Setup Runtime

Runtime configuration guide for GPU acceleration and optimal performance

This configuration step is optional. By default, Sapientia is designed to run immediately using the CPU without any additional configuration. However, to achieve maximum inference performance and faster response times, we highly recommend setting up the runtime environment to enable hardware acceleration (GPU).

If you skip this step, the application will still function normally using your processor (CPU) resources.


Windows

For Windows users who want to enable GPU acceleration, ensure that graphics drivers and supporting software are up to date.

GPU Configuration

1. NVIDIA (CUDA):

  1. For optimal acceleration, you will need the Driver and CUDA Toolkit.
    • Update Driver: Download the latest driver through NVIDIA GeForce Experience or NVIDIA Official Website.
    • Install CUDA Toolkit: Download and install CUDA Toolkit from NVIDIA Developer. This toolkit contains essential compute libraries required by the inference engine.

2. AMD / Intel (Vulkan):

Vulkan drivers are typically installed automatically with graphics card driver updates (Radeon Software or Intel Graphics Command Center). No additional SDK installation is required for general users.


Linux

Linux users need to manually install Vulkan libraries or configure NVIDIA settings.

1. Vulkan SDK (Required for AMD/Intel)

Some Linux distributions do not include Vulkan libraries by default. Vulkan SDK installation can be performed by visiting the following link:

https://vulkan.lunarg.com/sdk/home

2. NVIDIA Drivers & CUDA Toolkit

If using an NVIDIA GPU, you must use the proprietary driver and install the CUDA Toolkit.

Install CUDA Toolkit: Download and install CUDA Toolkit from NVIDIA Developer. This toolkit contains essential compute libraries required by the inference engine.

After installation is complete, restart your computer to apply changes and ensure the driver is loaded correctly.


macOS

For Apple ecosystem users, runtime configuration is much simpler thanks to tight hardware and software integration.

Apple Silicon

Sapientia leverages Metal API natively for hardware acceleration on Apple Silicon chips. Graphics acceleration (Metal) support on Apple Silicon Macs will be active by default without additional configuration.

Intel Mac

Graphics acceleration (Metal) support on Intel-based Macs has architectural limitations and is disabled by default. The application will automatically run in CPU mode.