze:
            best_name = best_name + '_freeze'
        best_name = best_name + '.pth.tar'


    print("=> loading the best model")
    print("=>", best_name)
    if args.gpu is None:
        checkpoint = torch.load(best_name)
    else:
        # Map model to be loaded to specified single gpu.
        loc = 'cuda:{}'.format(args.gpu)
        checkpoint = torch.load(best_name, map_location=loc)
    
    model.load_state_dict(checkpoint['state_dict'])

    print("=> start testing ...")
    top1, top5, test_loss = validate(test_loader, model, criterion, args)
    wandb.log({
        "Test Acc@1": top1,
        "Test Acc@5": top5,
        "Test Loss": test_loss
    })


def train(train_loader, model, criterion, optimizer, epoch, args):
    batch_time = AverageMeter('Time', ':6.3f')
    data_time = AverageMeter('Data', ':6.3f')
    losses = AverageMeter('Loss', ':.4e')
    top1 = AverageMeter('Acc@1', ':6.2f')
    top5 = AverageMeter('Acc@5', ':6.2f')
    progress = ProgressMeter(
        len(train_loader),
        [batch_time, data_time, losses, top1, top5],
        prefix="Epoch: [{}]".format(epoch))

    # switch to train mode
    if args.freeze:
        """
        Switch to eval mode:
        Under the protocol of linear classification on frozen features/models,
        it is not legitimate to change any part of the pre-trained model.
        BatchNorm in train mode may revise running mean/std (even if it receives
        no gradient), which are part of the model parameters too.
        """
        model.eval()
    else:
        model.train()

    end = time.time()
    for i, (images, target) in tqdm(enumerate(train_loader)):
        # measure data loading time

        if args.gpu is not None:
            images = images.cuda(args.gpu, non_blocking=True)
        if torch.cuda.is_available():
            target = target.cuda(args.gpu, non_blocking=True)

        data_time.update(time.time() - end)

        # compute output
        output = model(images)
        loss = criterion(output, target)

        # measure accuracy and record loss
        acc1, acc5 = accuracy(output, target, topk=(1, 5))
        losses.update(loss.item(), images.size(0))
        top1.update(acc1[0], images.size(0))
        top5.update(acc5[0], images.size(0))

        # compute gradient and do SGD step
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # measure elapsed time
        batch_time.update(time.time() - end)
        end = time.time()

        if i % args.print_freq == 0:
            progress.display(i)

    return losses.avg, top1.avg, top5.avg



def validate(val_loader, model, criterion, args):
    batch_time = AverageMeter('Time', ':6.3f')
    losses = AverageMeter('Loss', ':.4e')
    top1 = AverageMeter('Acc@1', ':6.2f')
    top5 = AverageMeter('Acc@5', ':6.2f')
    progress = ProgressMeter(
        len(val_loader),
        [batch_time, losses, top1, top5],
        prefix='Test: ')

    # switch to evaluate mode
    model.eval()

    with torch.no_grad():
        end = time.time()
        for i, (images, target) in enumerate(val_loader):
            if args.gpu is not None:
                images = images.cuda(args.gpu, non_blocking=True)
            if torch.cuda.is_available():
                target = target.cuda(args.gpu, non_blocking=True)

            # compute output
            output = model(images)
            loss = criterion(output, target)

            # measure accuracy and record loss
            acc1, acc5 = accuracy(output, target, topk=(1, 5))
            losses.update(loss.item(), images.size(0))
            top1.update(acc1[0], images.size(0))
            top5.update(acc5[0], images.size(0))

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()

            if i % args.print_freq == 0:
                progress.display(i)

        # TODO: this should also be done with the ProgressMeter
        print(' * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f} Loss {losses.avg:.4f}'
              .format(top1=top1, top5=top5, losses=losses))

    return top1.avg, top5.avg, losses.avg


def save_checkpoint(state, is_best, img_num, model_name, seed, freeze, filename='checkpoint.pth.tar'):
    filename = 'checkpoint_' + model_name + '_' + str(img_num) + 'k_' + str(seed)
    if freeze:
        filename = filename + '_freeze'
    filename = filename + '.pth.tar'

    torch.save(state, filename)
    if is_best:
        best_name = 'model_best_' + model_name + '_' + str(img_num) + 'k_' + str(seed)
        if freeze:
            best_name = best_name + '_freeze'
        best_name = best_name + '.pth.tar'
        shutil.copyfile(filename, best_name)


class AverageMeter(object):
    """Computes and stores the average and current value"""
    def __init__(self, name, fmt=':f'):
        self.name = name
        self.fmt = fmt
        self.reset()

    def reset(self):
        self.val = 0
        self.avg = 0
        self.sum = 0
        self.count = 0

    def update(self, val, n=1):
        self.val = val
        self.sum += val * n
        self.count += n
        self.avg = self.sum / self.count

    def __str__(self):
        fmtstr = '{name} {val' + self.fmt + '} ({avg' + self.fmt + '})'
        return fmtstr.format(**self.__dict__)


class ProgressMeter(object):
    def __init__(self, num_batches, meters, prefix=""):
        self.batch_fmtstr = self._get_batch_fmtstr(num_batches)
        self.meters = meters
        self.prefix = prefix

    def display(self, batch):
        entries = [self.prefix + self.batch_fmtstr.format(batch)]
        entries += [str(meter) for meter in self.meters]
        print('\t'.join(entries))

    def _get_batch_fmtstr(self, num_batches):
        num_digits = len(str(num_batches // 1))
        fmt = '{:' + str(num_digits) + 'd}'
        return '[' + fmt + '/' + fmt.format(num_batches) + ']'


def adjust_learning_rate(optimizer, epoch, args):
    """Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
    lr = args.lr * (0.1 ** (epoch // 40))
    
    wandb.log({
        "epoch": epoch,
        "lr": lr
    })

    for param_group in optimizer.param_groups:
        param_group['lr'] = lr


def accuracy(output, target, topk=(1,)):
    """Computes the accuracy over the k top predictions for the specified values of k"""
    with torch.no_grad():
        maxk = max(topk)
        batch_size = target.size(0)

        _, pred = output.topk(maxk, 1, True, True)
        pred = pred.t()
        correct = pred.eq(target.view(1, -1).expand_as(pred))

        res = []
        for k in topk:
            correct_k = correct[:k].reshape(-1).float().sum(0, keepdim=True)
            res.append(correct_k.mul_(100.0 / batch_size))
        return res


if __name__ == '__main__':
    main()