train_distill.py

import argparse
import logging
import os
import random
import time
from copy import deepcopy
from pathlib import Path

import numpy as np
import torch.distributed as dist
import torch.nn as nn
import torch.optim as optim
import torch.optim.lr_scheduler as lr_scheduler
import torch.utils.data
import yaml
from torch.cuda import amp
from torch.nn.parallel import DistributedDataParallel as DDP
from torch.utils.tensorboard import SummaryWriter
from tqdm import tqdm

import test
from models.experimental import attempt_load
from models.yolo import Model
from utils.datasets import create_dataloader
from utils.general import increment_path, init_seeds, \
    fitness, strip_optimizer, get_latest_run, check_dataset, \
    check_file, check_img_size, \
    check_requirements, print_mutation, set_logging, one_cycle, colorstr
from utils.google_utils import attempt_download
from utils.loss import ComputeLoss, ComputeDstillLoss, ComputeOutbasedDstillLoss
from utils.plots import plot_results, plot_evolution
from utils.torch_utils import ModelEMA, select_device, intersect_dicts, torch_distributed_zero_first, is_parallel
from utils.wandb_logging.wandb_utils import WandbLogger, check_wandb_resume
from teacher import TeacherModel
logger = logging.getLogger(__name__)


def train(hyp, opt, device, tb_writer=None):

    logger.info(colorstr('hyperparameters: ') +
                ', '.join(f'{k}={v}' for k, v in hyp.items()))
    save_dir, epochs, batch_size, total_batch_size, weights, rank = \
        Path(opt.save_dir), opt.epochs, opt.batch_size, opt.total_batch_size, opt.weights, opt.global_rank

    # Directories
    wdir = save_dir / 'weights'
    wdir.mkdir(parents=True, exist_ok=True)  # make dir
    last = wdir / 'last.pt'
    best = wdir / 'best.pt'
    results_file = save_dir / 'results.txt'

    # Save run settings
    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:
        yaml.safe_dump(vars(opt), f, sort_keys=False)

    # Configure
    plots = not opt.evolve  # create plots
    cuda = device.type != 'cpu'
    init_seeds(2 + rank)
    with open(opt.data) as f:
        data_dict = yaml.safe_load(f)  # data dict
    is_coco = opt.data.endswith('coco.yaml')

    # Logging- Doing this before checking the dataset. Might update data_dict
    loggers = {'wandb': None}  # loggers dict
    if rank in [-1, 0]:
        opt.hyp = hyp  # add hyperparameters
        run_id = torch.load(weights).get('wandb_id') if weights.endswith(
            '.pt') and os.path.isfile(weights) else None
        wandb_logger = WandbLogger(opt, save_dir.stem, run_id, data_dict)
        loggers['wandb'] = wandb_logger.wandb
        data_dict = wandb_logger.data_dict
        if wandb_logger.wandb:
            # WandbLogger might update weights, epochs if resuming
            weights, epochs, hyp = opt.weights, opt.epochs, opt.hyp

    nc = 1 if opt.single_cls else int(data_dict['nc'])  # number of classes
    names = ['item'] if opt.single_cls and len(
        data_dict['names']) != 1 else data_dict['names']  # class names
    assert len(names) == nc, '%g names found for nc=%g dataset in %s' % (
        len(names), nc, opt.data)  # check

    # Log
    tags = ['train/box_loss', 'train/obj_loss', 'train/cls_loss', 'train/soft_loss',  # train loss
            'metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95',
            'val/box_loss', 'val/obj_loss', 'val/cls_loss',  # val loss
            'x/lr0', 'x/lr1', 'x/lr2']  # params

