2021SC@SDUSC山东大学软件学院软件工程应用与实践——yolov5代码分析——第六篇——train.py(2)

2021SC@SDUSC

目录

main函数

---

请不要忽视代码中的注释

main函数

def main(opt, callbacks=Callbacks()):
    # Checks
    set_logging(RANK)
    if RANK in [-1, 0]:
        # 输出所有训练opt参数 
        print_args(FILE.stem, opt)
        # 检查代码版本是否是最新的
        check_git_status()
        # 检查requirements.txt所需包是否都满足
        check_requirements(exclude=['thop'])

logging和wandb初始化

 # Resume
    if opt.resume and not check_wandb_resume(opt) and not opt.evolve:  # resume an interrupted run
        # 使用断点续训 就从last.pt中读取相关参数
        # 如果resume是str，则表示传入的是模型的路径地址
        # 如果resume是True，则通过get_lastest_run()函数找到runs为文件夹中最近的权重文件last.pt
        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()  # specified or most recent path
        assert os.path.isfile(ckpt), 'ERROR: --resume checkpoint does not exist'
        # 相关的opt参数也要替换成last.pt中的opt参数
        with open(Path(ckpt).parent.parent / 'opt.yaml') as f:
            opt = argparse.Namespace(**yaml.safe_load(f))  # replace
        opt.cfg, opt.weights, opt.resume = '', ckpt, True  # reinstate
        LOGGER.info(f'Resuming training from {ckpt}')
    else:
        # 不使用断点续训 就从文件中读取相关参数
        opt.data, opt.cfg, opt.hyp = check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp)  # check YAMLs
        assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
        if opt.evolve:
            opt.project = 'runs/evolve'
            opt.exist_ok, opt.resume = opt.resume, False  # pass resume to exist_ok and disable resume
        # 根据opt.project生成目录
        opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))

判断是否使用断点续训resume, 读取参数

使用断点续训 就从last.pt中读取相关参数；不使用断点续训 就从文件中读取相关参数

# DDP mode
# 选择设备  cpu/cuda:0
    device = select_device(opt.device, batch_size=opt.batch_size)
    if LOCAL_RANK != -1:
        # LOCAL_RANK != -1 进行多GPU训练
        from datetime import timedelta
        assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command'
        assert opt.batch_size % WORLD_SIZE == 0, '--batch-size must be multiple of CUDA device count'
        assert not opt.image_weights, '--image-weights argument is not compatible with DDP training'
        assert not opt.evolve, '--evolve argument is not compatible with DDP training'
        torch.cuda.set_device(LOCAL_RANK)
        # 根据GPU编号选择设备
        device = torch.device('cuda', LOCAL_RANK)
        # 初始化进程组  distributed backend
        dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")

DDP mode设置

 # Train
    # 不使用进化算法 正常Train
    if not opt.evolve:
        # 如果不进行超参进化 那么就直接调用train()函数，开始训练
        train(opt.hyp, opt, device, callbacks)
         # 如果是使用多卡训练, 那么销毁进程组
        if WORLD_SIZE > 1 and RANK == 0:
            LOGGER.info('Destroying process group... ')
            dist.destroy_process_group()

不进行算法，正常训练

    else:
        # Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
        meta = {'lr0': (1, 1e-5, 1e-1),  # initial learning rate (SGD=1E-2, Adam=1E-3)
                'lrf': (1, 0.01, 1.0),  # final OneCycleLR learning rate (lr0 * lrf)
                'momentum': (0.3, 0.6, 0.98),  # SGD momentum/Adam beta1
                'weight_decay': (1, 0.0, 0.001),  # optimizer weight decay
                'warmup_epochs': (1, 0.0, 5.0),  # warmup epochs (fractions ok)
                'warmup_momentum': (1, 0.0, 0.95),  # warmup initial momentum
                'warmup_bias_lr': (1, 0.0, 0.2),  # warmup initial bias lr
                'box': (1, 0.02, 0.2),  # box loss gain
                'cls': (1, 0.2, 4.0),  # cls loss gain
                'cls_pw': (1, 0.5, 2.0),  # cls BCELoss positive_weight
                'obj': (1, 0.2, 4.0),  # obj loss gain (scale with pixels)
                'obj_pw': (1, 0.5, 2.0),  # obj BCELoss positive_weight
                'iou_t': (0, 0.1, 0.7),  # IoU training threshold
                'anchor_t': (1, 2.0, 8.0),  # anchor-multiple threshold
                'anchors': (2, 2.0, 10.0),  # anchors per output grid (0 to ignore)
                'fl_gamma': (0, 0.0, 2.0),  # focal loss gamma (efficientDet default gamma=1.5)
                'hsv_h': (1, 0.0, 0.1),  # image HSV-Hue augmentation (fraction)
                'hsv_s': (1, 0.0, 0.9),  # image HSV-Saturation augmentation (fraction)
                'hsv_v': (1, 0.0, 0.9),  # image HSV-Value augmentation (fraction)
                'degrees': (1, 0.0, 45.0),  # image rotation (+/- deg)
                'translate': (1, 0.0, 0.9),  # image translation (+/- fraction)
                'scale': (1, 0.0, 0.9),  # image scale (+/- gain)
                'shear': (1, 0.0, 10.0),  # image shear (+/- deg)
                'perspective': (0, 0.0, 0.001),  # image perspective (+/- fraction), range 0-0.001
                'flipud': (1, 0.0, 1.0),  # image flip up-down (probability)
                'fliplr': (0, 0.0, 1.0),  # image flip left-right (probability)
                'mosaic': (1, 0.0, 1.0),  # image mixup (probability)
                'mixup': (1, 0.0, 1.0),  # image mixup (probability)
                'copy_paste': (1, 0.0, 1.0)}  # segment copy-paste (probability)

