YOLOv9 Hyperparameter Tuning & Layer Freezing

💡You can run the below code in Kaggle by using the copy and edit buttons. Kaggle Notebook Link

Notebook to train the Yolov9 model on the Yolo Format Dataset

• The notebook uses the model Yolov9 from Github at the Yolov9
• The notebook will train the Yolov9 model on the PinotNoirGrapes dataset available in kaggle.
• Wandb is used for tracking the training process.
• The weights are downloaded from the Yolov9

In the notebook

• First an initial training is performed
• Then the model is trained with different optimizers
• Also freezing of layers is performed.
• Then the model is again trained with the best optimizer
• Then Hyperparameter tuning is done without freezing layers and selecting the default optimizer.
• Same is also done for freezing layers upto 28.
• Finally the model is trained with the best hyperparameters after hyperparameter tuning for 100 epochs with resuming training from previous checkpoints.
• Also the model is trained with the best hyperparameters with the freezing layers upto 100 epochs
• The best model is thereafter evaluated on the test set.

If the evolve (hyperparameter tuning doesn't work) then check the Appendix Section to modify your train.py file as instructed

Feel free to reach out to me if you have any questions or suggestions.

Importing the libraries

#Importing the necessary libraries
import torch
import numpy as np
import random
import matplotlib.pyplot as plt
import os
import shutil
import wandb
from IPython.display import Image

Wandb initialization

#Setting the wandb key
#If you dont want to use wandb then uncomment the below code and comment out the remaining code
# os.environ["WANDB_API_KEY"] = ""
wandb.login(key='token_here')

#Dont run this code unless you are sure that you made a mistake in copying wrong dataset or made wrong file structure
!rm -r /kaggle/working/*

Model Download and Initialization

#Cloning the Yolov9 model
!git clone https://github.com/WongKinYiu/yolov9.git

#Setting the working directory and installing the requirements
%cd yolov9
HOME = os.getcwd()
print(HOME)
!pip install -r requirements.txt

Downloading the weights of the model

#Downloading the weights of model
!wget -P {HOME}/weights -q https://github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9-c.pt
!wget -P {HOME}/weights -q https://github.com/WongKinYiu/yolov9/releases/download/v0.1/yolov9-e.pt
!wget -P {HOME}/weights -q https://github.com/WongKinYiu/yolov9/releases/download/v0.1/gelan-c.pt
!wget -P {HOME}/weights -q https://github.com/WongKinYiu/yolov9/releases/download/v0.1/gelan-e.pt

#Setting the device
!nvidia-smi
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print('Using device:', device)
print()

#Copying the dataset from the input to the working
# Make sure the dataset.yaml is directly inside the datasets folder
!rm -r {HOME}/datasets
source_dir=('/kaggle/input/pinotnoirgrapes')
dest_dir=('/kaggle/working/yolov9/datasets')
shutil.copytree(source_dir, dest_dir)

# Run this code only if there is an issue with the wand initialization
!rm -r /kaggle/working/yolov9/wandb

Training the Model with the default setting

#Training the initial model without changing any default setting
!python train.py --img 512 --batch 16 --epochs 100 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-initial --cfg models/detect/gelan-e.yaml --hyp hyp.scratch-high.yaml

Training the Model with the freezing layers upto 28

#Training the model with the freezing layers upto head for the gelan-e model 
!python train.py --img 512 --batch 16 --epochs 100 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --freeze 28 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-freeze --cfg models/detect/gelan-e.yaml --hyp hyp.scratch-high.yaml --exist-ok

Training the Gelan-e model freezing upto head for layers 28 and different optimizers

#Training the model with the freezing layers upto head for the gelan-e model and using different optimizers like Adam, AdamW.
!python train.py --img 512 --batch 16 --epochs 100 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --freeze 24 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-freeze-optimizer-Adam --cfg models/detect/gelan-e.yaml --hyp hyp.scratch-high.yaml --exist-ok --optimizer Adam

# Fitness function used for the Genetic Optimization Algorithm
def fitness(x):
    # Model fitness as a weighted combination of metrics
    w = [0.0, 0.0, 0.1, 0.9]  # weights for [P, R, mAP@0.5, mAP@0.5:0.95]
    return (x[:, :4] * w).sum(1)

# Used to remove the evolve directory for faulty training configurations(Don't run unless otherwise you are sure that you made a mistake in training code or hyperparameters)
#!rm -r {HOME}/runs/evolve

If you encounter an error dfl not found then please check the Appendix section to modify your train.py file as instructed.

