目录 前言 1. parse_opt函数 2. main函数 2.1 main函数打印关键词/安装环境 2.2 main函数是否进行断点训练 2.3 main函数是否分布式训练 2.4 main函数是否进化训练/遗传算法调参 3. train函数 3.1 trai
目录
- 前言
- 1. parse_opt函数
- 2. main函数
- 2.1 main函数——打印关键词/安装环境
- 2.2 main函数——是否进行断点训练
- 2.3 main函数——是否分布式训练
- 2.4 main函数——是否进化训练/遗传算法调参
- 3. train函数
- 3.1 train函数——基本配置信息
- 3.2 train函数——模型加载/断点训练
- 3.3 train函数——冻结训练/冻结层设置
- 3.4 train函数——图片大小/batchsize设置
- 3.5 train函数——优化器选择 / 分组优化设置
- 3.6 train函数——学习率/ema/归一化/单机多卡
- 3.7 train函数——数据加载 / anchor调整
- 3.8 train函数——训练配置/多尺度训练/热身训练
- 3.9 train函数——训练结束/打印信息/保存结果
- 4. run函数
- 5.全部代码注释
- 使用教程
- 总结
前言
最近在用yolov5参加比赛，yolov5的技巧很多，仅仅用来参加比赛，着实有点浪费，所以有必要好好学习一番，在认真学习之前，首先向yolov5的作者致敬，对了我是用的版本是v6。每每看到这些大神的作品，实在是有点惭愧，要学的太多了
1. parse_opt函数
def parse_opt(known=False):
"""
argparse 使用方法：
parse = argparse.ArgumentParser()
parse.add_argument('--s', type=int, default=2, help='flag_int')
"""
parser = argparse.ArgumentParser()
# weights 权重的路径./weights/yolov5s.pt....
# yolov5提供4个不同深度不同宽度的预训练权重 用户可以根据自己的需求选择下载
parser.add_argument('--weights', type=str, default=ROOT / 'yolov5s.pt', help='initial weights path')
# cfg 配置文件（网络结构） anchor/backbone/numclasses/head，训练自己的数据集需要自己生成
# 生成方式——例如我的yolov5s_mchar.yaml 根据自己的需求选择复制./models/下面.yaml文件，5个文件的区别在于模型的深度和宽度依次递增
parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
# data 数据集配置文件（路径） train/val/label/， 该文件需要自己生成
# 生成方式——例如我的/data/mchar.yaml 训练集和验证集的路径 + 类别数 + 类别名称
parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path')
# hpy超参数设置文件（lr/sgd/mixup）./data/hyps/下面有5个超参数设置文件，每个文件的超参数初始值有细微区别，用户可以根据自己的需求选择其中一个
parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path')
# epochs 训练轮次， 默认轮次为300次
parser.add_argument('--epochs', type=int, default=300)
# batchsize 训练批次， 默认bs=16
parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs, -1 for autobatch')
# imagesize 设置图片大小, 默认640*640
parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)')
# rect 是否采用矩形训练，默认为False
parser.add_argument('--rect', action='store_true', help='rectangular training')
# resume 是否接着上次的训练结果，继续训练
parser.add_argument('--resume', nargs='?', const=True, default=False, help='resume most recent training')
# nosave 不保存模型 默认False(保存) 在./runs/exp*/train/weights/保存两个模型 一个是最后一次的模型 一个是最好的模型
# best.pt/ last.pt 不建议运行代码添加 --nosave
parser.add_argument('--nosave', action='store_true', help='only save final checkpoint')
# noval 最后进行测试, 设置了之后就是训练结束都测试一下， 不设置每轮都计算mAP, 建议不设置
parser.add_argument('--noval', action='store_true', help='only validate final epoch')
# noautoanchor 不自动调整anchor, 默认False, 自动调整anchor
parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor')
# evolve参数进化， 遗传算法调参
parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations')
# bucket谷歌优盘 / 一般用不到
parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
# cache 是否提前缓存图片到内存，以加快训练速度，默认False
parser.add_argument('--cache', type=str, nargs='?', const='ram', help='--cache images in "ram" (default) or "disk"')
# mage-weights 使用图片采样策略，默认不使用
parser.add_argument('--image-weights', action='store_true', help='use weighted image selection for training')
# device 设备选择
parser.add_argument('--device', default='', help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
# multi-scale 多测度训练
parser.add_argument('--multi-scale', action='store_true', help='vary img-size +/- 50%%')
# single-cls 数据集是否多类/默认True
parser.add_argument('--single-cls', action='store_true', help='train multi-class data as single-class')
# optimizer 优化器选择 / 提供了三种优化器
parser.add_argument('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW'], default='SGD', help='optimizer')
# sync-bn:是否使用跨卡同步BN,在DDP模式使用
parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
# workers/dataloader的最大worker数量
parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
