main_vgg_cifar10.py

import sys, os, random, shutil, time
import copy
import random

sys.path.append(os.path.dirname(os.path.dirname(os.path.realpath(__file__))))

from models.vgg_cifar import vgg

import torch_pruning_tool.torch_pruning as tp
import argparse
import torch
import torch.backends.cudnn as cudnn
from torchvision.datasets import CIFAR10, CIFAR100
from torchvision import transforms
import torch.nn.functional as F
import torch.nn as nn
import numpy as np
from utils.utils import AverageMeter, RecorderMeter, time_string
from utils.utils import convert_secs2time, get_ncc_sim_matrix, get_n_flops_, get_n_params_

from scipy.spatial import distance

parser = argparse.ArgumentParser()

parser.add_argument("--data_path", type=str, default="./data/cifar.python/")
parser.add_argument('--dataset', type=str, default='cifar10')
parser.add_argument('--pretrain_path', type=str, help='..path of pre-trained model')
parser.add_argument('--baseline_path', type=str, help='..path of pre-trained model')
parser.add_argument('--pruned_path', type=str, help='..path of pre-trained model')
parser.add_argument("--ckpt_path", type=str, default='')
parser.add_argument('--save_path', type=str, default='./', help='Folder to save checkpoints and log.')
parser.add_argument('--mode', type=str, required=True, choices=['train', 'eval', 'prune'])
parser.add_argument('--batch_size', type=int, default=128)
parser.add_argument('--test_batch_size', type=int, default=256, metavar='N', help='input batch size for testing (default: 256)')
parser.add_argument('--verbose', action='store_true', default=False)
parser.add_argument('--total_epoches', type=int, default=160) 
parser.add_argument('--recover_epoch', type=int, default=2)
parser.add_argument('--prune_epoch', type=int, default=60)
parser.add_argument('--exp_round', type=int, default=1)
parser.add_argument('--lr', type=float, default=0.01, metavar='LR', help='learning rate (default: 0.1)')
parser.add_argument('--schedule', type=int, nargs='+', default=[150, 225],
                    help='Decrease learning rate at these epochs.')
parser.add_argument('--gammas', type=float, nargs='+', default=[0.1, 0.1],
                    help='LR is multiplied by gamma on schedule, number of gammas should be equal to schedule')
parser.add_argument('--momentum', type=float, default=0.9, metavar='M', help='SGD momentum (default: 0.9)')
parser.add_argument('--decay', '--wd', default=1e-4, type=float, metavar='W', help='weight decay (default: 1e-4)')
parser.add_argument('--seed', type=int, default=1, metavar='S', help='random seed (default: 1)')
parser.add_argument('--depth', default=16, type=int, help='depth of the neural network')
parser.add_argument('--no-cuda', action='store_true', default=False, help='disables CUDA training')

# Acceleration
parser.add_argument('--ngpu', type=int, default=1, help='0 = CPU.')
parser.add_argument('--workers', type=int, default=2, help='number of data loading workers (default: 2)')

# compress rate
parser.add_argument('--rate_flop', type=float, default=0.342, help='This is flop reduction rate')

# random seed
parser.add_argument('--manualSeed', type=int, help='manual seed')

args = parser.parse_args()
args.cuda = not args.no_cuda and torch.cuda.is_available()
device = torch.device('cuda:0' if args.cuda else 'cpu')
args.use_cuda = args.ngpu > 0 and torch.cuda.is_available()

if args.manualSeed is None:
    args.manualSeed = random.randint(1, 10000)
random.seed(args.manualSeed)
torch.manual_seed(args.manualSeed)
if args.use_cuda:
    torch.cuda.manual_seed_all(args.manualSeed)
cudnn.benchmark = True

def main():

