train_crossview_sat.py

import os
import time
import copy
import yaml
import random
import logging
_logger = logging.getLogger('train')
import argparse
import builtins
import warnings

import torch
import torch.nn.parallel
import torch.distributed as dist
import torch.optim
import torch.multiprocessing as mp
import torch.utils.data
import torch.utils.data.distributed
import torchvision.models as models

from timm.utils import setup_default_logging
from datetime import datetime
from dataset.CVACT import CVACTTrainSat,CVACTVal
from dataset.CVUSA import CVUSATrainSat,CVUSAVal
from dataset.VIGOR import VIGORTrain,VigorVal

from torch.utils.data import Subset
from eval.evaluate import evaluate
from model.sample4geo import Sample4Geo
from model.infonce import InfoNCELoss
from utils import *

os.environ["CUDA_DEVICE_ORDER"] = "PCI_BUS_ID"
os.environ["TORCH_DISTRIBUTED_DEBUG"]="DETAIL"
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('-j', '--workers', default=16, type=int, metavar='N',
                    help='number of data loading workers (default: 16)')
parser.add_argument('--epochs', default=40, 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('--warmup-epoch', default=6, type=int, metavar='N',
                    help='warmup epoch for semi-supervised learning')
parser.add_argument('-b', '--batch-size', default=128, type=int,
                    metavar='N',
                    help='mini-batch size (default: 128), this is the total '
                         'batch size of all GPUs on the current node when '
                         'using Data Parallel or Distributed Data Parallel')
parser.add_argument('--eval-batch-size', default=128, type=int,
                    metavar='N',
                    help='mini-batch size (default: 128), 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', default=0.0001, type=float,
                    help='initial learning rate')
parser.add_argument('-p', '--print-freq', default=10, type=int,
                    metavar='N', help='print frequency (default: 10)')
parser.add_argument('--checkpoint', default=None, type=str, metavar='PATH',
                    help='path to latest checkpoint (default: none)')
parser.add_argument('--save_path', default='./result/', type=str, metavar='PATH',
                    help='path to save checkpoint (default: none)')
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://localhost:10000', 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('--data-folder', default='./data/CVACT', type=str, metavar='PATH',
                    help='path to dataset')
parser.add_argument('--dataset', default='cvact', type=str,
                    help='vigor, cvusa, cvact')
parser.add_argument('--same-area',default=True)
parser.add_argument('--cities',default=1, type=int,help="the trained cities")
parser.add_argument('--op', default='adamw', type=str)
parser.add_argument('--grd-size',type=int, nargs='+', default=[384, 384],help="the size of ground images")
parser.add_argument('--sat-size',type=int, nargs='+', default=[384, 384],help="the size of satellite images")
parser.add_argument('--mean',type=int, nargs='+', default=[0.485, 0.456, 0.406],help="the mean of normalized images")
parser.add_argument('--std',type=int, nargs='+', default=[0.229, 0.224, 0.225],help="the std of normalized images")
parser.add_argument('--eval-freq',default=4, type=int,help="the frequency of evaluation")
parser.add_argument('--gt-ratio', default=0., type=float,
                    help='the ratio of ground-truth labels for training')
parser.add_argument('--labeled-ratio', default=0.4, type=float,
                    help='the ratio of labeled images. When the ratio of labeled images is largger than default value, we will introduce the unlabeled images as negative samples.')
parser.add_argument('--threshold', default=0.035, type=float,
                    help='the threshold of threshold filter')
parser.add_argument('-e', '--evaluate', dest='evaluate', action='store_true',
                    help='evaluate model on validation set')


def main():
    args = parser.parse_args()
    print(args)
    args_dict = vars(args)
    
    if args.multiprocessing_distributed or (dist.is_initialized() and dist.get_rank() == 0):
        if not os.path.exists(args.save_path):
            os.makedirs(args.save_path)
        else:
            timestamp = time.time()
            local_time = time.localtime(timestamp)
            time_str = time.strftime("%Y-%m-%d-%H-%M-%S", local_time)
            args.save_path = os.path.join(args.save_path,args.dataset,'crossview_sat',time_str)
            os.makedirs(args.save_path)
            
    with open(os.path.join(args.save_path,"args.yaml"), 'w') as file:
        yaml.dump(args_dict, file)

