What changed

Ultralytics YOLO, the popular open-source framework for real-time computer vision, now officially supports AMD GPUs. Developers can train and run YOLO26 models on AMD GPUs using PyTorch on ROCm™, and deploy exported ONNX models with MIGraphX, the AMD graph inference engine.

This integration has been tested on AMD Instinct™ GPUs, AMD Radeon™ GPUs, and AMD Ryzen™ AI processors, from data center servers to embedded devices.

What this means in practice

The integration allows developers to use AMD GPUs for training and inference tasks with the same Ultralytics API. The Ultralytics validation tests showed that MIGraphX ran YOLO26 models with up to 67% lower latency than PyTorch at the same accuracy.

Developers can choose between using a Docker image or installing via pip to set up the environment. The Docker image includes PyTorch on ROCm, Ultralytics, ONNX Runtime, and MIGraphX, while the pip installation requires Python 3.11 or newer and specific system configurations.

Limitations and availability

AMD GPU support is available on Linux x86_64. Windows is not supported, and WSL has not been tested.

The tested platforms include AMD Instinct MI350X, Radeon AI PRO R9700, Ryzen AI Max+ PRO 395, Ryzen AI Embedded X100, and Ryzen AI Embedded P185. A dash indicates that the workflow was not tested on that platform.

The tests used Ubuntu 24.04, the amdgpu driver included in the Linux kernel, and ROCm 10.0.0 from the PyTorch wheels.

Source facts for Ultralytics YOLO Now Supports AMD GPUs for Training and Deployment: Ultralytics YOLO now officially supports AMD GPUs. In this blog, you’ll learn to install Ultralytics on a system with an AMD GPU, train a YOLO26 model, run it, and deploy it with MIGraphX. For the full reference, see the Ultralytics AMD integration guide.
Original AINewsTrend explainer based on the cited source. This graphic summarizes source statements; it is not an independent benchmark.

Sources and further reading

  1. amd: Ultralytics YOLO on AMD GPUs: Train, Run, and Deploy with ROCm™