cifar10_train_crossval.py

# coding: utf-8

# This code extends the functionality of
# https://github.com/pytorch/examples/tree/master/imagenet
# to support cross-validation training, allowing you compute the out of sample
# predicted probabilities for the entire CIFAR training set:
# a necessary step for confident learning and the cleanlab package.
# 
# Example showing how to obtain 4-fold cross-validated predicted probabilities:
# $ python3 cifar10_train_crossval.py \
#     -a resnet50 -b 256 --lr 0.1 --gpu 0 --cvn 4 --cv 0  \
#     --train-labels LABELS_PATH.json CIFAR10_PATH
# $ python3 cifar10_train_crossval.py \
#     -a resnet50 -b 256 --lr 0.1 --gpu 1 --cvn 4 --cv 1  \
# #   --train-labels LABELS_PATH.json CIFAR10_PATH
# $ python3 cifar10_train_crossval.py \
#     -a resnet50 -b 256 --lr 0.1 --gpu 2 --cvn 4 --cv 2  \
# #   --train-labels LABELS_PATH.json CIFAR10_PATH
# $ python3 cifar10_train_crossval.py \
#     -a resnet50 -b 256 --lr 0.1 --gpu 3 --cvn 4 --cv 3  \
# #   --train-labels LABELS_PATH.json CIFAR10_PATH
# 
# Combine the cross-validation folds into a single predicted prob matrix
# $ python3 cifar10_train_crossval.py \
#     -a resnet50 --cvn 4 --combine-folds CIFAR10_PATH
#
# This script can also be used to train on CLEANED datasets, like this:
# python3 cifar10_train_crossval.py \
#     -a resnet50 -b 256 --lr 0.1 --gpu 0 --train-labels LABELS_PATH.json \
#     --dir-train-mask PATH_TO_CLEAN_DATA_BOOL_MASK.npy CIFAR10_PATH


# These imports enhance Python2/3 compatibility.
from __future__ import (
    print_function, absolute_import, division, unicode_literals, with_statement,
)

import argparse
import copy
import json
import os
import random
import shutil
import time
import warnings

import numpy as np
import torch
import torch.backends.cudnn as cudnn
import torch.distributed as dist
import torch.multiprocessing as mp
import torch.nn as nn
import torch.nn.parallel
import torch.optim
import torch.utils.data
import torch.utils.data.distributed
import torchvision.datasets as datasets
import torchvision.models as models
import torchvision.transforms as transforms
from sklearn.model_selection import StratifiedKFold

num_classes = 10


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: resnet18)')
parser.add_argument('-j', '--workers', default=4, type=int, metavar='N',
                    help='number of data loading workers (default: 4)')
parser.add_argument('--epochs', default=350, type=int, metavar='N',
                    help='number of epochs to run. Use 250 for Co-Teaching')
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=64, type=int,
                    metavar='N',
                    help='mini-batch size (default: 64), this is the total '
                         'batch size of all GPUs on the current node when '
                         'using Data Parallel or Distributed Data Parallel.'
                         'Use 128 for Co-Teaching.')
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=50, 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('--cv-seed', default=0, type=int,
                    help='seed for determining the cv folds. ')
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('--cv', '--cv-fold', type=int, default=None,
                    metavar='N', help='The fold to holdout')
parser.add_argument('--cvn', '--cv-n-folds', default=0, type=int,
                    metavar='N', help='The number of folds')
parser.add_argument('-m', '--dir-train-mask', default=None, type=str,
                    help='Boolean mask with True for indices to '
                         'train with and false for indices to skip.')
parser.add_argument('--combine-folds', action='store_true', default=False,
                    help='Pass this flag and -a arch to combine probs from all'
                         'folds. You must pass -a and -cvn flags as well!')
parser.add_argument('--train-labels', type=str, default=None,
                    help='DIR of training labels format: json filename2integer')
parser.add_argument('--coteaching', action='store_true',
                    help='Use Co-Teaching algorithm to train (Han et al 2018).')
parser.add_argument('--num-iter-per-epoch', type=int, default=400,
                    help='In each epoch, only train for this many iterations.'
                        'Total number of examples trained per epoch is'
                         'args.num_iter_per_epoch * args.batch_size')
parser.add_argument('--forget-rate', type=float, default=0.2,
                    help='Co-Teaching forget rate. Set to % noise if you can.')
parser.add_argument('--num_gradual', type=int, default=10,
                    help='Co-Teaching num epochs for linear drop rate,'
                         'can be 5, 10, 15. This parameter is Tk for R(T) in'
                         'Co-teaching paper.')
parser.add_argument('--exponent', type=float, default=1,
                    help='Co-Teaching exponent of the forget rate, can be'
                         '0.5, 1, 2. This parameter is equal to c in Tc for'
                         'R(T) in Co-teaching paper.')
parser.add_argument('--epoch_decay_start', type=int, default=80,
                    help='Co-Teaching number of epochs to train before'
                         'starting to decay the learning rate.')
parser.add_argument('--turn-off-save-checkpoint', action='store_true',
                    help='Prevents saving model at every epoch of training.')

