Remote GPU Cluster YOLOv10 Training All Sweeps Result in < null >

W&B Help
1

Hello, WandB Community!
I run YOLOv10 training script on a remote GPU cluster and want to finetune hyperparameters before running the main model training.
I run 20 sweeps and all of them result in < null > (see image below)

sweeps
sweeps1383×566 139 KB

However, in my yolov10_training/yolov10_finetune_LARD_13/02_19_52/ I can see results.csv which are not null, see the image below

epoch time train/box_loss train/cls_loss train/dfl_loss metrics/precision(B) metrics/recall(B) metrics/mAP50(B) metrics/mAP50-95(B) val/box_loss val/cls_loss val/dfl_loss lr/pg0 lr/pg1 lr/pg2
1 165.006 2.91836 10.6008 1.95599 0.48503 0.276 0.28385 0.14981 3.57982 6.09934 2.40669 0.000663743 0.000663743 0.000663743
2 340.683 2.55896 4.95659 1.93792 0.59701 0.44838 0.43939 0.24511 3.36526 4.47328 2.29902 0.00106699 0.00106699 0.00106699
3 516.641 2.36457 2.97833 1.85849 0.5841 0.48533 0.4661 0.26107 3.38215 4.35041 2.25128 0.00120623 0.00120623 0.00120623
4 689.54 2.12222 2.12358 1.8063 0.58889 0.37333 0.40226 0.24237 3.13558 3.75161 2.21274 0.000812 0.000812 0.000812
5 865.638 1.82564 1.71369 1.77575 0.66677 0.45888 0.48731 0.2987 3.04279 3.14339 2.14769 0.000416 0.000416 0.000416

this is my train_model.py:

import ultralytics
import wandb
from ultralytics import YOLO
import sys
import yaml
import datetime
import torch

if name == ‘main’:
sys.stdout.reconfigure(encoding=‘utf-8’)
ultralytics.checks()
wandb.login(key=“my_api_key_here”)

dataset = "LARD"
time = datetime.datetime.now().strftime("%d/%m_%H_%M")

sweep_configuration = {
    "method": "random",
    "name": "yolov10-sweep",
    "metric": {"name": "loss", "goal": "minimize"},
    "parameters": {
        "batch_size": {"values": [16, 32, 64]},
        "epochs": {"values": [5, 10, 15]},
        "lr": {"max": 0.1, "min": 0.0001},
    },
}

sweep_id = wandb.sweep(sweep=sweep_configuration, project="RLD-training")

def train_yolo():
    wandb.init(project="RLD-training", name=f"RLD_Train_{dataset}_{time}")
    
    config_file = "yolo/config/yolov10_config.yaml"
    config_data = {
        "train": "dataset/images/train",
        "val": "dataset/images/train",  
        "test": "dataset/images/test",
        "nc": 1,  
        "names": ["runway"] 
    }
    with open(config_file, "w") as f:
        yaml.dump(config_data, f)

    model = YOLO("yolo/weights/yolov10n.pt").to(torch.device("cuda"))
    model_path = "best_yolov10.pt"
    
    model.train(
        data="yolo/config/yolov10_config.yaml", 
        epochs=wandb.config.epochs,   
        batch=wandb.config.batch_size,    
        lr0=wandb.config.lr,        
        project="yolov10_training",
        name=f"yolov10_finetune_{dataset}_{time}",
    )

    model.save(model_path)
    wandb.log_model(model_path)

wandb.agent(sweep_id, function=train_yolo, count=10)
wandb.finish()

Please, help me fix the visualization of wandb sweeps.
Regards,
Yulian.