search

ultralytics / yolov5

YOLOv5 🚀 in PyTorch > ONNX > CoreML > TFLite
https://docs.ultralytics.com
GNU Affero General Public License v3.0
50.89k stars 16.39k forks source link

How to use a second GPU outside of the default GPU on yolov5? #13379

Open ijnrghjkdsmigywneig203 opened 3 weeks ago

ijnrghjkdsmigywneig203 commented 3 weeks ago

Search before asking

Question

I have tried doing torch.cuda.set_device and self.model.to(device), etc... None of it seems to work. It always defaults to main GPU.

Additional

No response

UltralyticsAssistant commented 3 weeks ago

👋 Hello @ijnrghjkdsmigywneig203, thank you for your interest in YOLOv5 🚀! This is an automated response, and an Ultralytics engineer will assist you soon.

If you're facing issues with utilizing a second GPU, please ensure you've set the appropriate device environment correctly. For detailed debugging, providing a minimum reproducible example would be extremely helpful. This will allow us to better understand your setup and identify any potential issues.

In the meantime, you might find it useful to review our documentation and tutorials for guidance on multi-GPU setups and other advanced configurations.

Requirements

Ensure you have Python version 3.8.0 or greater and all necessary packages installed according to the requirements.txt. Start by cloning the repository and installing dependencies if you haven't already.

Environments

YOLOv5 can be run in various verified environments, such as Google Colab, Kaggle, or through Docker containers, which come pre-installed with all necessary dependencies.

Status

Check the continuous integration (CI) status on our GitHub Actions page to verify that all tests are currently passing. This helps ensure that the codebase is functioning as expected across various features like training, validation, and inference.

Introducing YOLOv8 🚀

We're excited to share the launch of YOLOv8, our latest and state-of-the-art object detection model for 2023! YOLOv8 is designed to be fast, accurate, and easy to use. It supports a wide range of tasks including object detection, image segmentation, and image classification. To try it out, simply install the new package.

Feel free to provide additional context or examples to help us assist you more effectively. Happy experimenting! 😊

pderrenger commented 6 days ago

@ijnrghjkdsmigywneig203 to use a specific GPU with YOLOv5, you can specify the device using the --device argument in your command, like this: python train.py --device 1 to use the second GPU.