Hyperparameter Tuning

# Hyperparameter tuning without freezing layers upto 20 generations
# You can also change the number of epochs if you want good results
#If you want to tune hyperparameters for more than 20 generations then set --evolve greater than 20
# More evolve takes more time, evolve function is time consuming. Make sure you have enough understanding about the hyperparameters range to effectively do the hyperparameter tuning.
!python train.py --img 512 --batch 16 --epochs 10 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0  --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-evolve --cfg models/detect/gelan-e.yaml --hyp hyp.scratch-high.yaml --exist-ok  --evolve 20 --resume 

# Plot of the Hyperparameter tuning runs
Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve/results.png", width=800)

# Hyperparameter ranges with the best values after hyperparameter tuning
Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve/evolve.png", width=800)

Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve/labels_correlogram.jpg", width=800)

Training the model with the best hyperparameters

#Training the model with the best hyperparameters after hyperparameter tuning
!python train.py --img 512 --batch 16 --epochs 100 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-initial --cfg models/detect/gelan-e.yaml --hyp {HOME}/runs/evolve/gelan-e-evolve/hyp_evolve.yaml  --close-mosaic 15 --name best-train-hyp-model

#Plotting the confusion matrix for the best model
Image(filename=f"{HOME}/runs/train/best-train-hyp-model2/confusion_matrix.png", width=800)

#Training Instances
Image(filename=f"{HOME}/runs/train/best-train-hyp-model2/labels.jpg", width=800)

#Training and the Validation loss
Image(filename=f"{HOME}/runs/train/best-train-hyp-model2/results.png", width=800)

Hyperparameter Tuning with Freezing layers upto 28(for Gelan-e model)

# Hyperparameter tuning with freezing layers 28 upto 20 generations
!python train.py --img 512 --batch 16 --epochs 10 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --freeze 28 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml --name gelan-e-evolve-freeze-28 --cfg models/detect/gelan-e.yaml --hyp hyp.scratch-high.yaml --exist-ok  --evolve 20 --resume 

# Plot of the Hyperparameter tuning runs
Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve-freeze-28/results.png", width=800)

# Plot of the Hyperparameter tuning runs
Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve-freeze-28/evolve.png", width=800)

# Plot of the Hyperparameter tuning runs
Image(filename=f"{HOME}/runs/evolve/gelan-e-evolve-freeze-28/labels_correlogram.jpg", width=800)

#Training the model with the best hyperparameters after hyperparameter tuning with 28 freezing layers
!python train.py --img 512 --batch 32 --epochs 100 --freeze 28 --data {HOME}/datasets/data.yaml --weights {HOME}/weights/gelan-e.pt --cache --device 0 --close-mosaic 15 --cfg {HOME}/models/detect/gelan-e.yaml  --hyp {HOME}/runs/evolve/gelan-e-evolve-freeze-28/hyp_evolve.yaml  --close-mosaic 15 --name best-train-hyp-model-freeze-28

#Plotting the confusion matrix for the best model
Image(filename=f"{HOME}/runs/train/best-train-hyp-model-freeze-28/confusion_matrix.png", width=800)

#Training Instances
Image(filename=f"{HOME}/runs/train/best-train-hyp-model-freeze-28/labels.jpg", width=800)

#Training and the Validation loss
Image(filename=f"{HOME}/runs/train/best-train-hyp-model-freeze-28/results.png", width=800)

Testing the Best model Accuracy over the test set

# Validation using the best model
!python val.py --weights {HOME}/runs/train/best-train-hyp-model2/weights/best.pt --conf 0.1  --device 0 --save-txt --save-conf --exist-ok  --iou-thres 0.1 --imgsz 512"