    # Model
    pretrained = weights.endswith('.pt')
    if pretrained:
        with torch_distributed_zero_first(rank):
            attempt_download(weights)  # download if not found locally
        ckpt = torch.load(weights, map_location=device)  # load checkpoint
        model = Model(opt.cfg or ckpt['model'].yaml, ch=3, nc=nc, anchors=hyp.get(
            'anchors')).to(device)  # create
        exclude = ['anchor'] if (opt.cfg or hyp.get(
            'anchors')) and not opt.resume else []  # exclude keys
        state_dict = ckpt['model'].float().state_dict()  # to FP32
        state_dict = intersect_dicts(
            state_dict, model.state_dict(), exclude=exclude)  # intersect
        model.load_state_dict(state_dict, strict=False)  # load
        logger.info('Transferred %g/%g items from %s' %
                    (len(state_dict), len(model.state_dict()), weights))  # report
    else:
        model = Model(opt.cfg, ch=3, nc=nc, anchors=hyp.get(
            'anchors')).to(device)  # create
    with torch_distributed_zero_first(rank):
        check_dataset(data_dict)  # check
    train_path = data_dict['train']
    test_path = data_dict['val']

    # Teacher Model
    assert os.path.exists(
        opt.teacher), '-[ERROR] tearcher weights do not exists.'
    teacher_model = TeacherModel(
        conf_thres=opt.t_conf_thres, iou_thres=opt.t_nms_thres, training=opt.full_output_loss)
    teacher_model.init_model(opt.teacher, opt.device, nc)

    # Freeze
    freeze = []  # parameter names to freeze (full or partial)
    for k, v in model.named_parameters():
        v.requires_grad = True  # train all layers
        if any(x in k for x in freeze):
            print('freezing %s' % k)
            v.requires_grad = False

    # Optimizer
    nbs = 64  # nominal batch size
    # accumulate loss before optimizing
    accumulate = max(round(nbs / total_batch_size), 1)
    hyp['weight_decay'] *= total_batch_size * \
        accumulate / nbs  # scale weight_decay
    logger.info(f"Scaled weight_decay = {hyp['weight_decay']}")

    pg0, pg1, pg2 = [], [], []  # optimizer parameter groups
    for k, v in model.named_modules():
        if hasattr(v, 'bias') and isinstance(v.bias, nn.Parameter):
            pg2.append(v.bias)  # biases
        if isinstance(v, nn.BatchNorm2d):
            pg0.append(v.weight)  # no decay
        elif hasattr(v, 'weight') and isinstance(v.weight, nn.Parameter):
            pg1.append(v.weight)  # apply decay

    if opt.adam:
        optimizer = optim.Adam(pg0, lr=hyp['lr0'], betas=(
            hyp['momentum'], 0.999))  # adjust beta1 to momentum
    else:
        optimizer = optim.SGD(
            pg0, lr=hyp['lr0'], momentum=hyp['momentum'], nesterov=True)

    # add pg1 with weight_decay
    optimizer.add_param_group(
        {'params': pg1, 'weight_decay': hyp['weight_decay']})
    optimizer.add_param_group({'params': pg2})  # add pg2 (biases)
    logger.info('Optimizer groups: %g .bias, %g conv.weight, %g other' %
                (len(pg2), len(pg1), len(pg0)))
    del pg0
    del pg1
    del pg2

    if opt.linear_lr:
        def lf(x): return (1 - x / (epochs - 1)) * \
            (1.0 - hyp['lrf']) + hyp['lrf']  # linear
    else:
        lf = one_cycle(1, hyp['lrf'], epochs)  # cosine 1->hyp['lrf']
    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
    # plot_lr_scheduler(optimizer, scheduler, epochs)

    # EMA
    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']

        # Results
        if ckpt.get('training_results') is not None:
            results_file.write_text(
                ckpt['training_results'])  # write results.txt

        # Epochs
        start_epoch = ckpt['epoch'] + 1
        if opt.resume:
            assert start_epoch > 0, '%s training to %g epochs is finished, nothing to resume.' % (
                weights, epochs)
        if epochs < start_epoch:
            logger.info('%s has been trained for %g epochs. Fine-tuning for %g additional epochs.' %
                        (weights, ckpt['epoch'], epochs))
            epochs += ckpt['epoch']  # finetune additional epochs

        del ckpt
        del state_dict

    # Image sizes
    gs = max(int(model.stride.max()), 32)  # grid size (max stride)
    # number of detection layers (used for scaling hyp['obj'])
    nl = model.model[-1].nl
    # verify imgsz are gs-multiples
    imgsz, imgsz_test = [check_img_size(x, gs) for x in opt.img_size]