best_acc1 = 0


def main(args):
    if args.seed is not None:
        random.seed(args.seed)
        torch.manual_seed(args.seed)
        torch.cuda.manual_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
    use_crossval = args.cvn > 0
    use_mask = args.dir_train_mask is not None
    cv_fold = args.cv
    cv_n_folds = args.cvn
    class_weights = None

    if use_crossval and use_mask:
        raise ValueError(
            'Either args.cvn > 0 or dir-train-mask not None, but not both.')

    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
    if args.coteaching:
        if args.gpu is None:
            raise AssertionError('if --co-teaching used, --gpu must be used.')
        torch.cuda.set_device(args.gpu)
        from cleanlab.models.cifar_cnn import CNN
        from cleanlab import coteaching
        model = CNN(input_channel=3, n_outputs=num_classes)
        model.cuda(args.gpu)
        model2 = CNN(input_channel=3, n_outputs=num_classes)
        model2.cuda(args.gpu)
    elif args.pretrained:
        print("=> using pre-trained model '{}'".format(args.arch))
        model = models.__dict__[args.arch](
            pretrained=True, num_classes=num_classes)
    else:
        print("=> creating model '{}'".format(args.arch))
        model = models.__dict__[args.arch](num_classes=num_classes)

    if args.coteaching:
        pass  # Do nothing
    elif args.distributed:
        # For multiprocessing, 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)
            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 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()

    if args.coteaching:
        optimizer = torch.optim.Adam(model.parameters(), lr=args.lr)
        optimizer2 = torch.optim.Adam(model2.parameters(), lr=args.lr)
        # Set-up learning rate scheduler alpha and betas for Adam Optimizer
        alpha_plan, beta1_plan = coteaching.initialize_lr_scheduler(
            lr=args.lr,
            epochs=args.epochs,
            epoch_decay_start=args.epoch_decay_start,
        )
        # Schedule fraction of examples to forget at each Co-Teaching epoch.
        forget_rate_schedule = coteaching.forget_rate_scheduler(
            args.epochs, args.forget_rate, args.num_gradual, args.exponent)
    else:
        optimizer = torch.optim.SGD(
            model.parameters(),
            args.lr,
            momentum=args.momentum,
            weight_decay=args.weight_decay,
        )

    # 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:
                # In case you load checkpoint from 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, 'test')

    train_dataset = datasets.ImageFolder(
        traindir,
        transforms.Compose([
            transforms.RandomCrop(32, padding=4),
            transforms.RandomHorizontalFlip(),
            transforms.ToTensor(),
            transforms.Normalize((0.4914, 0.4822, 0.4465),
                                 (0.2023, 0.1994, 0.2010)),
        ]),
    )

    # if training labels are provided use those instead of dataset labels
    if args.train_labels is not None:
        with open(args.train_labels, 'r') as rf:
            train_labels_dict = json.load(rf)
        train_dataset.imgs = [(fn, train_labels_dict[fn]) for fn, _ in
                              train_dataset.imgs]
        train_dataset.samples = train_dataset.imgs

    # If training only on cross-validated portion & make val_set = train_holdout
    if use_crossval:
        checkpoint_fn = "model_{}__fold_{}__checkpoint.pth.tar".format(
            args.arch, cv_fold)
        print('Computing fold indices. This takes 15 seconds.')
        # Prepare labels
        labels = [label for img, label in datasets.ImageFolder(traindir).imgs]
        # Split train into train and holdout for particular cv_fold.
        kf = StratifiedKFold(n_splits=cv_n_folds, shuffle=True,
                             random_state=args.cv_seed)
        cv_train_idx, cv_holdout_idx = (
            list(kf.split(range(len(labels)), labels))[cv_fold])
        # Separate datasets
        np.random.seed(args.cv_seed)
        holdout_dataset = copy.deepcopy(train_dataset)
        holdout_dataset.imgs = [train_dataset.imgs[i] for i in cv_holdout_idx]
        holdout_dataset.samples = holdout_dataset.imgs
        train_dataset.imgs = [train_dataset.imgs[i] for i in cv_train_idx]
        train_dataset.samples = train_dataset.imgs
        print('Train size:', len(cv_train_idx), len(train_dataset.imgs))
        print('Holdout size:', len(cv_holdout_idx), len(holdout_dataset.imgs))
    else:
        checkpoint_fn = "model_{}__checkpoint.pth.tar".format(args.arch)
        if use_mask:
            checkpoint_fn = "model_{}__masked__checkpoint.pth.tar".format(
                args.arch)
            orig_class_counts = np.bincount(
                [lab for img, lab in datasets.ImageFolder(traindir).imgs],
                minlength=num_classes,
            )
            train_bool_mask = np.load(args.dir_train_mask)
            # Mask labels
            train_dataset.imgs = [img for i, img in
                                  enumerate(train_dataset.imgs) if
                                  train_bool_mask[i]]
            train_dataset.samples = train_dataset.imgs
            clean_class_counts = np.bincount(
                [lab for img, lab in train_dataset.imgs],
                minlength=num_classes,
            )
            # We divide by this, so make sure no count is zero
            clean_class_counts = np.asarray(
                [1 if z == 0 else z for z in clean_class_counts])
            print('Train size:', len(train_dataset.imgs))
            # Compute class weights to re-weight loss during training
            # Should use the confident joint to estimate the noise matrix then
            # class_weights = 1 / p(s=k, y=k) for each class k.
            # Here we approximate this with a simpler approach
            # class_weights = count(y=k) / count(s=k, y=k)
            class_weights = torch.Tensor(orig_class_counts / clean_class_counts)
            class_weights = class_weights.cuda(args.gpu)

