Accelerating LLM Inference on AMD ROCm with AITER and ATOM

For the next about 20 minutes, I'll talk about our inferencing framework called ATOM. And what's behind that is a collection of the kernels, which is called AITER. I think earlier today, when you hear about the keynote speech, you might realize that AMD has delivered a lot of the very good hardware.But to light up those GPUs, we actually need a very mature and sophisticated software stacks to make sure that we could leverage whatever that they offer by our hardware. So that's exactly what I'm going to prese ...