    # DP mode
    if cuda and rank == -1 and torch.cuda.device_count() > 1:
        model = torch.nn.DataParallel(model)

    # SyncBatchNorm
    if opt.sync_bn and cuda and rank != -1:
        model = torch.nn.SyncBatchNorm.convert_sync_batchnorm(model).to(device)
        logger.info('Using SyncBatchNorm()')
    # Process 0
    if rank in [-1, 0]:
        testloader = create_dataloader(test_path, imgsz_test, batch_size * 2, gs, opt,  # testloader
                                       hyp=hyp, cache=opt.cache_images and not opt.notest, rect=True, rank=-1,
                                       world_size=opt.world_size, workers=opt.workers,
                                       pad=0.5, prefix=colorstr('val: '))[0]

        if not opt.resume:
            model.half().float()  # pre-reduce anchor precision

    # Trainloader
    dataloader, _ = create_dataloader(train_path, imgsz, batch_size, gs, opt,
                                      hyp=hyp, augment=True, cache=opt.cache_images, rect=opt.rect, rank=rank,
                                      world_size=opt.world_size, workers=opt.workers,
                                      image_weights=opt.image_weights, quad=opt.quad, prefix=colorstr('train: '))
    nb = len(dataloader)  # number of batches

    # DDP mode
    if cuda and rank != -1:
        model = DDP(model, device_ids=[opt.local_rank], output_device=opt.local_rank,
                    # nn.MultiheadAttention incompatibility with DDP https://github.com/pytorch/pytorch/issues/26698
                    find_unused_parameters=any(isinstance(layer, nn.MultiheadAttention) for layer in model.modules()))

    # Model parameters
    hyp['box'] *= 3. / nl  # scale to layers
    hyp['cls'] *= nc / 80. * 3. / nl  # scale to classes and layers
    # scale to image size and layers
    hyp['obj'] *= (imgsz / 640) ** 2 * 3. / nl
    hyp['label_smoothing'] = opt.label_smoothing
    model.nc = nc  # attach number of classes to model
    model.hyp = hyp  # attach hyperparameters to model
    model.gr = 1.0  # iou loss ratio (obj_loss = 1.0 or iou)
    model.class_weights = 1.0
    model.names = names

    # Start training
    t0 = time.time()
    # number of warmup iterations, max(3 epochs, 1k iterations)
    nw = max(round(hyp['warmup_epochs'] * nb), 1000)
    # nw = min(nw, (epochs - start_epoch) / 2 * nb)  # limit warmup to < 1/2 of training
    # P, R, mAP@.5, mAP@.5-.95, val_loss(box, obj, cls)
    results = (0, 0, 0, 0, 0, 0, 0)
    scheduler.last_epoch = start_epoch - 1  # do not move
    scaler = amp.GradScaler(enabled=cuda)
    compute_loss = ComputeLoss(model)  # init loss class
    if opt.full_output_loss:
        compute_distill_loss = ComputeOutbasedDstillLoss(
            nc=nc, distill_ratio=opt.distill_ratio)
    else:
        compute_distill_loss = ComputeDstillLoss(
            model, distill_ratio=opt.distill_ratio)
    logger.info(f'Image sizes {imgsz} train, {imgsz_test} test\n'
                f'Using {dataloader.num_workers} dataloader workers\n'
                f'Logging results to {save_dir}\n'
                f'Starting training for {epochs} epochs...')