    # Init logger
    if not os.path.isdir(args.save_path):
        os.makedirs(args.save_path)

    log = open(os.path.join(args.save_path, 'log_seed_{}.txt'.format(args.manualSeed)), 'w')

    print_log('save path : {}'.format(args.save_path), log)
    state = {k: v for k, v in args._get_kwargs()}
    print_log(state, log)
    print_log("Random Seed: {}".format(args.manualSeed), log)
    print_log("python version : {}".format(sys.version.replace('\n', ' ')), log)
    print_log("torch  version : {}".format(torch.__version__), log)
    print_log("cudnn  version : {}".format(torch.backends.cudnn.version()), log)
    print_log("Pretrain path: {}".format(args.pretrain_path), log)

    # Init dataset
    if not os.path.isdir(args.data_path):
        os.makedirs(args.data_path)

    if args.dataset == 'cifar10':
        mean = [x / 255 for x in [125.3, 123.0, 113.9]]
        std = [x / 255 for x in [63.0, 62.1, 66.7]]
    elif args.dataset == 'cifar100':
        mean = [x / 255 for x in [129.3, 124.1, 112.4]]
        std = [x / 255 for x in [68.2, 65.4, 70.4]]
    else:
        assert False, "Unknow dataset : {}".format(args.dataset)

    transform_train = transforms.Compose(
            [transforms.RandomHorizontalFlip(), transforms.RandomCrop(32, padding=4), transforms.ToTensor(),
            transforms.Normalize(mean, std)])
    transform_test = transforms.Compose([
            transforms.ToTensor(), transforms.Normalize(mean, std)])

    if args.dataset == 'cifar10':
        dataset_train = CIFAR10(args.data_path, train=True, transform=transform_train, download=True)
        dataset_test = CIFAR10(args.data_path, train=False, transform=transform_test, download=True)
    elif args.dataset == 'cifar100':
        dataset_train = CIFAR100(args.data_path, train=True, transform=transform_train, download=True)
        dataset_test = CIFAR100(args.data_path, train=False, transform=transform_test, download=True)
    else:
        assert False, 'Do not support dataset : {}'.format(args.dataset)

    train_loader = torch.utils.data.DataLoader(dataset_train, batch_size=args.batch_size, shuffle=True, 
                                              num_workers=args.workers)
    test_loader = torch.utils.data.DataLoader(dataset_test, batch_size=args.test_batch_size, shuffle=False, 
                                              num_workers=args.workers)
    # subset of train dataset, 10000
    train_subset_index = random.sample(range(0, 49999), 10000)
    dataset_train_subset = torch.utils.data.Subset(dataset_train, train_subset_index)
    train_loader_subset = torch.utils.data.DataLoader(dataset_train_subset, batch_size=256, shuffle=False, 
                                              num_workers=args.workers)

    # model 
    model = vgg(dataset=args.dataset, depth=args.depth)
    # define loss function (criterion) and optimizer
    criterion = torch.nn.CrossEntropyLoss()

    if args.use_cuda:
        criterion.cuda()

    if args.mode == 'prune':
        
        if os.path.isfile(args.pretrain_path):
            pretrain = torch.load(args.pretrain_path)
            model = pretrain['state_dict']
        else:
            print("No pretrain model is given, training and pruning from scrach")

        if args.use_cuda:
            torch.cuda.empty_cache()
            model.cuda()
        
        # train for 0 ~ args.prune_epoch
        train(model, train_loader, test_loader, criterion, 0, args.prune_epoch, log)

        pruned_model = prune(model, train_loader, train_loader_subset, test_loader, criterion, log)
        print_log("=> network after pruning:\n {}".format(pruned_model), log, False)