    val_dataset = datasets.ImageFolder(
        valdir,
        transforms.Compose([
            transforms.ToTensor(),
            transforms.Normalize((0.4914, 0.4822, 0.4465),
                                 (0.2023, 0.1994, 0.2010)),
        ]),
    )

    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=not args.distributed,
        num_workers=args.workers,
        pin_memory=True,
        sampler=train_sampler,
        drop_last=True,  # Don't train on last batch: could be 1 noisy example.
    )

    val_loader = torch.utils.data.DataLoader(
        val_dataset,
        batch_size=args.batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True,
    )

    # define loss function (criterion)
    criterion = nn.CrossEntropyLoss(weight=class_weights).cuda(args.gpu)
    # define separate loss function for val set that does not use class_weights
    val_criterion = nn.CrossEntropyLoss(weight=None).cuda(args.gpu)

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

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

        # Train for one epoch
        if args.coteaching:
            coteaching.adjust_learning_rate(
                optimizer, epoch, alpha_plan, beta1_plan)
            coteaching.adjust_learning_rate(
                optimizer2, epoch, alpha_plan, beta1_plan)
            coteaching.train(train_loader, epoch, model, optimizer, model2,
                             optimizer2, args, forget_rate_schedule,
                             class_weights, accuracy)
            # Evaluate
            val_acc1, val_acc2 = coteaching.evaluate(val_loader, model, model2)
            print('Epoch [%d/%d] Test Accuracy on the %s test images: Model1 '
                  '%.4f %% Model2 %.4f %%' % (
                    epoch + 1, args.epochs, len(val_dataset), val_acc1,
                    val_acc2))
        else:
            adjust_learning_rate(optimizer, epoch, args)
            train(train_loader, model, criterion, optimizer, epoch, args)

        # Evaluate on validation set
        acc1 = validate(val_loader, model, criterion, args)
        if args.coteaching:
            print('Model 2:')
            validate(val_loader, model2, val_criterion, args)

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

        if (
                not args.turn_off_save_checkpoint
                and 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=is_best,
                filename=checkpoint_fn,
                cv_fold=cv_fold,
                use_mask=use_mask,
            )
    if use_crossval:
        holdout_loader = torch.utils.data.DataLoader(
            holdout_dataset,
            batch_size=args.batch_size, shuffle=False,
            num_workers=args.workers, pin_memory=True,
        )

        if not args.turn_off_save_checkpoint:  # Load best of saved checkpoints
            print("=> loading best model_{}__fold_{}_best.pth.tar".format(
                args.arch, cv_fold))
            checkpoint = torch.load(
                "model_{}__fold_{}_best.pth.tar".format(args.arch, cv_fold))
            model.load_state_dict(checkpoint['state_dict'])
        print("Running forward pass on holdout set of size:",
              len(holdout_dataset.imgs))
        probs = get_probs(holdout_loader, model, args)
        np.save('model_{}__fold_{}__probs.npy'.format(args.arch, cv_fold),
                probs)


def train(train_loader, model, criterion, optimizer, epoch, args):
    batch_time = AverageMeter()
    data_time = AverageMeter()
    losses = AverageMeter()
    top1 = AverageMeter()
    top5 = AverageMeter()

    # switch to train mode
    model.train()

    end = time.time()
    for i, (input, target) in enumerate(train_loader):
        # if batch is size 1, skip because batch-norm will fail
        if len(input) <= 1:
            continue

