"""train.py * modified from https://github.com/pytorch/examples/blob/master/imagenet/main.py * supports NASNet search spaces """ import argparse import os import sys import random import shutil import time import warnings import functools 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 nasnet import thop model_names = sorted(name for name in models.__dict__ if name.islower() and not name.startswith("__") and callable(models.__dict__[name])) model_names += ('nasnet',) parser = argparse.ArgumentParser(description='PyTorch ImageNet Training') parser.add_argument('data', metavar='DIR', help='path to dataset') parser.add_argument('--synthetic', action='store_true', help='use synthetic data') parser.add_argument('-a', '--arch', metavar='ARCH', default='mobilenet_v2', choices=model_names, help='model architecture: ' + ' | '.join(model_names) + ' (default: resnet18)') parser.add_argument('-g', '--genotype', default='/path/to/genotype.txt', help='a protobuf text file defining the architecture') parser.add_argument('-j', '--workers', default=32, type=int, metavar='N', help='number of data loading workers (default: 16)') parser.add_argument('--epochs', default=250, 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=4e-5, type=float, metavar='W', help='weight decay (default: 4e-5)', dest='weight_decay') parser.add_argument('--label-smooth', default=0.1, type=float, help='label smooth for cross entropy loss') parser.add_argument('--auxiliary', action='store_true', help='use auxiliary tower') parser.add_argument('--auxiliary-weight', default=0.4, type=float) parser.add_argument('--grad-clip', default=5., type=float) 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('--save', default='', type=str, metavar='PATH', help='path to saved checkpoint (default: none)') parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true', help='evaluate model on validation set') 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('--ip', default='224.66.41.62', type=str, help='url used to set up distributed training') parser.add_argument('--port', default='2222', 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('--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('--validate-architecture', action='store_true') best_acc1 = 0 print = functools.partial(print, flush=True) def adjust_lr(optimizer, epoch, args): # Smaller slope for the last 5 epochs because lr * 1/250 is relatively large if args.epochs - epoch > 5: lr = args.lr * (args.epochs - 5 - epoch) / (args.epochs - 5) else: lr = args.lr * (args.epochs - epoch) / ((args.epochs - 5) * 5) for param_group in optimizer.param_groups: param_group['lr'] = lr return lr def main(): args = parser.parse_args() if args.seed is not None: random.seed(args.seed) torch.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.') dist_url = "tcp://{}:{}".format(args.ip,args.port) if 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)) dist_url = "tcp://{}:{}".format(args.ip,args.port) if args.distributed: if 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=dist_url, world_size=args.world_size, rank=args.rank) # create model print("=> creating model '{}'".format(args.arch)) if args.arch == 'nasnet': assert args.auxiliary is True #model = nasnet.pdarts() model = nasnet.network(genotype_file=args.genotype) else: assert args.auxiliary is False model = models.__dict__[args.arch]() if args.validate_architecture: print('=> validate architecture') data = torch.rand(1, 3, 224, 224) flops, params = thop.profile(model, (data,), verbose=False) print('* FLOPs {:.2f}M Params {:.2f}M'.format(flops/1e6, params/1e6)) if 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() if args.validate_architecture: assert flops <= 600e6, 'FLOPs > 600M, invalid' assert params <= 20e6, 'Params > 20M, invalid' data = torch.rand(16, 3, 224, 224).cuda(args.gpu) for _ in range(10): model(data) return # define loss function (criterion), optimizer, and learning rate scheduler criterion = CrossEntropyLabelSmooth(1000, args.label_smooth).cuda(args.gpu) 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)) checkpoint = torch.load(args.resume) 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']) # scheduler.load_state_dict(checkpoint['scheduler']) 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]) train_dataset = datasets.ImageFolder( traindir, transforms.Compose([ transforms.RandomResizedCrop(224), transforms.RandomHorizontalFlip(), transforms.ColorJitter( brightness=0.4, contrast=0.4, saturation=0.4, hue=0.2), transforms.ToTensor(), normalize, ])) if not args.synthetic else \ datasets.FakeData(102400, num_classes=1000, transform=transforms.ToTensor()) if args.distributed: #train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset) train_sampler = InfiniteBatchSampler(train_dataset, args.batch_size) else: train_sampler = torch.utils.data.BatchSampler(torch.utils.data.RandomSampler(train_dataset), args.batch_size, True) train_loader = torch.utils.data.DataLoader( train_dataset, num_workers=args.workers, pin_memory=True, batch_sampler=train_sampler) #train_dataset, batch_size=args.batch_size, shuffle=(train_sampler is None), #num_workers=args.workers, pin_memory=True, sampler=train_sampler) train_loader = iter(train_loader) steps_per_epoch = 1280000 // args.batch_size // args.world_size val_loader = torch.utils.data.DataLoader( datasets.ImageFolder(valdir, transforms.Compose([ transforms.Resize(256), transforms.CenterCrop(224), transforms.ToTensor(), normalize, ])) if not args.synthetic else datasets.FakeData(20000, num_classes=1000, transform=transforms.ToTensor()), batch_size=args.batch_size, shuffle=False, num_workers=args.workers, pin_memory=True) 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) current_lr = adjust_lr(optimizer, epoch, args) if epoch < 5 and args.world_size * args.batch_size > 256: print('=> warmup epoch', epoch, 'lr', current_lr * (epoch + 1) / 5.0) for param_group in optimizer.param_groups: param_group['lr'] = current_lr * (epoch + 1) / 5.0 #else: # print('=> epoch', epoch, 'lr', scheduler.base_lrs[0] * (1 - epoch / args.epochs)) # for param_group in optimizer.param_groups: # param_group['lr'] = scheduler.base_lrs[0] * (1 - epoch / args.epochs) # train for one epoch train(train_loader, model, criterion, optimizer, epoch, steps_per_epoch, args) # evaluate on validation set #acc1 = validate(val_loader, model, criterion, args) # remember best acc@1 and save checkpoint #is_best = acc1 > best_acc1 #best_acc1 = max(acc1, best_acc1) #print('best_acc', best_acc1) #if not args.multiprocessing_distributed or (args.multiprocessing_distributed # and args.rank == 0): if args.rank == 0: save_checkpoint({ 'epoch': epoch + 1, 'arch': args.arch, 'state_dict': model.state_dict(), 'best_acc1': best_acc1, 'optimizer': optimizer.state_dict(), }, False, savedir=args.save) for param_group in optimizer.param_groups: param_group['lr'] = 0 train(train_loader, model, criterion, optimizer, 250, 100, args) acc1 = validate(val_loader, model, criterion, args) print('Final Acc', acc1) if args.rank == 0: save_checkpoint({ 'epoch': 250, 'arch': args.arch, 'state_dict': model.state_dict(), 'best_acc1': acc1, 'optimizer': optimizer.state_dict(), }, True, savedir=args.save) open(os.path.join(args.save, 'SCORE'), 'w').write(str(acc1)) def train(train_loader, model, criterion, optimizer, epoch, steps, 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( steps, [batch_time, data_time, losses, top1, top5], prefix="Epoch: [{}]".format(epoch)) # switch to train mode model.train() end = time.time() #for i, (images, target) in enumerate(train_loader): i = 0 for images, target in train_loader: # measure data loading time data_time.update(time.time() - end) if args.gpu is not None: images = images.cuda(args.gpu, non_blocking=True) target = target.cuda(args.gpu, non_blocking=True) # compute output output = model(images) if args.auxiliary: loss = criterion(output[0], target) loss_aux = criterion(output[1], target) loss += args.auxiliary_weight * loss_aux output = output[0] else: 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() #nn.utils.clip_grad_norm_(model.parameters(), args.grad_clip) optimizer.step() # measure elapsed time batch_time.update(time.time() - end) end = time.time() if i % args.print_freq == 0: progress.display(i) i = i + 1 if i >= steps: break 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) target = target.cuda(args.gpu, non_blocking=True) # compute output output = model(images) if args.auxiliary: output = output[0] 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}' .format(top1=top1, top5=top5)) return top1.avg def save_checkpoint(state, is_best, savedir='', filename='checkpoint.pth.tar'): torch.save(state, os.path.join(savedir, filename)) if is_best: shutil.copyfile(filename, os.path.join(savedir, 'model_best.pth.tar')) class CrossEntropyLabelSmooth(nn.Module): def __init__(self, num_classes, epsilon): super(CrossEntropyLabelSmooth, self).__init__() self.num_classes = num_classes self.epsilon = epsilon self.logsoftmax = nn.LogSoftmax(dim=1) def forward(self, inputs, targets): log_probs = self.logsoftmax(inputs) targets = torch.zeros_like(log_probs).scatter_(1, targets.unsqueeze(1), 1) targets = (1 - self.epsilon) * targets + self.epsilon / self.num_classes loss = (-targets * log_probs).mean(0).sum() return loss 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 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].view(-1).float().sum(0, keepdim=True) res.append(correct_k.mul_(100.0 / batch_size)) return res class InfiniteBatchSampler(torch.utils.data.distributed.DistributedSampler): def __init__(self, dataset, batch_size): super().__init__(dataset) self.batch_size = batch_size def __iter__(self): g = torch.Generator() g.manual_seed(self.rank) while True: yield torch.randint(len(self.dataset), [self.batch_size], generator=g).tolist() def __len__(self): return sys.maxsize # viewed as infinite if __name__ == '__main__': main() print("over")