        # train for remaining epoch: args.prune_epoch ~ args.total_epoches 
        train(pruned_model, train_loader, test_loader, criterion, args.prune_epoch, args.total_epoches, log)

    if args.mode == 'train':

        print_log("=> network :\n {}".format(model), log, False)

        if args.use_cuda:
            torch.cuda.empty_cache()
            model.cuda()

        train(model, train_loader, test_loader, criterion, 0, args.total_epoches,log)

    elif args.mode == 'eval':       

        base_size = get_n_params_(model)
        base_ops = get_n_flops_(model, img_size=(32, 32))

        if os.path.isfile(args.baseline_path):
            print("Loading Baseline Model from: ", args.baseline_path)
            baseline = torch.load(args.baseline_path)
            model = baseline['state_dict']
            print_log("=> Baseline network :\n {}".format(model), log, False)

            if args.use_cuda:
                torch.cuda.empty_cache()
                model.cuda()

            val_acc_top1, val_acc_top5, val_loss = validate(test_loader, model, criterion)
            print_log("Baseline Val Acc@1: %0.3lf, Acc@5: %0.3lf,  Loss: %0.5f" % (val_acc_top1, val_acc_top5, val_loss), log)

        if os.path.isfile(args.pruned_path):
            print("Load Pruned Model from %s" % (args.pruned_path))
            pruned = torch.load(args.pruned_path)
            model_pruned = pruned['state_dict']
            print_log("=> pruned network :\n {}".format(model_pruned), log, True)
            
            if args.use_cuda:
                torch.cuda.empty_cache()
                model_pruned.cuda()

            pruned_size = get_n_params_(model_pruned)
            pruned_ops = get_n_flops_(model_pruned, img_size=(32, 32))
            val_acc_top1, val_acc_top5, val_loss = validate(test_loader, model_pruned, criterion)
            print_log("Pruned Val Acc@1: %0.3lf, Acc@5: %0.3lf,  Loss: %0.5f" % (val_acc_top1, val_acc_top5, val_loss), log)

            print_log(
            "Params: {:.2f} M => {:.2f} M, (Param RR {:.2f}%)".format(
                base_size / 1e6, pruned_size / 1e6, (1.0 - pruned_size / base_size) * 100 ), log)
            print_log(
            "FLOPs: {:.2f} M => {:.2f} M (FLOPs RR {:.2f}%, Speed-Up {:.2f}X )".format(
                base_ops / 1e6,
                pruned_ops / 1e6,
                (1.0 - pruned_ops / base_ops) * 100,
                base_ops / pruned_ops ), log)

    log.close()

def prune(model, train_loader, train_loader_subset, test_loader, criterion, log):

    sub_inputs, sub_targets = get_train_subset_in_memory(train_loader_subset)

    flops_baseline = get_n_flops_(model, img_size=(32, 32))
    print_log("Baseline Model Flops: %lf" % flops_baseline, log)

    flop_rate_temp = 0
    current_flop_reduction_rate = 0.0
    step_prune_rate = 0.1
    filter_prune_limit_per_layer = 0.85

    layerwise_filter_count_org = get_conv_filter_count(model)

    methods = ['l1norm', 'l2norm', 'eucl', 'cos']
    prune_stat = {}
    for m in methods:
        prune_stat[m] = 0

    while current_flop_reduction_rate < args.rate_flop:
        small_loss = 100000000000.0
        small_loss_lindex = -1
        opt_method = ''

        for prune_conv_idx in layerwise_filter_count_org.keys():
            for method in methods:
                # model copy to prune 
                model_copy = copy.deepcopy(model)
                DG_temp = tp.DependencyGraph().build_dependency(model_copy, torch.rand(1,3,32,32))
                step_prune_count = prune_ith_conv_layer(model_copy, DG_temp, prune_conv_idx, layerwise_filter_count_org[prune_conv_idx], 
                                        step_prune_rate, filter_prune_limit_per_layer, method)
                if step_prune_count == 0:
                    continue

                # calc loss after prune 
                if args.use_cuda:
                    torch.cuda.empty_cache()
                    model_copy.cuda()
                with torch.no_grad():
                    _, sample_loss = validate_fast(sub_inputs, sub_targets, model_copy, criterion)

                # store conv layer index with small loss
                if sample_loss < small_loss: 
                    small_loss_lindex = prune_conv_idx
                    small_loss = sample_loss
                    opt_method = method