        # measure data loading time
        data_time.update(time.time() - end)

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

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

        # measure accuracy and record loss
        acc1, acc5 = accuracy(output, target, topk=(1, 5))
        losses.update(loss.item(), input.size(0))
        top1.update(acc1[0], input.size(0))
        top5.update(acc5[0], input.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:
            print('Epoch: [{0}][{1}/{2}]\t'
                  'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
                  'Data {data_time.val:.3f} ({data_time.avg:.3f})\t'
                  'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
                  'Acc@1 {top1.val:.3f} ({top1.avg:.3f})\t'
                  'Acc@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
                    epoch, i, len(train_loader), batch_time=batch_time,
                    data_time=data_time, loss=losses, top1=top1, top5=top5))


def get_probs(loader, model, args):
    # Switch to evaluate mode.
    model.eval()
    n_total = len(loader.dataset.imgs) / float(loader.batch_size)
    outputs = []
    with torch.no_grad():
        end = time.time()
        for i, (input, target) in enumerate(loader):
            print("\rComplete: {:.1%}".format(i / n_total), end="")
            if args.gpu is not None:
                input = input.cuda(args.gpu, non_blocking=True)
            target = target.cuda(args.gpu, non_blocking=True)

            # compute output
            outputs.append(model(input))

    # Prepare outputs as a single matrix
    probs = np.concatenate([
        torch.nn.functional.softmax(z, dim=1) if args.gpu is None else
        torch.nn.functional.softmax(z, dim=1).cpu().numpy()
        for z in outputs
    ])

    return probs


def validate(val_loader, model, criterion, args):
    batch_time = AverageMeter()
    losses = AverageMeter()
    top1 = AverageMeter()
    top5 = AverageMeter()

    # switch to evaluate mode
    model.eval()

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

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

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

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

            if i % args.print_freq == 0:
                print('Test: [{0}/{1}]\t'
                      'Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t'
                      'Loss {loss.val:.4f} ({loss.avg:.4f})\t'
                      'Acc@1 {top1.val:.3f} ({top1.avg:.3f})\t'
                      'Acc@5 {top5.val:.3f} ({top5.avg:.3f})'.format(
                    i, len(val_loader), batch_time=batch_time, loss=losses,
                    top1=top1, top5=top5))

        print(' * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f}'
              .format(top1=top1, top5=top5))

    return top1.avg


def save_checkpoint(state, is_best, filename='checkpoint.pth.tar', cv_fold=None,
                    use_mask=False):
    torch.save(state, filename)
    if is_best:
        sm = "__masked" if use_mask else ""
        sf = "__fold_{}".format(cv_fold) if cv_fold is not None else ""
        wfn = 'model_{}{}{}_best.pth.tar'.format(state['arch'], sm, sf)
        shutil.copyfile(filename, wfn)


class AverageMeter(object):
    """Computes and stores the average and current value"""

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

    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 adjust_learning_rate(optimizer, epoch, args):
    """Sets the learning rate to the initial LR decayed by 10 every 30 epochs
    0.1 for epoch [0,150)
    0.01 for epoch [150,250)
    0.001 for epoch [250,350)"""

    if epoch < int(150. / 350 * args.epochs):
        lr = args.lr
    elif epoch < int(250. / 350 * args.epochs):
        lr = args.lr * 0.1
    else:
        lr = args.lr * 0.01

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

    print('epoch:', epoch + 1, '| 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.reshape(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


def combine_folds(args):
    wfn = 'cifar10__train__model_{}__pyx.npy'.format(args.arch)
    print('Make sure you specified the model architecture with flag -a.')
    print('This method will overwrite file: {}'.format(wfn))
    print('Computing fold indices. This takes 15 seconds.')
    # Prepare labels
    labels = [label for img, label in
              datasets.ImageFolder(os.path.join(args.data, "train/")).imgs]
    # Initialize pyx array (output of trained network)
    pyx = np.empty((len(labels), num_classes))

    # Split train into train and holdout for each cv_fold.
    kf = StratifiedKFold(n_splits=args.cvn, shuffle=True,
                         random_state=args.cv_seed)
    for k, (cv_train_idx, cv_holdout_idx) in enumerate(
            kf.split(range(len(labels)), labels)):
        probs = np.load('model_{}__fold_{}__probs.npy'.format(args.arch, k))
        pyx[cv_holdout_idx] = probs[:, :num_classes]
    print('Writing final predicted probabilities.')
    np.save(wfn, pyx)

    # Compute overall accuracy
    print('Computing Accuracy.', flush=True)
    acc = sum(np.array(labels) == np.argmax(pyx, axis=1)) / float(len(labels))
    print('Accuracy: {:.25}'.format(acc))


if __name__ == '__main__':
    arg_parser = parser.parse_args()
    if arg_parser.combine_folds:
        combine_folds(arg_parser)
    else:
        main(arg_parser)