    for epoch in range(start_epoch, epochs):
        model.train()

        mloss = torch.zeros(4 + 1, device=device)  # mean losses
        if rank != -1:
            dataloader.sampler.set_epoch(epoch)
        pbar = enumerate(dataloader)
        logger.info(('\n' + '%10s' * 9) % ('Epoch', 'gpu_mem', 'box',
                                           'obj', 'cls', 'soft', 'total', 'obj_num', 'img_size'))
        if rank in [-1, 0]:
            pbar = tqdm(pbar, total=nb)  # progress bar
        optimizer.zero_grad()
        for i, (imgs, targets, _, _) in pbar:
            # number integrated batches (since train start)
            ni = i + nb * epoch
            imgs = imgs.to(device, non_blocking=True).float() / \
                255.0  # uint8 to float32, 0-255 to 0.0-1.0

            # Warmup
            if ni <= nw:
                xi = [0, nw]  # x interp
                # model.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 / total_batch_size]).round())
                for j, x in enumerate(optimizer.param_groups):
                    # 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']])

            # Forward
            with amp.autocast(enabled=cuda):
                pred = model(imgs)  # forward
                if opt.full_output_loss:
                    _, t_targets = teacher_model.generate_batch_targets(
                        imgs, opt.img_size)
                else:
                    t_targets, _ = teacher_model.generate_batch_targets(
                        imgs, opt.img_size)
                    t_targets = t_targets.to(device)

                loss, loss_items = compute_distill_loss(pred, t_targets)

                gt_loss, gt_loss_items = compute_loss(
                    pred, targets.to(device))
                loss += gt_loss
                loss_items += gt_loss_items
                pred_num = targets.shape[0]

                if rank != -1:
                    loss *= opt.world_size  # gradient averaged between devices in DDP mode
                if opt.quad:
                    loss *= 4.

            # Backward
            scaler.scale(loss).backward()

            # Optimize
            if ni % accumulate == 0:
                scaler.step(optimizer)  # optimizer.step
                scaler.update()
                optimizer.zero_grad()
                if ema:
                    ema.update(model)

            # Print
            if rank in [-1, 0]:
                mloss = (mloss * i + loss_items) / \
                    (i + 1)  # update mean losses
                mem = '%.3gG' % (torch.cuda.memory_reserved(
                ) / 1E9 if torch.cuda.is_available() else 0)  # (GB)
                s = ('%10s' * 2 + '%10.4g' * 7) % (
                    '%g/%g' % (epoch, epochs - 1), mem, *mloss, pred_num, imgs.shape[-1])
                pbar.set_description(s)
            # end batch ------------------------------------------------------------------------------------------------
        # end epoch ----------------------------------------------------------------------------------------------------

        # Scheduler
        lr = [x['lr'] for x in optimizer.param_groups]  # for tensorboard
        scheduler.step()

        # DDP process 0 or single-GPU
        if rank in [-1, 0]:
            # mAP
            ema.update_attr(
                model, include=['yaml', 'nc', 'hyp', 'gr', 'names', 'stride', 'class_weights'])
            final_epoch = epoch + 1 == epochs
            if not opt.notest or final_epoch:  # Calculate mAP
                wandb_logger.current_epoch = epoch + 1
                results, _, _ = test.test(data_dict,
                                          batch_size=batch_size * 2,
                                          imgsz=imgsz_test,
                                          model=ema.ema,
                                          single_cls=opt.single_cls,
                                          dataloader=testloader,
                                          save_dir=save_dir,
                                          verbose=nc < 50 and final_epoch,
                                          plots=plots and final_epoch,
                                          wandb_logger=wandb_logger,
                                          compute_loss=compute_loss,
                                          is_coco=is_coco)

            # Write
            with open(results_file, 'a') as f:
                # append metrics, val_loss
                f.write(s + '%10.4g' * 7 % results + '\n')
            if len(opt.name) and opt.bucket:
                os.system('gsutil cp %s gs://%s/results/results%s.txt' %
                          (results_file, opt.bucket, opt.name))

            for x, tag in zip(list(mloss[:-1]) + list(results) + lr, tags):
                if tb_writer:
                    tb_writer.add_scalar(tag, x, epoch)  # tensorboard
                if wandb_logger.wandb:
                    wandb_logger.log({tag: x})  # W&B