Object Detection on the Images

# Detection using the best model
#Upload the images in the detection named folder in the kaggle/input/ the results will be generated in the kaggle/working/Detection folder

submission_images_dir=r'/kaggle/input/detection'

results_directory = r'/kaggle/working/Detection'
for filename in os.listdir(submission_images_dir):
  if os.path.isfile(os.path.join(submission_images_dir, filename)) and filename.lower().endswith('.jpg'):
    image_path = os.path.join(submission_images_dir, filename)
    print(f"Making a prediction on {filename}")

    # Call detect.py using subprocess module
    import subprocess

    # Construct the command (modify according to your detect.py arguments)
    command = f"python detect.py --source {image_path} --weights {HOME}/runs/train/best-train-hyp-model2/weights/best.pt --conf 0.1  --device 0 --save-txt --save-conf --exist-ok  --iou-thres 0.1 --imgsz 512 --project {results_directory} "
    subprocess.run(command, shell=True)

    # You can potentially access the output files generated by detect.py here
    # (requires knowledge of the output format and directory structure)

    print("Output files generated (if applicable).")

# Verifying the results
Image(filename=f"/kaggle/working/Detection/exp/Image name.jpg", width=600)

# Verifying the results
Image(filename=f"/kaggle/working/Detection/exp/Image name.jpg", width=600)

Appendix

To correctly run the evolve function the df1 loss range need to be added in the hyperparameter evolve of the train.py or train_dual.py file as like below

• 'iou_t': (0, 0.1, 0.7), # IoU training threshold
• 'dfl': (0,0.1,2.0), # Distribution Focal Loss
• 'anchor_t': (1, 2.0, 8.0)

In order to make the evolve function correctly working otherwise it will throw an error as dfl not found in hyp(opt) To do this follow the steps below

1. use %load then provide the path of the train.py file or the train_dual.py file
2. Make the modifications in the train.py or train_dual.py file as stated above
3. Replace %load with %%writefile and provide the path of the train.py file or the train_dual.py file in the same cell.

#%%writefile /kaggle/working/yolov9/train.py
import argparse
import math
import os
import random
import sys
import time
from copy import deepcopy
from datetime import datetime
from pathlib import Path

import numpy as np
import torch
import torch.distributed as dist
import torch.nn as nn
import yaml
from torch.optim import lr_scheduler
from tqdm import tqdm

FILE = Path(__file__).resolve()
ROOT = FILE.parents[0]  # root directory
if str(ROOT) not in sys.path:
    sys.path.append(str(ROOT))  # add ROOT to PATH
ROOT = Path(os.path.relpath(ROOT, Path.cwd()))  # relative

import val as validate  # for end-of-epoch mAP
from models.experimental import attempt_load
from models.yolo import Model
from utils.autoanchor import check_anchors
from utils.autobatch import check_train_batch_size
from utils.callbacks import Callbacks
from utils.dataloaders import create_dataloader
from utils.downloads import attempt_download, is_url
from utils.general import (LOGGER, TQDM_BAR_FORMAT, check_amp, check_dataset, check_file, check_img_size,
                           check_suffix, check_yaml, colorstr, get_latest_run, increment_path, init_seeds,
                           intersect_dicts, labels_to_class_weights, labels_to_image_weights, methods,
                           one_cycle, one_flat_cycle, print_args, print_mutation, strip_optimizer, yaml_save)
from utils.loggers import Loggers
from utils.loggers.comet.comet_utils import check_comet_resume
from utils.loss_tal import ComputeLoss
from utils.metrics import fitness
from utils.plots import plot_evolve
from utils.torch_utils import (EarlyStopping, ModelEMA, de_parallel, select_device, smart_DDP,
                               smart_optimizer, smart_resume, torch_distributed_zero_first)

LOCAL_RANK = int(os.getenv('LOCAL_RANK', -1))  # https://pytorch.org/docs/stable/elastic/run.html
RANK = int(os.getenv('RANK', -1))
WORLD_SIZE = int(os.getenv('WORLD_SIZE', 1))
GIT_INFO = None


def train(hyp, opt, device, callbacks):  # hyp is path/to/hyp.yaml or hyp dictionary
    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
    callbacks.run('on_pretrain_routine_start')

    # Directories
    w = save_dir / 'weights'  # weights dir
    (w.parent if evolve else w).mkdir(parents=True, exist_ok=True)  # make dir
    last, best = w / 'last.pt', w / 'best.pt'
    last_striped, best_striped = w / 'last_striped.pt', w / 'best_striped.pt'

