How to set up Vulkan for GPU-accelerated inference with ailia SDK on Windows and Linux.
Vulkan is a cross-platform graphics and compute API developed by the Khronos Group. ailia SDK and ailia LLM can use Vulkan for GPU-accelerated inference on both NVIDIA and AMD GPUs. Unlike CUDA, Vulkan is not limited to NVIDIA hardware.
On Windows, the Vulkan runtime is included with GPU drivers from NVIDIA, AMD, and Intel. No additional installation is required.
If you have an up-to-date GPU driver installed, Vulkan is already available on your system.
Use an ailia-models sample to check that Vulkan appears in the environment list:
cd ailia-models/image_classification/resnet50
python3 resnet50.py --env_list
If Vulkan is available, you will see an environment entry like:
env[3]=Environment(id=3, type='GPU',
name='Vulkan-NVIDIA GeForce RTX 3080',
backend='Vulkan', props=[])
On Linux, the Vulkan loader library needs to be installed separately. Install it with:
sudo apt install libvulkan1
This installs the Vulkan ICD (Installable Client Driver) loader. You also need a GPU driver that includes Vulkan support.
If you are using an NVIDIA GPU, install the proprietary NVIDIA driver:
sudo apt install nvidia-driver-560
560 with the latest available driver version for your GPU. You can check available versions with apt search nvidia-driver.
If you are using an AMD GPU, install the Mesa Vulkan driver:
sudo apt install mesa-vulkan-drivers
You can verify the Vulkan installation using vulkaninfo:
sudo apt install vulkan-tools
vulkaninfo --summary
You can also verify with ailia SDK:
cd ailia-models/image_classification/resnet50
python3 resnet50.py --env_list
If Vulkan is available, you will see an environment entry with backend='Vulkan'.