- With Vulkan, the model now runs on the first dedicated GPU instead of whisper.cpp's default device 0, which can be integrated graphics. Integrated graphics are used only when there is no dedicated GPU.
- Processing device now shows what the loaded model runs on, for example
In use: Vulkan · NVIDIA GeForce RTX 4060 Ti, orIn use: CPU (Vulkan unavailable)when the Vulkan runtime cannot be used. The model load log records the GPU name.
Source commit: 2fda6a6