        with open(opt.hyp) as f:
            hyp = yaml.safe_load(f)  # load hyps dict
            if 'anchors' not in hyp:  # anchors commented in hyp.yaml
                hyp['anchors'] = 3
        opt.noval, opt.nosave, save_dir = True, True, Path(opt.save_dir)  # only val/save final epoch
        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
        evolve_yaml, evolve_csv = save_dir / 'hyp_evolve.yaml', save_dir / 'evolve.csv'
        if opt.bucket:
            os.system(f'gsutil cp gs://{opt.bucket}/evolve.csv {save_dir}')  # download evolve.csv if exists

        for _ in range(opt.evolve):  # generations to evolve
            if evolve_csv.exists():  # if evolve.csv exists: select best hyps and mutate
                # Select parent(s)
                # 选择超参进化方式 只用single和weighted两种
                parent = 'single'  # parent selection method: 'single' or 'weighted'
                x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
                # 选取至多前五次进化的结果
                n = min(5, len(x))  # number of previous results to consider
                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
                # 根据resluts计算hyp权重
                w = fitness(x) - fitness(x).min() + 1E-6  # weights (sum > 0)
                # 根据不同进化方式获得base hyp
                if parent == 'single' or len(x) == 1:
                    # x = x[random.randint(0, n - 1)]  # random selection
                    x = x[random.choices(range(n), weights=w)[0]]  # weighted selection
                elif parent == 'weighted':
                    x = (x * w.reshape(n, 1)).sum(0) / w.sum()  # weighted combination

                # Mutate 超参进化
                mp, s = 0.8, 0.2  # mutation probability, sigma
                npr = np.random
                npr.seed(int(time.time()))
                # 获取突变初始值
                g = np.array([meta[k][0] for k in hyp.keys()])  # gains 0-1
                ng = len(meta)
                v = np.ones(ng)
                # 设置突变
                while all(v == 1):  # mutate until a change occurs (prevent duplicates)
                    v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
                 # 将突变添加到base hyp上
                for i, k in enumerate(hyp.keys()):  # plt.hist(v.ravel(), 300)
                    hyp[k] = float(x[i + 7] * v[i])  # mutate

            # Constrain to limits
            for k, v in meta.items():
                hyp[k] = max(hyp[k], v[1])  # lower limit
                hyp[k] = min(hyp[k], v[2])  # upper limit
                hyp[k] = round(hyp[k], 5)  # significant digits

            # 训练 使用突变后的参超 测试其效果
            results = train(hyp.copy(), opt, device, callbacks)

            # Write mutation results
            print_mutation(results, hyp.copy(), save_dir, opt.bucket)

        # Plot results
        plot_evolve(evolve_csv)
        print(f'Hyperparameter evolution finished\n'
              f"Results saved to {colorstr('bold', save_dir)}\n"
              f'Use best hyperparameters example: $ python train.py --hyp {evolve_yaml}')

遗传进化算法，边进化边训练。

使用遗传算法进行参数进化，默认是进化三百代。此处的进化算法是：根据之前训练时的base hyp再进行突变；

通过之前每次进化得到的results来确定之前每个hyp的权重

有了每个hyp和每个hyp的权重之后有两种进化方式：

1.根据每个htp的权重随机算则一个之前的hyp作为base hyp，random.choices(range(n),weights=w)

2.根据每个hyp的权重对之前所有的hyp进行融合获得一个base hyp， （x*w.reshape(n,1)).sum(0)/w.sum()

evolve.txt会记录每次进化之后的results进行从大到小的排序；

在根据fitness函数计算之前每次进化的到的hyp的权重，再确定哪一种进化方式，从而进行进化