# 保存路径 / 默认保存路径 ./runs/ train
parser.add_argument('--project', default=ROOT / 'runs/train', help='save to project/name')
# 实验名称
parser.add_argument('--name', default='exp', help='save to project/name')
# 项目位置是否存在 / 默认是都不存在
parser.add_argument('--exist-ok', action='store_true', help='existing project/name ok, do not increment')
parser.add_argument('--quad', action='store_true', help='quad dataloader')
# cos-lr 余弦学习率
parser.add_argument('--cos-lr', action='store_true', help='cosine LR scheduler')
# 标签平滑 / 默认不增强， 用户可以根据自己标签的实际情况设置这个参数，建议设置小一点 0.1 / 0.05
parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
# 早停止忍耐次数 / 100次不更新就停止训练
parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)')
# --freeze冻结训练 可以设置 default = [0] 数据量大的情况下，建议不设置这个参数
parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2')
# --save-period 多少个epoch保存一下checkpoint
parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)')
# --local_rank 进程编号 / 多卡使用
parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify')
# Weights & Biases arguments
# 在线可视化工具，类似于tensorboard工具，想了解这款工具可以查看https://zhuanlan.zhihu.com/p/266337608
parser.add_argument('--entity', default=None, help='W&B: Entity')
# upload_dataset: 是否上传dataset到wandb tabel(将数据集作为交互式 dsviz表 在浏览器中查看、查询、筛选和分析数据集) 默认False
parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='W&B: Upload data, "val" option')
# bbox_interval: 设置界框图像记录间隔 Set bounding-box image logging interval for W&B 默认-1 opt.epochs // 10
parser.add_argument('--bbox_interval', type=int, default=-1, help='W&B: Set bounding-box image logging interval')
# 使用数据的版本
parser.add_argument('--artifact_alias', type=str, default='latest', help='W&B: Version of dataset artifact to use')
# 传入的基本配置中没有的参数也不会报错# parse_args()和parse_known_args()
# parse = argparse.ArgumentParser()
# parse.add_argument('--s', type=int, default=2, help='flag_int')
# parser.parse_args() / parse_args()
opt = parser.parse_known_args()[0] if known else parser.parse_args()
return opt
2. main函数
2.1 main函数——打印关键词/安装环境
def main(opt, callbacks=Callbacks()):
############################################### 1. Checks ##################################################
if RANK in [-1, 0]:
# 输出所有训练参数 / 参数以彩色的方式表现
print_args(FILE.stem, opt)
# 检查代码版本是否更新
check_git_status()
# 检查安装是否都安装了 requirements.txt， 缺少安装包安装。
# 缺少安装包：建议使用 pip install -i https://pypi.tuna.tsinghua.edu.cn/simple -r requirements.txt
check_requirements(exclude=['thop'])
2.2 main函数——是否进行断点训练
############################################### 2. Resume ##################################################
# 初始化可视化工具wandb,wandb使用教程看https://zhuanlan.zhihu.com/p/266337608
# 断点训练使用教程可以查看：https://blog.csdn.net/CharmsLUO/article/details/123410081
if opt.resume and not check_wandb_resume(opt) and not opt.evolve: # resume an interrupted run
# isinstance()是否是已经知道的类型
# 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
# 判断是否是文件
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 safe_load()yaml文件加载数据
with open(Path(ckpt).parent.parent / 'opt.yaml', errors='ignore') as f:
# argparse.Namespace 可以理解为字典
opt = argparse.Namespace(**yaml.safe_load(f)) # replace
opt.cfg, opt.weights, opt.resume = '', ckpt, True # reinstate
# 打印断点训练信息
LOGGER.info(f'Resuming training from {ckpt}')
else:
# 不使用断点训练就在加载输入的参数
opt.data, opt.cfg, opt.hyp, opt.weights, opt.project = \
check_file(opt.data), check_yaml(opt.cfg), check_yaml(opt.hyp), str(opt.weights), str(opt.project) # checks
assert len(opt.cfg) or len(opt.weights), 'either --cfg or --weights must be specified'
# opt.evolve=False,opt.name='exp' opt.evolve=True,opt.name='evolve'
if opt.evolve:
if opt.project == str(ROOT / 'runs/train'): # if default project name, rename to runs/evolve