            # Update best mAP
            # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
            fi = fitness(np.array(results).reshape(1, -1))
            if fi > best_fitness:
                best_fitness = fi
            wandb_logger.end_epoch(best_result=best_fitness == fi)

            # Save model
            if (not opt.nosave) or (final_epoch and not opt.evolve):  # if save
                ckpt = {'epoch': epoch,
                        'best_fitness': best_fitness,
                        'training_results': results_file.read_text(),
                        'model': deepcopy(model.module if is_parallel(model) else model).half(),
                        'ema': deepcopy(ema.ema).half(),
                        'updates': ema.updates,
                        'optimizer': optimizer.state_dict(),
                        'wandb_id': wandb_logger.wandb_run.id if wandb_logger.wandb else None}

                # Save last, best and delete
                torch.save(ckpt, last)
                if best_fitness == fi:
                    torch.save(ckpt, best)
                if wandb_logger.wandb:
                    if ((epoch + 1) % opt.save_period == 0 and not final_epoch) and opt.save_period != -1:
                        wandb_logger.log_model(
                            last.parent, opt, epoch, fi, best_model=best_fitness == fi)
                del ckpt

        # end epoch ----------------------------------------------------------------------------------------------------
    # end training
    if rank in [-1, 0]:
        # Plots
        if plots:
            plot_results(save_dir=save_dir)  # save as results.png
            if wandb_logger.wandb:
                files = ['results.png', 'confusion_matrix.png', *
                         [f'{x}_curve.png' for x in ('F1', 'PR', 'P', 'R')]]
                wandb_logger.log({"Results": [wandb_logger.wandb.Image(str(save_dir / f), caption=f) for f in files
                                              if (save_dir / f).exists()]})
        # Test best.pt
        logger.info('%g epochs completed in %.3f hours.\n' %
                    (epoch - start_epoch + 1, (time.time() - t0) / 3600))
        if opt.data.endswith('coco.yaml') and nc == 80:  # if COCO
            for m in (last, best) if best.exists() else (last):  # speed, mAP tests
                results, _, _ = test.test(opt.data,
                                          batch_size=batch_size * 2,
                                          imgsz=imgsz_test,
                                          conf_thres=0.0001,
                                          iou_thres=0.7,
                                          model=attempt_load(m, device).half(),
                                          single_cls=opt.single_cls,
                                          dataloader=testloader,
                                          save_dir=save_dir,
                                          save_json=True,
                                          plots=False,
                                          is_coco=is_coco)

        # Strip optimizers
        final = best if best.exists() else last  # final model
        for f in last, best:
            if f.exists():
                strip_optimizer(f)  # strip optimizers
        if opt.bucket:
            os.system(f'gsutil cp {final} gs://{opt.bucket}/weights')  # upload
        if wandb_logger.wandb and not opt.evolve:  # Log the stripped model
            wandb_logger.wandb.log_artifact(str(final), type='model',
                                            name='run_' + wandb_logger.wandb_run.id + '_model',
                                            aliases=['last', 'best', 'stripped'])
        wandb_logger.finish_run()
    else:
        dist.destroy_process_group()
    torch.cuda.empty_cache()
    return results