    # Hyperparameters
    if isinstance(hyp, str):
        with open(hyp, errors='ignore') as f:
            hyp = yaml.safe_load(f)  # load hyps dict
    LOGGER.info(colorstr('hyperparameters: ') + ', '.join(f'{k}={v}' for k, v in hyp.items()))
    hyp['anchor_t'] = 5.0
    opt.hyp = hyp.copy()  # for saving hyps to checkpoints

    # Save run settings
    if not evolve:
        yaml_save(save_dir / 'hyp.yaml', hyp)
        yaml_save(save_dir / 'opt.yaml', vars(opt))

    # Loggers
    data_dict = None
    if RANK in {-1, 0}:
        loggers = Loggers(save_dir, weights, opt, hyp, LOGGER)  # loggers instance

        # Register actions
        for k in methods(loggers):
            callbacks.register_action(k, callback=getattr(loggers, k))

        # Process custom dataset artifact link
        data_dict = loggers.remote_dataset
        if resume:  # If resuming runs from remote artifact
            weights, epochs, hyp, batch_size = opt.weights, opt.epochs, opt.hyp, opt.batch_size

    # Config
    plots = not evolve and not opt.noplots  # create plots
    cuda = device.type != 'cpu'
    init_seeds(opt.seed + 1 + RANK, deterministic=True)
    with torch_distributed_zero_first(LOCAL_RANK):
        data_dict = data_dict or check_dataset(data)  # check if None
    train_path, val_path = data_dict['train'], data_dict['val']
    nc = 1 if single_cls else int(data_dict['nc'])  # number of classes
    names = {0: 'item'} if single_cls and len(data_dict['names']) != 1 else data_dict['names']  # class names
    #is_coco = isinstance(val_path, str) and val_path.endswith('coco/val2017.txt')  # COCO dataset
    is_coco = isinstance(val_path, str) and val_path.endswith('val2017.txt')  # COCO dataset

    # Model
    check_suffix(weights, '.pt')  # check weights
    pretrained = weights.endswith('.pt')
    if pretrained:
        with torch_distributed_zero_first(LOCAL_RANK):
            weights = attempt_download(weights)  # download if not found locally
        ckpt = torch.load(weights, map_location='cpu')  # load checkpoint to CPU to avoid CUDA memory leak
        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
        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:
        model = Model(cfg, ch=3, nc=nc, anchors=hyp.get('anchors')).to(device)  # create
    amp = check_amp(model)  # check AMP

    # Freeze
    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 TODO: uncomment this line as in master
        # v.register_hook(lambda x: torch.nan_to_num(x))  # NaN to 0 (commented for erratic training results)
        if any(x in k for x in freeze):
            LOGGER.info(f'freezing {k}')
            v.requires_grad = False

    # Image size
    gs = max(int(model.stride.max()), 32)  # grid size (max stride)
    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, amp)
        loggers.on_params_update({"batch_size": batch_size})

    # Optimizer
    nbs = 64  # nominal batch size
    accumulate = max(round(nbs / batch_size), 1)  # accumulate loss before optimizing
    hyp['weight_decay'] *= batch_size * accumulate / nbs  # scale weight_decay
    optimizer = smart_optimizer(model, opt.optimizer, hyp['lr0'], hyp['momentum'], hyp['weight_decay'])

    # Scheduler
    if opt.cos_lr:
        lf = one_cycle(1, hyp['lrf'], epochs)  # cosine 1->hyp['lrf']
    elif opt.flat_cos_lr:
        lf = one_flat_cycle(1, hyp['lrf'], epochs)  # flat cosine 1->hyp['lrf']        
    elif opt.fixed_lr:
        lf = lambda x: 1.0
    else:
        lf = lambda x: (1 - x / epochs) * (1.0 - hyp['lrf']) + hyp['lrf']  # linear

    scheduler = lr_scheduler.LambdaLR(optimizer, lr_lambda=lf)
    # from utils.plots import plot_lr_scheduler; plot_lr_scheduler(optimizer, scheduler, epochs)