        # prune selected layer with given prune rate
        DG = tp.DependencyGraph().build_dependency(model, torch.rand(1,3,32,32))
        exp_step = prune_ith_conv_layer(model, DG, small_loss_lindex, layerwise_filter_count_org[small_loss_lindex],
                            step_prune_rate, filter_prune_limit_per_layer, opt_method)

        prune_stat[opt_method] += exp_step
        flops_pruned = get_n_flops_(model, img_size=(32, 32))
        current_flop_reduction_rate = 1.0 - flops_pruned / flops_baseline
        print("[Pruning %d filters using %s from layer %d] Flop Reduction Rate: %lf/%lf" % 
                        (exp_step, opt_method, small_loss_lindex, current_flop_reduction_rate, args.rate_flop))
 
        # train single epoch to recover recently pruned layer 
        if current_flop_reduction_rate - flop_rate_temp > 0.03:
            if args.use_cuda:
                torch.cuda.empty_cache()
                model.cuda()
            train(model, train_loader, test_loader, criterion, args.prune_epoch, args.prune_epoch + args.recover_epoch, log)

            flop_rate_temp = current_flop_reduction_rate

    print_log('Prune Stats: ' + ''.join(str(prune_stat)), log)

    layerwise_filter_count_prune = {}
    for idx, m in enumerate(model.modules()):
        if isinstance(m, nn.Conv2d):
            layerwise_filter_count_prune[idx] = m.weight.shape[0]

    print_log('Final Flop Reduction Rate: %.4lf' % current_flop_reduction_rate, log)
    print_log("Conv Filters Before Pruning: " + ''.join(str(layerwise_filter_count_org)), log)
    print_log("Conv Filters After Pruning: " + ''.join(str(layerwise_filter_count_prune)), log)    

    filter_prune_rate = {}
    for idx in layerwise_filter_count_org.keys():
        filter_prune_rate[idx] = 1.0 - float(layerwise_filter_count_prune[idx]/float(layerwise_filter_count_org[idx]))
    print_log("Layerwise Pruning Rate: " + ''.join(str(filter_prune_rate)), log)    
    
    # save pruned model before finetuning
    save_checkpoint({
        'epoch': 0,
        'arch': args.depth,
        'state_dict': model,
    }, False, args.save_path, 'vgg%s.init.prune.pth.tar' % (args.depth) )

    return model

def get_conv_filter_count(model):
    conv_filter_count = {}
    for idx, m in enumerate(model.modules()):
        if isinstance(m, nn.Conv2d):
            weight = m.weight
            conv_filter_count[idx] = int(weight.shape[0])
    return conv_filter_count

def prune_ith_conv_layer(model, DG, conv_idx, org_filter_count, step_prune_rate=0.1, 
                            max_pruning_rate=0.7, method='norm'):
    step_prune_count = 0
    for idx, m in enumerate(model.modules()):
        if isinstance(m, nn.Conv2d) and idx == conv_idx:
            nfilters = m.weight.shape[0]
            last_prune_rate = 1 - float(nfilters) / float(org_filter_count)
            if last_prune_rate < max_pruning_rate:
                step_prune_count = int(nfilters * step_prune_rate + 0.5)
                prune_conv(DG, m, step_prune_count, method)
            break
    return step_prune_count

def get_train_subset_in_memory(train_loader_subset):
    sub_inputs = []
    sub_targets = []
    for _, (input, target) in enumerate(train_loader_subset):
        if args.use_cuda:
            target = target.cuda(non_blocking=True)
            input = input.cuda(non_blocking=True)
        sub_targets.append(target)
        sub_inputs.append(input)
    return sub_inputs, sub_targets