opt.project = str(ROOT / 'runs/evolve')
opt.exist_ok, opt.resume = opt.resume, False # pass resume to exist_ok and disable resume
# 保存相关信息
opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))
2.3 main函数——是否分布式训练
# ############################################## 3.DDP mode ###############################################
# 选择设备cpu/cuda
device = select_device(opt.device, batch_size=opt.batch_size)
# 多卡训练GPU
if LOCAL_RANK != -1:
msg = 'is not compatible with YOLOv5 Multi-GPU DDP training'
assert not opt.image_weights, f'--image-weights {msg}'
assert not opt.evolve, f'--evolve {msg}'
assert opt.batch_size != -1, f'AutoBatch with --batch-size -1 {msg}, please pass a valid --batch-size'
assert opt.batch_size % WORLD_SIZE == 0, f'--batch-size {opt.batch_size} must be multiple of WORLD_SIZE'
assert torch.cuda.device_count() > LOCAL_RANK, 'insufficient CUDA devices for DDP command'
# 根据编号选择设备
#使用torch.cuda.set_device()可以更方便地将模型和数据加载到对应GPU上, 直接定义模型之前加入一行代码即可
# torch.cuda.set_device(gpu_id) #单卡
# torch.cuda.set_device('cuda:'+str(gpu_ids)) #可指定多卡
torch.cuda.set_device(LOCAL_RANK)
device = torch.device('cuda', LOCAL_RANK)
# 初始化多进程
dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")
2.4 main函数——是否进化训练/遗传算法调参
################################################ 4. Train #################################################
# 不设置evolve直接调用train训练
if not opt.evolve:
train(opt.hyp, opt, device, callbacks)
# 分布式训练 WORLD_SIZE=主机的数量
# 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... ')
# 使用多卡训练, 那么销毁进程组
dist.destroy_process_group()
# Evolve hyperparameters (optional)
# 遗传净化算法/一边训练一遍进化
# 了解遗传算法可以查看我的博客：
else:
# Hyperparameter evolution metadata (mutation scale 0-1, lower_limit, upper_limit)
# 超参数列表(突变范围 - 最小值 - 最大值)
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, errors='ignore') as f:
# 加载yaml超参数
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
# 保存进化的超参数列表
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 {evolve_csv}') # download evolve.csv if exists
"""
遗传算法调参：遵循适者生存、优胜劣汰的法则，即寻优过程中保留有用的，去除无用的。
遗传算法需要提前设置4个参数: 群体大小/进化代数/交叉概率/变异概率
"""
# 默认选择进化300代
for _ in range(opt.evolve): # generations to evolve
if evolve_csv.exists(): # if evolve.csv exists: select best hyps and mutate
# Select parent(s)
# 进化方式--single / --weight
parent = 'single' # parent selection method: 'single' or 'weighted'
# 加载evolve.txt文件
x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
# 选取进化结果代数
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
# # 获取突变初始值
mp, s = 0.8, 0.2 # mutation probability, sigma
npr = np.random
npr.seed(int(time.time()))
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)
# 将突变添加到base hyp上
# [i+7]是因为x中前7个数字为results的指标(P,R,mAP,F1,test_loss=(box,obj,cls)),之后才是超参数hyp
v = (g * (npr.random(ng) < mp) * npr.randn(ng) * npr.random() * s + 1).clip(0.3, 3.0)
for i, k in enumerate(hyp.keys()): # plt.hist(v.ravel(), 300)
hyp[k] = float(x[i + 7] * v[i]) # mutate
# Constrain to limits
# 限制超参再规定范围
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
# Train mutation
# 训练 使用突变后的参超 测试其效果
results = train(hyp.copy(), opt, device, callbacks)
callbacks = Callbacks()
# Write mutation results
# Write mutation results
# 将结果写入results 并将对应的hyp写到evolve.txt evolve.txt中每一行为一次进化的结果
# 每行前七个数字 (P, R, mAP, F1, test_losses(GIOU, obj, cls)) 之后为hyp
# 保存hyp到yaml文件
print_mutation(results, hyp.copy(), save_dir, opt.bucket)
# Plot results
# 将结果可视化 / 输出保存信息
plot_evolve(evolve_csv)
LOGGER.info(f'Hyperparameter evolution finished {opt.evolve} generations\n'
f"Results saved to {colorstr('bold', save_dir)}\n"
f'Usage example: $ python train.py --hyp {evolve_yaml}')
3. train函数
3.1 train函数——基本配置信息
