GPU/CUDA Support
#10
by
brunjo
- opened
Is it possible to run ten-vad on a NVIDIA GPU?
@brunjo For ONNX, it's possible. For prebuilt-libs, I think not. But the point is that the model itself is small enough and the model inference is in a streaming fashion. So the parallel performance of GPU would not help such a task, or even the inference could be slower, since the data transfer from CPU to GPU may spend more time than the potential speed gain using GPU.