def get_similar_matrix_cuda(weight_vec_after_norm, dist_type="eucl"):
    if dist_type == "eucl":
        similar_matrix = torch.cdist(weight_vec_after_norm, weight_vec_after_norm, p=2)
    elif dist_type == "cos":
        weight_vec_after_norm = F.normalize(weight_vec_after_norm, p=2, dim=1)
        similar_matrix = torch.matmul(weight_vec_after_norm, weight_vec_after_norm.t())
        similar_matrix = 1 - similar_matrix
        similar_matrix[torch.isnan(similar_matrix)] = 1
    return similar_matrix

def similarity_prune_cuda(conv, amount=0.2, method='eucl'):
    weight = conv.weight.detach().clone().cuda()
    total_filters = weight.shape[0]
    weight = weight.view(total_filters, -1)
    num_pruned = int(total_filters * amount) if amount < 1.0 else amount

    similar_matrix = get_similar_matrix_cuda(weight, method)
    similar_sum = torch.sum(similar_matrix, dim=1)
    _, pruning_index = torch.sort(similar_sum)
    pruning_index = pruning_index[:num_pruned].tolist()

    return pruning_index

def norm_prune(conv, amount=1, ltype='l1norm'):
    if ltype == 'l1norm':
        strategy = tp.strategy.L1Strategy()
    else:
        strategy = tp.strategy.L2Strategy()
    pruning_index = strategy(conv.weight, amount=amount)
    return pruning_index

def prune_conv(DG, conv, amount, method):
    # get index of filters to be pruned        
    if 'norm' in method:
        pruning_index = norm_prune(conv, amount, method)
        # apply pruning
        plan = DG.get_pruning_plan(conv, tp.prune_conv_out_channel, pruning_index)
        plan.exec()
    else:
        pruning_index = similarity_prune_cuda(conv, amount, method)
        # apply pruning
        plan = DG.get_pruning_plan(conv, tp.prune_conv_out_channel, pruning_index)
        plan.exec()

def train(model, train_loader, test_loader, criterion, start_epoch, total_epoch, log):

    optimizer = torch.optim.SGD(model.parameters(), args.lr, momentum=args.momentum,
            weight_decay=args.decay, nesterov=True)

    if args.use_cuda:
        torch.cuda.empty_cache()
        model.cuda()

    recorder = RecorderMeter(total_epoch)
    start_time = time.time()
    epoch_time = AverageMeter()
    best_accuracy = 0

    # Main loop
    for epoch in range(start_epoch, total_epoch):

        current_learning_rate = adjust_learning_rate(optimizer, epoch, args.gammas, args.schedule)
        #current_learning_rate = adjust_learning_rate_vgg(optimizer, epoch)

        need_hour, need_mins, need_secs = convert_secs2time(epoch_time.avg * (total_epoch - epoch))
        need_time = '[Need: {:02d}:{:02d}:{:02d}]'.format(need_hour, need_mins, need_secs)

        # train for one epoch
        train_acc, train_los = train_epoch(train_loader, model, criterion, optimizer)

        # validate
        val_acc_2, val_los_2,_ = validate(test_loader, model, criterion)
        print("Epoch %d/%d [learning_rate=%lf] Val [Acc|Loss]: %.3f %% | %0.5f" % (epoch, total_epoch, current_learning_rate, val_acc_2, val_los_2))

        is_best = recorder.update(epoch, train_los, train_acc, val_los_2, val_acc_2)
        if recorder.max_accuracy(False) > best_accuracy:
            print_log('\n==>>{:s} [Epoch={:03d}/{:03d}] {:s} [learning_rate={:6.4f}]'.format(time_string(), epoch, total_epoch,
                                                                                    need_time, current_learning_rate) \
                + ' [Best : Accuracy={:.2f}, Error={:.2f}]'.format(recorder.max_accuracy(False),
                                                                100 - recorder.max_accuracy(False)), log)
            best_accuracy = recorder.max_accuracy(False)
            is_best = True

        save_checkpoint({
            'epoch': epoch + 1,
            'arch': args.depth,
            'state_dict': model,
            'recorder': recorder,
            'optimizer': optimizer.state_dict(),
        }, is_best, args.save_path, 'checkpoint.pth.tar')