    # EMA
    ema = ModelEMA(model) if RANK in {-1, 0} else None

    # Resume
    best_fitness, start_epoch = 0.0, 0
    if pretrained:
        if resume:
            best_fitness, start_epoch, epochs = smart_resume(ckpt, optimizer, ema, weights, epochs, resume)
        del ckpt, csd

    # DP mode
    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.')
        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()')

    # Trainloader
    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,
                                              close_mosaic=opt.close_mosaic != 0,
                                              quad=opt.quad,
                                              prefix=colorstr('train: '),
                                              shuffle=True,
                                              min_items=opt.min_items)
    labels = np.concatenate(dataset.labels, 0)
    mlc = int(labels[:, 0].max())  # max label class
    assert mlc < nc, f'Label class {mlc} exceeds nc={nc} in {data}. Possible class labels are 0-{nc - 1}'

    # Process 0
    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]

        if not resume:
            # if not opt.noautoanchor:
            #     check_anchors(dataset, model=model, thr=hyp['anchor_t'], imgsz=imgsz)  # run AutoAnchor
            model.half().float()  # pre-reduce anchor precision

        callbacks.run('on_pretrain_routine_end', labels, names)

    # DDP mode
    if cuda and RANK != -1:
        model = smart_DDP(model)

    # 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
    hyp['label_smoothing'] = opt.label_smoothing
    model.nc = nc  # attach number of classes to model
    model.hyp = hyp  # attach hyperparameters to model
    model.class_weights = labels_to_class_weights(dataset.labels, nc).to(device) * nc  # attach class weights
    model.names = names

    # Start training
    t0 = time.time()
    nb = len(train_loader)  # number of batches
    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
    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)
    scheduler.last_epoch = start_epoch - 1  # do not move
    scaler = torch.cuda.amp.GradScaler(enabled=amp)
    stopper, stop = EarlyStopping(patience=opt.patience), False
    compute_loss = ComputeLoss(model)  # init loss class
    callbacks.run('on_train_start')
    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...')
    for epoch in range(start_epoch, epochs):  # epoch ------------------------------------------------------------------
        callbacks.run('on_train_epoch_start')
        model.train()

        # Update image weights (optional, single-GPU only)
        if opt.image_weights:
            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
        if epoch == (epochs - opt.close_mosaic):
            LOGGER.info("Closing dataloader mosaic")
            dataset.mosaic = False

        # 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' + '%11s' * 7) % ('Epoch', 'GPU_mem', 'box_loss', 'cls_loss', 'dfl_loss', 'Instances', 'Size'))
        if RANK in {-1, 0}:
            pbar = tqdm(pbar, total=nb, bar_format=TQDM_BAR_FORMAT)  # progress bar
        optimizer.zero_grad()
        for i, (imgs, targets, paths, _) in pbar:  # batch -------------------------------------------------------------
            callbacks.run('on_train_batch_start')
            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

            # 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 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 == 0 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:
                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
            with torch.cuda.amp.autocast(amp):
                pred = model(imgs)  # forward
                loss, loss_items = compute_loss(pred, targets.to(device))  # loss scaled by batch_size
                if RANK != -1:
                    loss *= WORLD_SIZE  # gradient averaged between devices in DDP mode
                if opt.quad:
                    loss *= 4.

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

            # Optimize - https://pytorch.org/docs/master/notes/amp_examples.html
            if ni - last_opt_step >= accumulate:
                scaler.unscale_(optimizer)  # unscale gradients
                torch.nn.utils.clip_grad_norm_(model.parameters(), max_norm=10.0)  # clip gradients
                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(('%11s' * 2 + '%11.4g' * 5) %
                                     (f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]))
                callbacks.run('on_train_batch_end', model, ni, imgs, targets, paths, list(mloss))
                if callbacks.stop_training:
                    return
            # end batch ------------------------------------------------------------------------------------------------