################################################ 1. 传入参数/基本配置 #############################################
# opt传入的参数
save_dir, epochs, batch_size, weights, single_cls, evolve, data, cfg, resume, noval, nosave, workers, freeze = \
Path(opt.save_dir), opt.epochs, opt.batch_size, opt.weights, opt.single_cls, opt.evolve, opt.data, opt.cfg, \
opt.resume, opt.noval, opt.nosave, opt.workers, opt.freeze
# Directories
w = save_dir / 'weights' # weights dir
# 新建文件夹 weights train evolve
(w.parent if evolve else w).mkdir(parents=True, exist_ok=True) # make dir
# 保存训练结果的目录 如runs/train/exp*/weights/last.pt
last, best = w / 'last.pt', w / 'best.pt'
# Hyperparameters # isinstance()是否是已知类型
if isinstance(hyp, str):
with open(hyp, errors='ignore') as f:
# 加载yaml文件
hyp = yaml.safe_load(f) # load hyps dict
# 打印超参数 彩色字体
LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
# Save run settings
# 如果不使用进化训练
if not evolve:
# safe_dump() python值转化为yaml序列化
with open(save_dir / 'hyp.yaml', 'w') as f:
yaml.safe_dump(hyp, f, sort_keys=False)
with open(save_dir / 'opt.yaml', 'w') as f:
# vars(opt) 的作用是把数据类型是Namespace的数据转换为字典的形式。
yaml.safe_dump(vars(opt), f, sort_keys=False)
# Loggers
data_dict = None
if RANK in [-1, 0]:
loggers = Loggers(save_dir, weights, opt, hyp, LOGGER) # loggers instance
if loggers.wandb:
data_dict = loggers.wandb.data_dict
if resume:
weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size
# Register actions
for k in methods(loggers):
callbacks.register_action(k, callback=getattr(loggers, k))
# Config 画图
plots = not evolve # create plots
# GPU / CPU
cuda = device.type != 'cpu'
# 随机种子
init_seeds(1 + RANK)
# 存在子进程-分布式训练
with torch_distributed_zero_first(LOCAL_RANK):
data_dict = data_dict or check_dataset(data) # check if None
# 训练集和验证集的位路径
train_path, val_path = data_dict['train'], data_dict['val']
# 设置类别 是否单类
nc = 1 if single_cls else int(data_dict['nc']) # number of classes
# 类别对应的名称
names = ['item'] if single_cls and len(data_dict['names']) != 1 else data_dict['names'] # class names
# 判断类别长度和文件是否对应
assert len(names) == nc, f'{len(names)} names found for nc={nc} dataset in {data}' # check
# 当前数据集是否是coco数据集(80个类别)
is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt') # COCO dataset
3.2 train函数——模型加载/断点训练
################################################### 2. Model ###########################################
# 检查文件后缀是否是.pt
check_suffix(weights, '.pt') # check weights
# 加载预训练权重 yolov5提供了5个不同的预训练权重，大家可以根据自己的模型选择预训练权重
pretrained = weights.endswith('.pt')
if pretrained:
# # torch_distributed_zero_first(RANK): 用于同步不同进程对数据读取的上下文管理器
with torch_distributed_zero_first(LOCAL_RANK):
# 如果本地不存在就从网站上下载
weights = attempt_download(weights) # download if not found locally
# 加载模型以及参数
ckpt = torch.load(weights, map_location='cpu') # load checkpoint to CPU to avoid CUDA memory leak
"""
两种加载模型的方式: opt.cfg / ckpt['model'].yaml
使用resume-断点训练: 选择ckpt['model']yaml创建模型, 且不加载anchor
使用断点训练时,保存的模型会保存anchor,所以不需要加载
"""
model = Model(cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
exclude = ['anchor'] if (cfg or hyp.get('anchors')) and not resume else [] # exclude keys
csd = ckpt['model'].float().state_dict() # checkpoint state_dict as FP32
# 筛选字典中的键值对 把exclude删除
csd = intersect_dicts(csd, model.state_dict(), exclude=exclude) # intersect
model.load_state_dict(csd, strict=False) # load
LOGGER.info(f'Transferred {len(csd)}/{len(model.state_dict())} items from {weights}') # report
else:
# 不适用预训练权重
model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device) # create
3.3 train函数——冻结训练/冻结层设置
################################################ 3. Freeze/冻结训练 #########################################
# 冻结训练的网络层
freeze = [f'model.{x}.' for x in (freeze if len(freeze) > 1 else range(freeze[0]))] # layers to freeze