if __name__ == '__main__':
    parser = argparse.ArgumentParser()
    parser.add_argument('--weights', type=str,
                        default='weights/yolov5s.pt', help='initial weights path')
    parser.add_argument('--teacher', type=str,
                        default='weights/yolov5m-voc.pt', help='teacher weights path')
    parser.add_argument('--distill-ratio', type=float,
                        default=0.5, help='distill loss ratio in total loss')
    parser.add_argument('--t_conf_thres', type=float,
                        default=0.1, help='teacher confidence threshold')
    parser.add_argument('--t_nms_thres', type=float,
                        default=0.3, help='teacher nms iou threshold')
    parser.add_argument('--full-output-loss',
                        action='store_true', help='use full output to compute distill loss')
    parser.add_argument('--cfg', type=str,
                        default='models/yolov5s.yaml', help='model.yaml path')
    parser.add_argument('--data', type=str,
                        default='data/voc.yaml', help='data.yaml path')
    parser.add_argument(
        '--hyp', type=str, default='data/hyp_distill.scratch.yaml', help='hyperparameters path')
    parser.add_argument('--epochs', type=int, default=30)
    parser.add_argument('--batch-size', type=int, default=2,
                        help='total batch size for all GPUs')
    parser.add_argument('--img-size', nargs='+', type=int,
                        default=[640, 640], help='[train, test] image sizes')
    parser.add_argument('--rect', action='store_true',
                        help='rectangular training')
    parser.add_argument('--resume', nargs='?', const=True,
                        default=False, help='resume most recent training')
    parser.add_argument('--nosave', action='store_true',
                        help='only save final checkpoint')
    parser.add_argument('--notest', action='store_true',
                        help='only test final epoch')
    parser.add_argument('--evolve', action='store_true',
                        help='evolve hyperparameters')
    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
    parser.add_argument('--cache-images', action='store_true',
                        help='cache images for faster training')
    parser.add_argument('--image-weights', action='store_true',
                        help='use weighted image selection for training')
    parser.add_argument('--device', default='',
                        help='cuda device, i.e. 0 or 0,1,2,3 or cpu')
    parser.add_argument('--multi-scale', action='store_true',
                        help='vary img-size +/- 50%%')
    parser.add_argument('--single-cls', action='store_true',
                        help='train multi-class data as single-class')
    parser.add_argument('--adam', action='store_true',
                        help='use torch.optim.Adam() optimizer')
    parser.add_argument('--sync-bn', action='store_true',
                        help='use SyncBatchNorm, only available in DDP mode')
    parser.add_argument('--local_rank', type=int, default=-1,
                        help='DDP parameter, do not modify')
    parser.add_argument('--workers', type=int, default=0,
                        help='maximum number of dataloader workers')
    parser.add_argument('--project', default='runs/train',
                        help='save to project/name')
    parser.add_argument('--entity', default=None, help='W&B entity')
    parser.add_argument('--name', default='exp', help='save to project/name')
    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')
    parser.add_argument('--linear-lr', action='store_true', help='linear LR')
    parser.add_argument('--label-smoothing', type=float,
                        default=0.0, help='Label smoothing epsilon')
    parser.add_argument('--upload_dataset', action='store_true',
                        help='Upload dataset as W&B artifact table')
    parser.add_argument('--bbox_interval', type=int, default=-1,
                        help='Set bounding-box image logging interval for W&B')
    parser.add_argument('--save_period', type=int, default=-1,
                        help='Log model after every "save_period" epoch')
    parser.add_argument('--artifact_alias', type=str, default="latest",
                        help='version of dataset artifact to be used')
    opt = parser.parse_args()

    # Set DDP variables
    opt.world_size = int(os.environ['WORLD_SIZE']
                         ) if 'WORLD_SIZE' in os.environ else 1
    opt.global_rank = int(os.environ['RANK']) if 'RANK' in os.environ else -1
    set_logging(opt.global_rank)
    if opt.global_rank in [-1, 0]:
        # check_git_status()
        check_requirements(exclude=('pycocotools', 'thop'))

    # Resume
    wandb_run = check_wandb_resume(opt)
    if opt.resume and not wandb_run:  # resume an interrupted run
        # specified or most recent path
        ckpt = opt.resume if isinstance(opt.resume, str) else get_latest_run()
        assert os.path.isfile(
            ckpt), 'ERROR: --resume checkpoint does not exist'
        apriori = opt.global_rank, opt.local_rank
        with open(Path(ckpt).parent.parent / 'opt.yaml') as f:
            opt = argparse.Namespace(**yaml.safe_load(f))  # replace
        opt.cfg, opt.weights, opt.resume, opt.batch_size, opt.global_rank, opt.local_rank = \
            '', ckpt, True, opt.total_batch_size, *apriori  # reinstate
        logger.info('Resuming training from %s' % ckpt)
    else:
        opt.data, opt.cfg, opt.hyp = check_file(opt.data), check_file(
            opt.cfg), check_file(opt.hyp)  # check files
        assert len(opt.cfg) or len(
            opt.weights), 'either --cfg or --weights must be specified'
        # extend to 2 sizes (train, test)
        opt.img_size.extend([opt.img_size[-1]] * (2 - len(opt.img_size)))
        opt.name = 'evolve' if opt.evolve else opt.name
        opt.save_dir = str(increment_path(
            Path(opt.project) / opt.name, exist_ok=opt.exist_ok | opt.evolve))

