Can I set the order in which the dataset is entered when I train the dataset?

[mark4653]

I understand that if you set the batch size to 16 for 64 images, the total images are learned by dividing them into 4 times, 16 each.
At this time, can I decide what image will be in each batch? (For example, when there are images from 1 to 64, I wonder if I can type from 17 to 32 in the first batch and from 1 to 16 in the second batch.)

If I have to modify the source because that feature is not supported, which part should I modify?

[UltralyticsAssistant]

👋 Hello @mark4653, thank you for your interest in Ultralytics 🚀! We recommend checking out our Docs for detailed information on custom training and dataset handling, as many questions like this may already be addressed there.

If this is a ❓ Question related to custom training or dataset order, please provide additional context if possible, including your dataset details, the framework version you are using, and any relevant code snippets to help us better understand your requirements. Additionally, ensure you’re following our Tips for Best Training Results.

If you aim to modify how batches are ordered, custom changes might be necessary in the source code. For reference, the DataLoader implementation typically handles batch formation, so you may need to explore areas like dataset.py or the related DataLoader configurations. Keep in mind that modifying source code may require some experience with Python and PyTorch.

For any 🐛 Bug Reports related to dataset loading or custom configurations, we kindly ask for a minimum reproducible example that demonstrates the behavior.

Join the Ultralytics community to discuss ideas or ask questions directly:

• For real-time help, join our Discord 🎧
• For in-depth discussions, check out our Discourse
• Or connect with others on our Subreddit

Upgrade

To ensure the latest features and bug fixes are applied, upgrade your environment to the latest ultralytics package with all requirements in a Python>=3.8 environment:

pip install -U ultralytics

Verified Environments

YOLO can be trained in the following verified environments. These come with all necessary dependencies pre-installed:

• 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 passing. CI tests verify correct operation of YOLO Modes and Tasks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

This is an automated response, but don’t worry, an Ultralytics engineer will assist you further soon 🚀!

[glenn-jocher]

@mark4653 to customize the batch order during training, you would need to modify the data loading logic. The relevant code is in the get_dataloader method where the DataLoader is constructed with shuffle parameter control. You can implement a custom sampler or adjust the dataset indices before batching. For specific implementation, we recommend reviewing PyTorch's DataLoader documentation on custom sampling strategies. This functionality isn't natively supported but can be achieved through code modifications.