        
    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
        mp.spawn(main_worker, nprocs=ngpus_per_node, args=(ngpus_per_node, args))
    else:
        main_worker(args.gpu, ngpus_per_node, args)


def main_worker(gpu, ngpus_per_node, args):
    is_best = False
    best_acc = 0.
    
    setup_default_logging(log_path=f'{args.save_path}/train.log')
    
    args.gpu = gpu
    args.ngpus_per_node = ngpus_per_node
    
    # suppress printing if not master
    if args.multiprocessing_distributed and args.gpu != 0:
        def print_pass(*args):
            pass
        builtins.print = print_pass

    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)
    
    set_up_system(args)
    
    # create model
    end = time.time()
    if args.gpu==0:
        _logger.info("=> creating model")
    
    if not args.multiprocessing_distributed or (dist.is_initialized() and args.gpu == 0):
        if not os.path.exists(args.save_path):
            os.mkdir(args.save_path)
    
        
    model = Sample4Geo(args)
    
    if args.distributed:
        # For multiprocessing distributed, DistributedDataParallel constructor
        # should always set the single device scope, otherwise,
        # DistributedDataParallel will use all available devices.
        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model)
        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],find_unused_parameters=True)
        else:
            model.cuda()
            model = torch.nn.parallel.DistributedDataParallel(model,find_unused_parameters=True)
            
    elif args.gpu is not None:
        model.cuda(args.gpu)
    else:
        # DataParallel will divide and allocate batch_size to all available GPUs
        model = torch.nn.DataParallel(model).cuda()
    
    parameters = list(filter(lambda p: p.requires_grad, model.parameters()))
    
    # compute_complexity(model, args)  # uncomment to see detailed computation cost
    criterion = InfoNCELoss(args).cuda(args.gpu)
    parameters.extend(criterion.parameters())
    
    optimizer = torch.optim.AdamW(parameters, args.lr)
    
    # optionally checkpoint from a checkpoint
    if args.checkpoint:
        if os.path.isfile(args.checkpoint):
            if args.gpu==0:
                _logger.info("=> loading checkpoint '{}'".format(args.checkpoint))
            if args.gpu is None:
                checkpoint = torch.load(args.checkpoint)
            else:
                # Map model to be loaded to specified single gpu.
                loc = 'cuda:{}'.format(args.gpu)
                checkpoint = torch.load(args.checkpoint, map_location=loc)
            # args.start_epoch = checkpoint['epoch']
            # best_acc = checkpoint['best_acc']
            best_acc = checkpoint['best_acc_cross']
            
            # model.load_state_dict(checkpoint['model'])
            model.load_state_dict(checkpoint['cross_model'])
            
            # optimizer.load_state_dict(checkpoint['optimizer'])
            optimizer.load_state_dict(checkpoint['cross_optimizer'])
            
            
            
            
            if args.gpu==0:
                _logger.info("=> loaded checkpoint '{}' (epoch {})".format(args.checkpoint, checkpoint['epoch']))
        else:
            if args.gpu==0:
                _logger.info("=> no checkpoint found at '{}'".format(args.checkpoint))

    if args.gpu==0:
        _logger.info(f"=> creating model cost '{time.time()-end}'")
    end = time.time()
    
    if args.gpu==0:
        _logger.info("=> creating dataset")
    
    if args.dataset.lower() == "cvusa":
        val_query_dataset = CVUSAVal(args)
    elif args.dataset.lower() == "cvact":
        val_query_dataset = CVACTVal(args)
    elif args.dataset.lower() == "vigor":
        val_query_dataset = VigorVal(args)
    else:
        print('not implemented!')
        raise Exception
    