        # Scheduler
        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)
            ema.update_attr(model, include=['yaml', 'nc', 'hyp', 'names', 'stride', 'class_weights'])
            final_epoch = (epoch + 1 == epochs) or stopper.possible_stop
            if not noval or final_epoch:  # Calculate mAP
                results, maps, _ = validate.run(data_dict,
                                                batch_size=batch_size // WORLD_SIZE * 2,
                                                imgsz=imgsz,
                                                half=amp,
                                                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
            fi = fitness(np.array(results).reshape(1, -1))  # weighted combination of [P, R, mAP@.5, mAP@.5-.95]
            stop = stopper(epoch=epoch, fitness=fi)  # early stop check
            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
            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(),
                    'opt': vars(opt),
                    'git': GIT_INFO,  # {remote, branch, commit} if a git repo
                    'date': datetime.now().isoformat()}

                # Save last, best and delete
                torch.save(ckpt, last)
                if best_fitness == fi:
                    torch.save(ckpt, best)
                if 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)

        # EarlyStopping
        if RANK != -1:  # if DDP training
            broadcast_list = [stop if RANK == 0 else None]
            dist.broadcast_object_list(broadcast_list, 0)  # broadcast 'stop' to all ranks
            if RANK != 0:
                stop = broadcast_list[0]
        if stop:
            break  # must break all DDP ranks

        # end epoch ----------------------------------------------------------------------------------------------------
    # end training -----------------------------------------------------------------------------------------------------
    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():
                if f is last:
                    strip_optimizer(f, last_striped)  # strip optimizers
                else:
                    strip_optimizer(f, best_striped)  # strip optimizers
                if f is best:
                    LOGGER.info(f'\nValidating {f}...')
                    results, _, _ = validate.run(
                        data_dict,
                        batch_size=batch_size // WORLD_SIZE * 2,
                        imgsz=imgsz,
                        model=attempt_load(f, device).half(),
                        single_cls=single_cls,
                        dataloader=val_loader,
                        save_dir=save_dir,
                        save_json=is_coco,
                        verbose=True,
                        plots=plots,
                        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)

        callbacks.run('on_train_end', last, best, epoch, results)

    torch.cuda.empty_cache()
    return results


def parse_opt(known=False):
    parser = argparse.ArgumentParser()
    # parser.add_argument('--weights', type=str, default=ROOT / 'yolo.pt', help='initial weights path')
    # parser.add_argument('--cfg', type=str, default='', help='model.yaml path')
    parser.add_argument('--weights', type=str, default='', help='initial weights path')
    parser.add_argument('--cfg', type=str, default='yolo.yaml', help='model.yaml path')
    parser.add_argument('--data', type=str, default=ROOT / 'data/coco128.yaml', help='dataset.yaml path')
    parser.add_argument('--hyp', type=str, default=ROOT / 'data/hyps/hyp.scratch-low.yaml', help='hyperparameters path')
    parser.add_argument('--epochs', type=int, default=100, help='total training epochs')
    parser.add_argument('--batch-size', type=int, default=16, help='total batch size for all GPUs, -1 for autobatch')
    parser.add_argument('--imgsz', '--img', '--img-size', type=int, default=640, help='train, val image size (pixels)')
    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('--noval', action='store_true', help='only validate final epoch')
    parser.add_argument('--noautoanchor', action='store_true', help='disable AutoAnchor')
    parser.add_argument('--noplots', action='store_true', help='save no plot files')
    parser.add_argument('--evolve', type=int, nargs='?', const=300, help='evolve hyperparameters for x generations')
    parser.add_argument('--bucket', type=str, default='', help='gsutil bucket')
    parser.add_argument('--cache', type=str, nargs='?', const='ram', help='image --cache ram/disk')
    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('--optimizer', type=str, choices=['SGD', 'Adam', 'AdamW', 'LION'], default='SGD', help='optimizer')
    parser.add_argument('--sync-bn', action='store_true', help='use SyncBatchNorm, only available in DDP mode')
    parser.add_argument('--workers', type=int, default=8, help='max dataloader workers (per RANK in DDP mode)')
    parser.add_argument('--project', default=ROOT / 'runs/train', help='save to project/name')
    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('--cos-lr', action='store_true', help='cosine LR scheduler')
    parser.add_argument('--flat-cos-lr', action='store_true', help='flat cosine LR scheduler')
    parser.add_argument('--fixed-lr', action='store_true', help='fixed LR scheduler')
    parser.add_argument('--label-smoothing', type=float, default=0.0, help='Label smoothing epsilon')
    parser.add_argument('--patience', type=int, default=100, help='EarlyStopping patience (epochs without improvement)')
    parser.add_argument('--freeze', nargs='+', type=int, default=[0], help='Freeze layers: backbone=10, first3=0 1 2')
    parser.add_argument('--save-period', type=int, default=-1, help='Save checkpoint every x epochs (disabled if < 1)')
    parser.add_argument('--seed', type=int, default=0, help='Global training seed')
    parser.add_argument('--local_rank', type=int, default=-1, help='Automatic DDP Multi-GPU argument, do not modify')
    parser.add_argument('--min-items', type=int, default=0, help='Experimental')
    parser.add_argument('--close-mosaic', type=int, default=0, help='Experimental')