        # measure elapsed time
        epoch_time.update(time.time() - start_time)
        start_time = time.time()
        recorder.plot_curve(os.path.join(args.save_path, 'curve.png'))

# train function (forward, backward, update)
def train_epoch(train_loader, model, criterion, optimizer):
    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):
        # measure data loading time
        data_time.update(time.time() - end)

        if args.use_cuda:
            target = target.cuda(non_blocking=True)
            input = input.cuda(non_blocking=True)

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

        # measure accuracy and record loss
        prec1, prec5 = accuracy(output.data, target, topk=(1, 5))
        losses.update(loss.item(), input.size(0))
        top1.update(prec1, input.size(0))
        top5.update(prec5, input.size(0))

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

        # Mask grad for iteration
        optimizer.step()

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

    return top1.avg, losses.avg

def accuracy(output, target, topk=(1,)):
    with torch.no_grad():
        maxk = max(topk)
        # Only need to do topk for highest k, reuse for the rest 
        _, pred = output.topk(k=maxk, dim=1, largest=True, sorted=True)
        pred = pred.t()
        correct = pred.eq(target.view(1, -1).expand_as(pred))
        
        batch_size = target.size(0)

        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).item())
        return res

def validate_single_epoch(input, target, model, criterion, losses_m, top1_m, top5_m):
    if args.use_cuda:
        target = target.cuda(non_blocking=True)
        input = input.cuda(non_blocking=True)

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

    # measure accuracy and record loss
    prec1, prec5 = accuracy(output.data, target, topk=(1, 5))

    losses_m.update(loss.item(), input.size(0))
    top1_m.update(prec1, input.size(0))
    top5_m.update(prec5, input.size(0))

    return losses_m, top1_m, top5_m    

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

    # switch to evaluate mode
    model.eval()
    for i, (input, target) in enumerate(val_loader):
        losses, top1, top5 = validate_single_epoch(input, target, model, criterion, losses, top1, top5)
    return top1.avg, top5.avg, losses.avg

def validate_fast(inputs, targets, model, criterion):
    losses = AverageMeter()
    top1 = AverageMeter()
    top5 = AverageMeter()

    # switch to evaluate mode
    model.eval()
    for (input, target) in zip(inputs, targets):
        losses, top1, top5 = validate_single_epoch(input, target, model, criterion, losses, top1, top5)
    return top1.avg, losses.avg

def print_log(print_string, log, display=True):
    if display:
        print("{}".format(print_string))
    log.write('{}\n'.format(print_string))
    log.flush()

def save_checkpoint(state, is_best, save_path, filename):
    filename = os.path.join(save_path, filename)
    torch.save(state, filename)
    if is_best:
        bestname = os.path.join(save_path, 'model_best.pth.tar')
        shutil.copyfile(filename, bestname)

def adjust_learning_rate(optimizer, epoch, gammas, schedule):
    """Sets the learning rate to the initial LR decayed by 10 every 30 epochs"""
    lr = args.lr
    assert len(gammas) == len(schedule), "length of gammas and schedule should be equal"
    for (gamma, step) in zip(gammas, schedule):
        if (epoch >= step):
            lr = lr * gamma
        else:
            break
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr
    return lr

def adjust_learning_rate_vgg(optimizer, epoch):
    if epoch < args.total_epoches * 0.20: 
        return args.lr
    
    if epoch in [args.total_epoches * 0.20, args.total_epoches * 0.40, args.total_epoches * 0.60, args.total_epoches * 0.80]:
        current_learning_rate = args.lr
        for param_group in optimizer.param_groups:
            param_group['lr'] *= 0.1
            current_learning_rate = param_group['lr']
        return current_learning_rate

    if epoch >= args.total_epoches * 0.2:
        for param_group in optimizer.param_groups:
            current_learning_rate = param_group['lr']
        return current_learning_rate

if __name__=='__main__':
    main()