for k, v in model.named_parameters():
v.requires_grad = True # train all layers
if any(x in k for x in freeze):
LOGGER.info(f'freezing {k}')
# 冻结训练的层梯度不更新
v.requires_grad = False
3.4 train函数——图片大小/batchsize设置
# Image size
gs = max(int(model.stride.max()), 32) # grid size (max stride)
# 检查图片的大小
imgsz = check_img_size(opt.imgsz, gs, floor=gs * 2) # verify imgsz is gs-multiple
# Batch size
if RANK == -1 and batch_size == -1: # single-GPU only, estimate best batch size
batch_size = check_train_batch_size(model, imgsz)
loggers.on_params_update({"batch_size": batch_size})
3.5 train函数——优化器选择 / 分组优化设置
############################################ 4. Optimizer/优化器 ###########################################
"""
nbs = 64
batchsize = 16
accumulate = 64 / 16 = 4
模型梯度累计accumulate次之后就更新一次模型 相当于使用更大batch_size
"""
nbs = 64 # nominal batch size
accumulate = max(round(nbs / batch_size), 1) # accumulate loss before optimizing
# 权重衰减参数
hyp['weight_decay'] *= batch_size * accumulate / nbs # scale weight_decay
# 打印日志
LOGGER.info(f"Scaled weight_decay = {hyp['weight_decay']}")
# 将模型参数分为三组(weights、biases、bn)来进行分组优化
g0, g1, g2 = [], [], [] # optimizer parameter groups
for v in model.modules():
if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter): # bias
g2.append(v.bias)
if isinstance(v, nn.BatchNorm2d): # weight (no decay)
g0.append(v.weight)
elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter): # weight (with decay)
g1.append(v.weight)
# 选择优化器 / 提供了三个优化器——g0
if opt.optimizer == 'Adam':
optimizer = Adam(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999)) # adjust beta1 to momentum
elif opt.optimizer == 'AdamW':
optimizer = AdamW(g0, lr=hyp['lr0'], betas=(hyp['momentum'], 0.999)) # adjust beta1 to momentum
else:
optimizer = SGD(g0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)
# 设置优化的方式——g1 / g2
optimizer.add_param_group({'params': g1, 'weight_decay': hyp['weight_decay']}) # add g1 with weight_decay
optimizer.add_param_group({'params': g2}) # add g2 (biases)
# 打印log日志 优化信息
LOGGER.info(f"{colorstr('optimizer:')} {type(optimizer).__name__} with parameter groups "
f"{len(g0)} weight (no decay), {len(g1)} weight, {len(g2)} bias")
# 删除变量
del g0, g1, g2
3.6 train函数——学习率/ema/归一化/单机多卡
############################################ 5. Scheduler ##############################################
# 是否余弦学习率调整方式
if opt.cos_lr:
lf = one_cycle(1, hyp['lrf'], epochs) # cosine 1->hyp['lrf']
else:
lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf'] # linear
scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf) # plot_lr_scheduler(optimizer, scheduler, epochs)
# EMA
# 使用EMA（指数移动平均）对模型的参数做平均, 一种给予近期数据更高权重的平均方法, 以求提高测试指标并增加模型鲁棒。
ema = ModelEMA(model) if RANK in [-1, 0] else None
# Resume
start_epoch, best_fitness = 0, 0.0
if pretrained:
# Optimizer
if ckpt['optimizer'] is not None:
optimizer.load_state_dict(ckpt['optimizer'])
best_fitness = ckpt['best_fitness']
# EMA
if ema and ckpt.get('ema'):
ema.ema.load_state_dict(ckpt['ema'].float().state_dict())
ema.updates = ckpt['updates']
# Epochs
start_epoch = ckpt['epoch'] + 1
if resume:
assert start_epoch > 0, f'{weights} training to {epochs} epochs is finished, nothing to resume.'
if epochs < start_epoch:
LOGGER.info(f"{weights} has been trained for {ckpt['epoch']} epochs. Fine-tuning for {epochs} more epochs.")
epochs += ckpt['epoch'] # finetune additional epochs
del ckpt, csd
# DP mode
# DP: 单机多卡模式
if cuda and RANK == -1 and torch.cuda.device_count() > 1:
LOGGER.warning('WARNING: DP not recommended, use torch.distributed.run for best DDP Multi-GPU results.\n'
'See Multi-GPU Tutorial at https://github.com/ultralytics/yolov5/issues/475 to get started.')
