Training is slow on the second step of each epoch

[Busterfake]

Search before asking

• I have searched the YOLOv5 issues and discussions and found no similar questions.

Question

Hello, i rencently made a ml rig to improve my training time. This allowed me to train my datasets (around 150k images) 10 times faster than before. I currently train with batch size = 1000 and each epoch is around 24 sec for the first step (against ~7min before) and the second step is around 1min30.
From my experience, the second step was always faster than the first one so I wonder why it is now almost 4 times longer

For the second step, the training time didn't improve that much and i don't really understand why. What does the second step does exactly ? Do you have any idea how I can improve the speed of the training.
I also have the feeling that i am CPU limited and that all the cores are not used during the training process. Is there a solution for that too ?

Additional

No response

[github-actions]

👋 Hello @Busterfake, thank you for your interest in YOLOv5 🚀! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

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.

Requirements

Python>=3.8.0 with all requirements.txt installed including PyTorch>=1.8. To get started:

git clone https://github.com/ultralytics/yolov5  # clone
cd yolov5
pip install -r requirements.txt  # install

Environments

YOLOv5 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 YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training, validation, inference, export and benchmarks on macOS, Windows, and Ubuntu every 24 hours and on every commit.

Introducing YOLOv8 🚀

We're excited to announce the launch of our latest state-of-the-art (SOTA) object detection model for 2023 - YOLOv8 🚀!

Designed to be fast, accurate, and easy to use, YOLOv8 is an ideal choice for a wide range of object detection, image segmentation and image classification tasks. With YOLOv8, you'll be able to quickly and accurately detect objects in real-time, streamline your workflows, and achieve new levels of accuracy in your projects.

Check out our YOLOv8 Docs for details and get started with:

pip install ultralytics

[glenn-jocher]

@Busterfake hello! Thanks for reaching out. It's great to hear that you've significantly improved your training times with your new setup. 🚀

The difference in duration between the first and second steps of each epoch could be due to several factors. The first step usually involves forward and backward passes through the network, while the second step often includes validation, logging, and possibly saving checkpoints. If the second step is taking longer, it might be due to the validation dataset size or complexity, I/O operations, or additional computations performed during this phase.

To address your concerns:

1. Validation Time: If your validation set is large, it could be slowing down the process. Consider reducing the size of the validation set for faster epochs.
2. I/O Bottlenecks: Ensure your data is on a fast SSD and that you're not experiencing any I/O bottlenecks.
3. CPU Utilization: YOLOv5 automatically uses all available CPU cores for data augmentation. However, if you suspect that the CPU is a bottleneck, you can monitor its usage during training. If not all cores are at high utilization, you might want to check if other system processes are affecting performance.
4. Hyperparameters: Experiment with different batch sizes and worker numbers (--workers) to optimize CPU and GPU usage.

For more detailed guidance on training and potential performance improvements, please refer to our documentation.

Keep in mind that training times can vary based on many factors, and finding the optimal setup can require some experimentation. If you continue to experience issues, please provide more details about your system configuration and training setup, and we'll do our best to assist you.

Happy training! 😊

[github-actions]

👋 Hello there! We wanted to give you a friendly reminder that this issue has not had any recent activity and may be closed soon, but don't worry - you can always reopen it if needed. If you still have any questions or concerns, please feel free to let us know how we can help.

For additional resources and information, please see the links below:

• Docs: https://docs.ultralytics.com
• HUB: https://hub.ultralytics.com
• Community: https://community.ultralytics.com

Feel free to inform us of any other issues you discover or feature requests that come to mind in the future. Pull Requests (PRs) are also always welcomed!

Thank you for your contributions to YOLO 🚀 and Vision AI ⭐