    val_reference_dataset = copy.deepcopy(val_query_dataset)
    val_reference_dataset.img_type = "sat"

    val_query_loader = torch.utils.data.DataLoader(
        val_query_dataset,batch_size=args.eval_batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True) # 512, 64
    val_reference_loader = torch.utils.data.DataLoader(
        val_reference_dataset, batch_size=args.eval_batch_size, shuffle=False,
        num_workers=args.workers, pin_memory=True) # 80, 128
    
    if args.evaluate:
        if not args.multiprocessing_distributed or args.gpu == 0:
            evaluate(args, model, val_reference_loader,val_query_loader)
        return
    
    if args.dataset.lower() == "cvusa":
        train_dataset = CVUSATrainSat(args)
    elif args.dataset.lower() == "cvact":
        train_dataset = CVACTTrainSat(args)
    elif args.dataset.lower() == "vigor":
        train_dataset = VIGORTrain(args)
    else:
        print('not implemented!')
        raise Exception
    
    scan_dataset = copy.deepcopy(train_dataset)
    scan_dataset.mode = "scan"
    
    if args.distributed:
        train_sampler = torch.utils.data.distributed.DistributedSampler(train_dataset)
        scan_sampler = torch.utils.data.distributed.DistributedSampler(scan_dataset)
    else:
        train_sampler = None
        scan_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, drop_last=True)
    
    scan_loader = torch.utils.data.DataLoader(
            scan_dataset, batch_size=args.eval_batch_size, shuffle=scan_sampler is None,
            num_workers=args.workers, pin_memory=True, sampler=scan_sampler, drop_last=False)
    
    if args.gpu==0:
        _logger.info(f"=> creating dataset cost {time.time() - end}")

    
    if args.gpu==0:
        _logger.info("cross-view semi-supervised with sat images")
        
    for epoch in range(args.start_epoch,args.epochs):
        if args.distributed:
            train_sampler.set_epoch(epoch)
        
        if args.gpu==0:
            _logger.info('epoch:{}, compute pseudo labels at :{}'.format(epoch, datetime.now()))
        
        end = time.time()
        pseudo = {}
        if epoch >= args.warmup_epoch:
            if args.dataset == "vigor":
                pseudo = get_pseudo_labels_vigor(scan_loader,model,args,epoch)
            else:
                pseudo = get_pseudo_labels(scan_loader,model,args,epoch)
            if args.gpu == 0:
                _logger.info(f'Compute pseudo labels cost time {time.time() - end}')
            
            true_counts = 0
            for key,value in pseudo.items():
                label1 = train_loader.dataset.samples[key]
                label2 = train_loader.dataset.samples[train_loader.dataset.shuffle_samples[value]]
                if label1 == label2:
                    true_counts +=1
                    
            if args.gpu == 0:
                _logger.info(f'In epoch {epoch}, the number of total labels is: {len(pseudo)}, where the number of true is: {true_counts} \n')
        else:
            if args.gpu == 0:
                _logger.info(f'the epoch is smaller than {args.warmup_epoch}, so we only train the model in the labeled images')
        
        train(train_loader, model, criterion, optimizer, epoch, args,pseudo=pseudo)
        
        
        if not args.multiprocessing_distributed or (args.multiprocessing_distributed and args.rank % ngpus_per_node == 0):
            if  (epoch + 1) % args.eval_freq == 0 or (epoch + 1) == args.epochs: 
                result = evaluate(args, model, val_reference_loader,val_query_loader)
                _logger.info(f"=========================Recall==========================\n {result}")
                # remember best acc@1 and save checkpoint
                is_best = result[0] > best_acc
                best_acc = max(result[0], best_acc)
        
            if not args.multiprocessing_distributed or (args.multiprocessing_distributed and args.gpu % ngpus_per_node == 0):
                save_checkpoint({
                    'epoch': epoch + 1,
                    'model': model.state_dict(),
                    'best_acc': best_acc,
                    'optimizer': optimizer.state_dict(),
                }, is_best, filename=f'checkpoint.pth.tar', save_path=args.save_path)
                