model = torch.nn.DataParallel(model)
# SyncBatchNorm 多卡归一化
if opt.sync_bn and cuda and RANK != -1:
model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
# 打印信息
LOGGER.info('Using SyncBatchNorm()')
3.7 train函数——数据加载 / anchor调整
# ############################################## 6. Trainloader / 数据加载 ######################################
# 训练集数据加载
train_loader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls,
hyp=hyp, augment=True, cache=None if opt.cache == 'val' else opt.cache,
rect=opt.rect, rank=LOCAL_RANK, workers=workers,
image_weights=opt.image_weights, quad=opt.quad,
prefix=colorstr('train: '), shuffle=True)
# 标签编号最大值
mlc = int(np.concatenate(dataset.labels, 0)[:, 0].max()) # max label class
# 类别总数
nb = len(train_loader) # number of batches
# 判断编号是否正确
assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'
# Process 0
# 验证集数据集加载
if RANK in [-1, 0]:
val_loader = create_dataloader(val_path, imgsz, batch_size // WORLD_SIZE * 2, gs, single_cls,
hyp=hyp, cache=None if noval else opt.cache,
rect=True, rank=-1, workers=workers * 2, pad=0.5,
prefix=colorstr('val: '))[0]
# 没有使用断点训练
if not resume:
labels = np.concatenate(dataset.labels, 0)
# c = torch.tensor(labels[:, 0]) # classes
# cf = torch.bincount(c.long(), minlength=nc) + 1. # frequency
# model._initialize_biases(cf.to(device))
if plots:
# 画出标签信息
plot_labels(labels, names, save_dir)
# Anchors
# 自适应anchor / anchor可以理解为程序预测的box
# 根据k-mean算法聚类生成新的锚框
if not opt.noautoanchor:
# 参数dataset代表的是训练集，hyp['anchor_t']是从配置文件hpy.scratch.yaml读取的超参数 anchor_t:4.0
# 当配置文件中的anchor计算bpr（best possible recall）小于0.98时才会重新计算anchor。
# best possible recall最大值1，如果bpr小于0.98，程序会根据数据集的label自动学习anchor的尺寸
check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)
# 半进度
model.half().float() # pre-reduce anchor precision
callbacks.run('on_pretrain_routine_end')
3.8 train函数——训练配置/多尺度训练/热身训练
# #################################################### 7. 训练 ###############################################
# DDP mode
# DDP：多机多卡
if cuda and RANK != -1:
model = DDP(model, device_ids=[LOCAL_RANK], output_device=LOCAL_RANK)
# Model attributes
nl = de_parallel(model).model[-1].nl # number of detection layers (to scale hyps)
hyp['box'] *= 3 / nl # scale to layers
hyp['cls'] *= nc / 80 * 3 / nl # scale to classes and layers
hyp['obj'] *= (imgsz / 640) ** 2 * 3 / nl # scale to image size and layers
# 标签平滑
hyp['label_smoothing'] = opt.label_smoothing
model.nc = nc # attach number of classes to model
model.hyp = hyp # attach hyperparameters to model
# 从训练样本标签得到类别权重（和类别中的目标数即类别频率成反比）
model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc # attach class weights
model.names = names
# Start training
t0 = time.time()
# # 获取热身迭代的次数iterations： 3
nw = max(round(hyp['warmup_epochs'] * nb), 100) # number of warmup iterations, max(3 epochs, 100 iterations)
# nw = min(nw, (epochs - start_epoch) / 2 * nb) # limit warmup to < 1/2 of training
last_opt_step = -1
# # 初始化maps(每个类别的map)和results
maps = np.zeros(nc) # mAP per class
results = (0, 0, 0, 0, 0, 0, 0) # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
# 设置学习率衰减所进行到的轮次，即使打断训练，使用resume接着训练也能正常衔接之前的训练进行学习率衰减
scheduler.last_epoch = start_epoch - 1 # do not move
# 设置amp混合精度训练
scaler = amp.GradScaler(enabled=cuda)
# 早停止，不更新结束训练
stopper = EarlyStopping(patience=opt.patience)
# 初始化损失函数
compute_loss = ComputeLoss(model) # init loss class
# 打印信息
LOGGER.info(f'Image sizes {imgsz} train, {imgsz} val\n'
f'Using {train_loader.num_workers * WORLD_SIZE} dataloader workers\n'
f"Logging results to {colorstr('bold', save_dir)}\n"
f'Starting training for {epochs} epochs...')