    # DDP mode
    opt.total_batch_size = opt.batch_size
    device = select_device(opt.device, batch_size=opt.batch_size)
    if opt.local_rank != -1:
        assert torch.cuda.device_count() > opt.local_rank
        torch.cuda.set_device(opt.local_rank)
        device = torch.device('cuda', opt.local_rank)
        # distributed backend
        dist.init_process_group(backend='nccl', init_method='env://')
        assert opt.batch_size % opt.world_size == 0, '--batch-size must be multiple of CUDA device count'
        opt.batch_size = opt.total_batch_size // opt.world_size

    # Hyperparameters
    with open(opt.hyp) as f:
        hyp = yaml.safe_load(f)  # load hyps

    # Train
    logger.info(opt)
    if not opt.evolve:
        tb_writer = None  # init loggers
        if opt.global_rank in [-1, 0]:
            prefix = colorstr('tensorboard: ')
            logger.info(
                f"{prefix}Start with 'tensorboard --logdir {opt.project}', view at http://localhost:6006/")
            tb_writer = SummaryWriter(opt.save_dir)  # Tensorboard
        train(hyp, opt, device, tb_writer)
    # Evolve hyperparameters (optional)
    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)
                # final OneCycleLR learning rate (lr0 * lrf)
                'lrf': (1, 0.01, 1.0),
                '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 per output grid (0 to ignore)
                'anchors': (2, 2.0, 10.0),
                # focal loss gamma (efficientDet default gamma=1.5)
                'fl_gamma': (0, 0.0, 2.0),
                # image HSV-Hue augmentation (fraction)
                'hsv_h': (1, 0.0, 0.1),
                # image HSV-Saturation augmentation (fraction)
                'hsv_s': (1, 0.0, 0.9),
                # image HSV-Value augmentation (fraction)
                'hsv_v': (1, 0.0, 0.9),
                '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)
                # image perspective (+/- fraction), range 0-0.001
                'perspective': (0, 0.0, 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)

        assert opt.local_rank == -1, 'DDP mode not implemented for --evolve'
        opt.notest, opt.nosave = True, True  # only test/save final epoch
        # ei = [isinstance(x, (int, float)) for x in hyp.values()]  # evolvable indices
        yaml_file = Path(opt.save_dir) / \
            'hyp_evolved.yaml'  # save best result here
        if opt.bucket:
            os.system('gsutil cp gs://%s/evolve.txt .' %
                      opt.bucket)  # download evolve.txt if exists

        for _ in range(300):  # generations to evolve
            if Path('evolve.txt').exists():  # if evolve.txt exists: select best hyps and mutate
                # Select parent(s)
                parent = 'single'  # parent selection method: 'single' or 'weighted'
                x = np.loadtxt('evolve.txt', ndmin=2)
                n = min(5, len(x))  # number of previous results to consider
                x = x[np.argsort(-fitness(x))][:n]  # top n mutations
                w = fitness(x) - fitness(x).min()  # weights
                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
                mp, s = 0.8, 0.2  # mutation probability, sigma
                npr = np.random
                npr.seed(int(time.time()))
                g = np.array([x[0] for x in meta.values()])  # gains 0-1
                ng = len(meta)
                v = np.ones(ng)
                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)
                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

            # Train mutation
            results = train(hyp.copy(), opt, device)

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

        # Plot results
        plot_evolution(yaml_file)
        print(f'Hyperparameter evolution complete. Best results saved as: {yaml_file}\n'
              f'Command to train a new model with these hyperparameters: $ python train.py --hyp {yaml_file}')