cache the last data dir, the second time to train will be updated about ultralytics

[github-actions]

👋 Hello @figodeng, thank you for your interest in Ultralytics YOLOv8 🚀! We recommend a visit to the Docs for new users where you can find many Python and CLI usage examples and where many of the most common questions may already be answered.

If this is a 🐛 Bug Report, please provide a minimum reproducible example to help us debug it.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset image examples and training logs, and verify you are following our Tips for Best Training Results.

Join the vibrant Ultralytics Discord 🎧 community for real-time conversations and collaborations. This platform offers a perfect space to inquire, showcase your work, and connect with fellow Ultralytics users.

Install

Pip install the ultralytics package including all requirements in a Python>=3.8 environment with PyTorch>=1.8.

pip install ultralytics

Environments

YOLOv8 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Notebooks with free GPU:
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Amazon Deep Learning AMI. See AWS Quickstart Guide
• Docker Image. See Docker Quickstart Guide

Status

If this badge is green, all Ultralytics CI tests are currently passing. CI tests verify correct operation of all YOLOv8 Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

@figodeng hello!

Thank you for your detailed bug report and for providing a minimal reproducible example. This is very helpful! 😊

From your description, it seems like the issue is related to the caching mechanism of the dataset directory, where the first training run uses the old directory and only updates to the new directory on the second run.

To help us investigate further, could you please confirm the following:

1. Are you using the latest versions of torch and ultralytics? If not, please update to the latest versions and try running your code again.
2. Could you provide any additional logs or error messages that appear during the first and second runs?

In the meantime, you might want to try explicitly clearing the cache before starting a new training session. Here’s a small modification to your code that might help:

from ultralytics.models.yolo.detect import DetectionTrainer
from ultralytics.utils import yaml_load, IterableSimpleNamespace
from ultralytics import settings

def train(data_dir, save_dir, epochs=3):
    # Clear the cache
    settings.clear_cache()
    
    # Update settings
    settings.update({'datasets_dir': data_dir})
    
    # Update config
    args = dict(model='./config/pretrained_models/yolov8n.pt',
                data='./config/coco8.yaml',
                save_dir=save_dir,
                epochs=epochs)
    config = yaml_load('./config/default.yaml')
    trainer = DetectionTrainer(cfg=IterableSimpleNamespace(**config), overrides=args)
    print(settings)
    trainer.train()

train(data_dir='./datasets/dataset_coco', save_dir='./model')

This should ensure that the cache is cleared and the new dataset directory is used immediately.

Let me know if this helps or if you have any further questions!