# 开始走起训练
for epoch in range(start_epoch, epochs): # epoch ------------------------------------------------------------------
model.train()
# Update image weights (optional, single-GPU only)
# opt.image_weights
if opt.image_weights:
"""
如果设置进行图片采样策略，
则根据前面初始化的图片采样权重model.class_weights以及maps配合每张图片包含的类别数
通过random.choices生成图片索引indices从而进行采样
"""
cw = model.class_weights.cpu().numpy() * (1 - maps) ** 2 / nc # class weights
iw = labels_to_image_weights(dataset.labels, nc=nc, class_weights=cw) # image weights
dataset.indices = random.choices(range(dataset.n), weights=iw, k=dataset.n) # rand weighted idx
# Update mosaic border (optional)
# b = int(random.uniform(0.25 * imgsz, 0.75 * imgsz + gs) // gs * gs)
# dataset.mosaic_border = [b - imgsz, -b] # height, width borders
mloss = torch.zeros(3, device=device) # mean losses
if RANK != -1:
train_loader.sampler.set_epoch(epoch)
pbar = enumerate(train_loader)
LOGGER.info(('\n' + '%10s' * 7) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'labels', 'img_size'))
if RANK in [-1, 0]:
# 进度条显示
pbar = tqdm(pbar, total=nb, bar_format='{l_bar}{bar:10}{r_bar}{bar:-10b}') # progress bar
# 梯度清零
optimizer.zero_grad()
for i, (imgs, targets, paths, _) in pbar: # batch -------------------------------------------------------------
ni = i + nb * epoch # number integrated batches (since train start)
imgs = imgs.to(device, non_blocking=True).float() / 255 # uint8 to float32, 0-255 to 0.0-1.0
"""
热身训练(前nw次迭代)
在前nw次迭代中, 根据以下方式选取accumulate和学习率
"""
# Warmup
if ni <= nw:
xi = [0, nw] # x interp
# compute_loss.gr = np.interp(ni, xi, [0.0, 1.0]) # iou loss ratio (obj_loss = 1.0 or iou)
accumulate = max(1, np.interp(ni, xi, [1, nbs / batch_size]).round())
for j, x in enumerate(optimizer.param_groups):
"""
bias的学习率从0.1下降到基准学习率lr*lf(epoch)，
其他的参数学习率从0增加到lr*lf(epoch).
lf为上面设置的余弦退火的衰减函数
动量momentum也从0.9慢慢变到hyp['momentum'](default=0.937)
"""
# bias lr falls from 0.1 to lr0, all other lrs rise from 0.0 to lr0
x['lr'] = np.interp(ni, xi, [hyp['warmup_bias_lr'] if j == 2 else 0.0, x['initial_lr'] * lf(epoch)])
if 'momentum' in x:
x['momentum'] = np.interp(ni, xi, [hyp['warmup_momentum'], hyp['momentum']])
# Multi-scale
if opt.multi_scale:
"""
Multi-scale 设置多尺度训练，从imgsz * 0.5, imgsz * 1.5 + gs随机选取尺寸
"""
sz = random.randrange(imgsz * 0.5, imgsz * 1.5 + gs) // gs * gs # size
sf = sz / max(imgs.shape[2:]) # scale factor
if sf != 1:
ns = [math.ceil(x * sf / gs) * gs for x in imgs.shape[2:]] # new shape (stretched to gs-multiple)
imgs = nn.functional.interpolate(imgs, size=ns, mode='bilinear', align_corners=False)
# Forward / 前向传播
with amp.autocast(enabled=cuda):
pred = model(imgs) # forward
# # 计算损失，包括分类损失，objectness损失，框的回归损失
# loss为总损失值，loss_items为一个元组，包含分类损失，objectness损失，框的回归损失和总损失
loss, loss_items = compute_loss(pred, targets.to(device)) # loss scaled by batch_size
if RANK != -1:
# 平均不同gpu之间的梯度
loss *= WORLD_SIZE # gradient averaged between devices in DDP mode
if opt.quad:
loss *= 4.
# Backward
scaler.scale(loss).backward()
# Optimize # 模型反向传播accumulate次之后再根据累积的梯度更新一次参数
if ni - last_opt_step >= accumulate:
scaler.step(optimizer) # optimizer.step
scaler.update()
optimizer.zero_grad()
if ema:
ema.update(model)
last_opt_step = ni
# Log
if RANK in [-1, 0]:
mloss = (mloss * i + loss_items) / (i + 1) # update mean losses
mem = f'{torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0:.3g}G' # (GB)
pbar.set_description(('%10s' * 2 + '%10.4g' * 5) % (
f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]))
callbacks.run('on_train_batch_end', ni, model, imgs, targets, paths, plots, opt.sync_bn)
if callbacks.stop_training:
return
# end batch ------------------------------------------------------------------------------------------------
# Scheduler 进行学习率衰减
lr = [x['lr'] for x in optimizer.param_groups] # for loggers
scheduler.step()
if RANK in [-1, 0]:
# mAP
callbacks.run('on_train_epoch_end', epoch=epoch)
# 将model中的属性赋值给ema
ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights'])
# 判断当前的epoch是否是最后一轮
final_epoch = (epoch + 1 == epochs) or stopper.possible_stop
# notest: 是否只测试最后一轮 True: 只测试最后一轮 False: 每轮训练完都测试mAP
if not noval or final_epoch: # Calculate mAP
"""
测试使用的是ema（指数移动平均 对模型的参数做平均）的模型
results: [1] Precision 所有类别的平均precision(最大f1时)
[1] Recall 所有类别的平均recall
[1] map@0.5 所有类别的平均mAP@0.5
[1] map@0.5:0.95 所有类别的平均mAP@0.5:0.95
[1] box_loss 验证集回归损失, obj_loss 验证集置信度损失, cls_loss 验证集分类损失
maps: [80] 所有类别的mAP@0.5:0.95
"""
results, maps, _ = val.run(data_dict,
batch_size=batch_size // WORLD_SIZE * 2,
imgsz=imgsz,
model=ema.ema,
single_cls=single_cls,
dataloader=val_loader,
save_dir=save_dir,
plots=False,
callbacks=callbacks,
compute_loss=compute_loss)
# Update best mAP
# Update best mAP 这里的best mAP其实是[P, R, mAP@.5, mAP@.5-.95]的一个加权值
# fi: [P, R, mAP@.5, mAP@.5-.95]的一个加权值 = 0.1*mAP@.5 + 0.9*mAP@.5-.95
fi = fitness(np.array(results).reshape(1, -1)) # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
if fi > best_fitness:
best_fitness = fi
log_vals = list(mloss) + list(results) + lr
callbacks.run('on_fit_epoch_end', log_vals, epoch, best_fitness, fi)
# Save model
"""
保存带checkpoint的模型用于inference或resuming training
保存模型, 还保存了epoch, results, optimizer等信息
optimizer将不会在最后一轮完成后保存
model保存的是EMA的模型
"""
if (not nosave) or (final_epoch and not evolve): # if save
ckpt = {'epoch': epoch,
'best_fitness': best_fitness,
'model': deepcopy(de_parallel(model)).half(),
'ema': deepcopy(ema.ema).half(),
'updates': ema.updates,
'optimizer': optimizer.state_dict(),
'wandb_id': loggers.wandb.wandb_run.id if loggers.wandb else None,
'date': datetime.now().isoformat()}
# Save last, best and delete
torch.save(ckpt, last)
if best_fitness == fi:
torch.save(ckpt, best)
if (epoch > 0) and (opt.save_period > 0) and (epoch % opt.save_period == 0):
torch.save(ckpt, w / f'epoch{epoch}.pt')
del ckpt
callbacks.run('on_model_save', last, epoch, final_epoch, best_fitness, fi)
# Stop Single-GPU
if RANK == -1 and stopper(epoch=epoch, fitness=fi):
break
# Stop DDP TODO: known issues shttps://github.com/ultralytics/yolov5/pull/4576
# stop = stopper(epoch=epoch, fitness=fi)
# if RANK == 0:
# dist.broadcast_object_list([stop], 0) # broadcast 'stop' to all ranks
# Stop DPP
# with torch_distributed_zero_first(RANK):
# if stop:
# break # must break all DDP ranks
3.9 train函数——训练结束/打印信息/保存结果
############################################### 8. 打印训练信息 ##########################################
if RANK in [-1, 0]:
LOGGER.info(f'\n{epoch - start_epoch + 1} epochs completed in {(time.time() - t0) / 3600:.3f} hours.')
for f in last, best:
if f.exists():
# 模型训练完后, strip_optimizer函数将optimizer从ckpt中删除
# 并对模型进行model.half() 将Float32->Float16 这样可以减少模型大小, 提高inference速度
strip_optimizer(f) # strip optimizers
if f is best:
LOGGER.info(f'\nValidating {f}...')
results, _, _ = val.run(data_dict,
batch_size=batch_size // WORLD_SIZE * 2,
imgsz=imgsz,
model=attempt_load(f, device).half(),
iou_thres=0.65 if is_coco else 0.60, # best pycocotools results at 0.65
single_cls=single_cls,
dataloader=val_loader,
save_dir=save_dir,
save_json=is_coco,
verbose=True,
plots=True,
callbacks=callbacks,
compute_loss=compute_loss) # val best model with plots
if is_coco:
callbacks.run('on_fit_epoch_end', list(mloss) + list(results) + lr, epoch, best_fitness, fi)
# 回调函数
callbacks.run('on_train_end', last, best, plots, epoch, results)
LOGGER.info(f"Results saved to {colorstr('bold', save_dir)}")
# 释放显存
torch.cuda.empty_cache()
return results
4. run函数
def run(**kwargs):
# 执行这个脚本/ 调用train函数 / 开启训练
# Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolov5m.pt')
opt = parse_opt(True)
for k, v in kwargs.items():
# setattr() 赋值属性，属性不存在则创建一个赋值
setattr(opt, k, v)
main(opt)
return opt
5.全部代码注释
# YOLOv5