    if not args.multiprocessing_distributed or (args.multiprocessing_distributed and args.gpu % ngpus_per_node == 0):        
        current_file_path = os.path.abspath(__file__)
        current_file_dir = os.path.dirname(current_file_path)
        ckpt_path = f"{current_file_dir}/ckpt/{args.dataset.lower()}"
        if not os.path.exists(ckpt_path):
            os.makedirs(ckpt_path)  
        # save_checkpoint({
        #                 'epoch': args.epochs,
        #                 'model': model.state_dict(),
        #                 'best_acc': best_acc,
        #                 'optimizer': optimizer.state_dict(),
        #             }, is_best=False, filename=f'train_crossviewsat.pth.tar', save_path=ckpt_path)
        
        shutil.copyfile(os.path.join(args.save_path,'model_best.pth.tar'), ckpt_path)
        

def get_pseudo_labels(scan_loader,model,args,epoch):
    batch_time = AverageMeter('Time', ':6.3f')
    progress = ProgressMeter(
        args,
        len(scan_loader),
        [batch_time],
        prefix='Extracting features: ')
    pseudo_labels = {}
    model =  model.eval()
    
    grd_features_list = []
    sat_features_list = []
    labels = []

    with torch.no_grad():
        end = time.time()
        # reference features
        for i, (grd,sat,label) in enumerate(scan_loader):
            if args.gpu is not None:
                grd = grd.cuda(args.gpu, non_blocking=True)
                sat = sat.cuda(args.gpu, non_blocking=True)
                label = label.cuda(args.gpu, non_blocking=True)
            grd, sat = model(grd,sat)
            
            grd_features = concat_all_gather(grd).detach().cpu().numpy()
            sat_features = concat_all_gather(sat).detach().cpu().numpy()
            label = concat_all_gather(label).detach().cpu().numpy()
            
            labels.append(label)
            grd_features_list.append(grd_features)
            sat_features_list.append(sat_features)

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()
            
            if i % args.print_freq == 0:
                progress.display(i)
        
        end = time.time()
        grd_features_list = np.concatenate(grd_features_list)
        sat_features_list = np.concatenate(sat_features_list)
        
        labels = np.concatenate(labels)
        
        if args.gpu == 0:
            _logger.info("mutual matching:")
            
        g_s_dist = np.matmul(grd_features_list,sat_features_list.transpose())
        s_g_dist = np.matmul(sat_features_list,grd_features_list.transpose())
        
        g2s_argmax = np.argmax(g_s_dist,axis=1)
        s2g_argmax = np.argmax(s_g_dist,axis=1)
        
        g_idx = np.arange(len(grd_features_list))
        dual_alignment_g2s = (s2g_argmax[g2s_argmax] == g_idx)
        
        if args.gpu == 0:
            _logger.info(f"the number of labels after mutual-matching is {np.sum(dual_alignment_g2s)}")
            _logger.info(f"threshold filter with a threshold {args.threshold}")
        
        logits = np.partition(g_s_dist[dual_alignment_g2s],-2,axis=1)[:,-2:]
        threshold = args.threshold * (1. + math.cos(math.pi * epoch / args.epochs)) / 2
        logits_positive = (logits[:,1] - logits[:,0]) >= threshold
        pseudo_np = np.dstack([labels[g_idx[dual_alignment_g2s][logits_positive]],labels[g2s_argmax[dual_alignment_g2s][logits_positive]]]).flatten()
        
        keys = pseudo_np[::2]
        values = pseudo_np[1::2]
        keys_array = np.fromiter(keys, dtype=np.int64)
        values_array = np.fromiter(values, dtype=np.int64)
        pseudo_labels = dict(zip(keys_array, values_array))
        
        if args.gpu == 0:
            _logger.info(f"the number of labels after threshold-filter is {len(pseudo_labels)}")
            