    # Logger arguments
    parser.add_argument('--entity', default=None, help='Entity')
    parser.add_argument('--upload_dataset', nargs='?', const=True, default=False, help='Upload data, "val" option')
    parser.add_argument('--bbox_interval', type=int, default=-1, help='Set bounding-box image logging interval')
    parser.add_argument('--artifact_alias', type=str, default='latest', help='Version of dataset artifact to use')

    return parser.parse_known_args()[0] if known else parser.parse_args()


def main(opt, callbacks=Callbacks()):
    # Checks
    if RANK in {-1, 0}:
        print_args(vars(opt))

    # Resume (from specified or most recent last.pt)
    if opt.resume and not check_comet_resume(opt) and not opt.evolve:
        last = Path(check_file(opt.resume) if isinstance(opt.resume, str) else get_latest_run())
        opt_yaml = last.parent.parent / 'opt.yaml'  # train options yaml
        opt_data = opt.data  # original dataset
        if opt_yaml.is_file():
            with open(opt_yaml, errors='ignore') as f:
                d = yaml.safe_load(f)
        else:
            d = torch.load(last, map_location='cpu')['opt']
        opt = argparse.Namespace(**d)  # replace
        opt.cfg, opt.weights, opt.resume = '', str(last), True  # reinstate
        if is_url(opt_data):
            opt.data = check_file(opt_data)  # avoid HUB resume auth timeout
    else:
        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'
        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
        if opt.name == 'cfg':
            opt.name = Path(opt.cfg).stem  # use model.yaml as name
        opt.save_dir = str(increment_path(Path(opt.project) / opt.name, exist_ok=opt.exist_ok))

    # DDP mode
    device = select_device(opt.device, batch_size=opt.batch_size)
    if LOCAL_RANK != -1:
        msg = 'is not compatible with YOLO 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'
        torch.cuda.set_device(LOCAL_RANK)
        device = torch.device('cuda', LOCAL_RANK)
        dist.init_process_group(backend="nccl" if dist.is_nccl_available() else "gloo")

    # Train
    if not opt.evolve:
        train(opt.hyp, opt, device, callbacks)

    # 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)
            '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
            'dfl': (0,0.1,2.0),        # Distribution Focal Loss
            '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:
            hyp = yaml.safe_load(f)  # load hyps dict
            if 'anchors' not in hyp:  # anchors commented in hyp.yaml
                hyp['anchors'] = 3
        if opt.noautoanchor:
            del hyp['anchors'], meta['anchors']
        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 {evolve_csv}')  # 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)
                parent = 'single'  # parent selection method: 'single' or 'weighted'
                x = np.loadtxt(evolve_csv, ndmin=2, delimiter=',', skiprows=1)
                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() + 1E-6  # weights (sum > 0)
                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([meta[k][0] for k in hyp.keys()])  # 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, callbacks)
            callbacks = Callbacks()
            # Write mutation results
            keys = ('metrics/precision', 'metrics/recall', 'metrics/mAP_0.5', 'metrics/mAP_0.5:0.95', 'val/box_loss',
                    'val/obj_loss', 'val/cls_loss')
            print_mutation(keys, results, hyp.copy(), save_dir, opt.bucket)

        # Plot results
        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}')


def run(**kwargs):
    # Usage: import train; train.run(data='coco128.yaml', imgsz=320, weights='yolo.pt')
    opt = parse_opt(True)
    for k, v in kwargs.items():
        setattr(opt, k, v)
    main(opt)
    return opt


if __name__ == "__main__":
    opt = parse_opt()
    main(opt)