    return pseudo_labels
   
def get_pseudo_labels_vigor(scan_loader,model,args,epoch):
    batch_time = AverageMeter('Time', ':6.3f')
    progress = ProgressMeter(
        args,
        len(scan_loader),
        [batch_time],
        prefix='Extracting features: ')
    pseudo_labels = {}
    model =  model.eval()
    
    grd_features_list = []
    sat_features_list = []
    labels = []

    with torch.no_grad():
        end = time.time()
        # reference features
        for i, (grd,sat,label) in enumerate(scan_loader):
            if args.gpu is not None:
                grd = grd.cuda(args.gpu, non_blocking=True)
                sat = sat.cuda(args.gpu, non_blocking=True)
                label = label.cuda(args.gpu, non_blocking=True)
            grd, sat = model(grd,sat)
            
            grd_features = concat_all_gather(grd).detach().cpu().numpy()
            sat_features = concat_all_gather(sat).detach().cpu().numpy()
            label = concat_all_gather(label).detach().cpu().numpy()
            
            labels.append(label)
            grd_features_list.append(grd_features)
            sat_features_list.append(sat_features)

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()
            
            if i % args.print_freq == 0:
                progress.display(i)
        
        end = time.time()
        grd_features_list = np.concatenate(grd_features_list)
        sat_features_list = np.concatenate(sat_features_list)
        
        labels = np.concatenate(labels)
        
        if args.gpu == 0:
            _logger.info("mutual matching:")
        
        step_size = 5000
        steps0 = len(grd_features_list) // step_size + 1
        g2s_argmax = []
        s2g_argmax = []
        g_s_dist=np.zeros([len(grd_features_list),10])

        for i in range(steps0):
            start = step_size * i
            
            end = start + step_size
            
            sim_tmp = np.matmul(grd_features_list[start:end],sat_features_list.transpose())
            
            g2s_max_value=np.max(sim_tmp,axis=1)
            g2s_argmax.extend([np.where(row==g2s_max_value[i])[0] for i,row in enumerate(sim_tmp)])
            
            g_s_dist[start:end,:] = np.partition(sim_tmp,-10,axis=-1)[:,-10:]
            
            del sim_tmp
    
        steps1 = len(sat_features_list) // step_size + 1
        for i in range(steps1):
            start = step_size * i
            
            end = start + step_size
            
            sim_tmp = np.matmul(sat_features_list[start:end],grd_features_list.transpose())
            
            s2g_argmax.extend(np.argmax(sim_tmp,axis=1))
            del sim_tmp
        del grd_features_list
        del sat_features_list 

        s2g_argmax=np.array(s2g_argmax)
        dual_alignment_g2s = []
        for i,j in enumerate(g2s_argmax):
            if i in s2g_argmax[j]:
                for j_0 in j:
                    dual_alignment_g2s.append([i,j_0])
        dual_alignment_g2s=np.array(dual_alignment_g2s)
        
        matched = g_s_dist[dual_alignment_g2s[:,0]]
        
        if args.gpu == 0:
            _logger.info(f"the number of labels after mutual-matching is {np.sum(dual_alignment_g2s)}")
            _logger.info(f"threshold filter with a threshold {args.threshold}")
        logits = []
        for i,_ in enumerate(matched):
            unique_values = np.unique(matched[i,:])
            logit = np.partition(unique_values,-2,axis=-1)[-2:]
            logits.append(logit)
        logits = np.array(logits)
        threshold = args.threshold * (1. + math.cos(math.pi * epoch / args.epochs)) / 2
        logits_positive = (logits[:,1] - logits[:,0]) >= threshold
        pseudo_np = np.dstack([labels[dual_alignment_g2s[:,0]][logits_positive],labels[dual_alignment_g2s[:,1]][logits_positive]]).flatten()
        
        keys = pseudo_np[::2]
        values = pseudo_np[1::2]
        keys_array = np.fromiter(keys, dtype=np.int64)
        values_array = np.fromiter(values, dtype=np.int64)
        pseudo_labels = dict(zip(keys_array, values_array))
        
        if args.gpu == 0:
            _logger.info(f"the number of labels after threshold-filter is {len(pseudo_labels)}")
            
    return pseudo_labels     
                
def train(train_loader, model, criterion, optimizer, epoch, args,pseudo=None):
    org_shuffle_samples = train_loader.dataset.shuffle_samples
    nums = len(org_shuffle_samples)
    gt_ratio = train_loader.dataset.gt_ratio
    if gt_ratio > 0.:
        if args.gpu == 0:
            _logger.info(f"your ground truth ratio is {gt_ratio}")
        pseudo_dict = dict(zip(org_shuffle_samples[:int(nums*gt_ratio)], org_shuffle_samples[:int(nums*gt_ratio)]))
        pseudo.update(pseudo_dict)

    keys = [key for key in pseudo.keys()]
    values = [values for values in pseudo.values()]
    shuffle_id = np.arange(nums)
    random.seed(42 if args.seed is None else args.seed)
    random.shuffle(shuffle_id)
    random_others = [item for item in shuffle_id if item not in values]
    shuffle_samples = np.zeros(nums,dtype=np.int64) - 1
    for key,value in pseudo.items():
        shuffle_samples[key] = org_shuffle_samples[value]
    shuffle_samples = [random.choice(random_others) if item==-1 else item for item in shuffle_samples]
    train_loader.dataset.shuffle_samples = shuffle_samples
    
    subset_dataloader = None
    if len(keys) < int(nums * args.labeled_ratio):
        if args.gpu == 0:
            _logger.info(f"your pseudo labels {len(keys)} are lower than {int(nums * args.labeled_ratio)}, so we only train the model without any other unlabeled images")
        
        num_duplication = max([1,(len(train_loader.dataset) // (len(keys)+1))+1])
            
        subset = Subset(train_loader.dataset,keys * num_duplication)
        
        if args.distributed:
            subset_train_sampler =torch.utils.data.DistributedSampler(subset)
        else:
            subset_train_sampler = None
        subset_dataloader = torch.utils.data.DataLoader(
            subset, batch_size=train_loader.batch_size, shuffle=(subset_train_sampler is None),
            num_workers=train_loader.num_workers, pin_memory=True, sampler=subset_train_sampler,drop_last=True)
        
        if args.distributed:
            subset_train_sampler.set_epoch(epoch)
    
    if args.gpu==0:
        _logger.info('start epoch:{}, date:{}'.format(epoch, datetime.now()))
    lr = adjust_learning_rate(optimizer, epoch,lr=args.lr,total_epoch=args.epochs)
    if args.gpu==0:
        _logger.info(f"The learning rate of epoch {epoch} is {lr}")
    
    current_loader = subset_dataloader if subset_dataloader is not None else train_loader
    batch_time = AverageMeter('Time', ':6.3f')
    data_time = AverageMeter('Data', ':6.3f')
    losses = AverageMeter('Loss', ':.4e')
    progress = ProgressMeter(
        args,
        len(current_loader),
        [batch_time, data_time, losses],
        prefix="Epoch: [{}]".format(epoch))

    model.train()
    
    end = time.time()
    for i, (grd,sat,labels) in enumerate(current_loader):
        GT = np.isin(labels,keys)
        # measure data loading time
        data_time.update(time.time() - end)
        if args.gpu is not None:
            grd = grd.cuda(args.gpu, non_blocking=True)
            sat = sat.cuda(args.gpu, non_blocking=True)
        embed_q, embed_k = model(im_q = grd, im_k=sat)
        
        loss = criterion(embed_q, embed_k,GT=GT,label_smoothing=0.1)
            
        losses.update(loss.item(), grd.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)
            
    train_loader.dataset.shuffle_samples = org_shuffle_samples



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

    def display(self, batch):
        entries = [self.prefix + self.batch_fmtstr.format(batch)]
        entries += [str(meter) for meter in self.meters]
        if self.args.gpu==0:
            _logger.info('\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) + ']'

if __name__ == '__main__':
    main()