Aristo94@Aristo94 · Oct 2A llama.cpp fork that speeds ROCm prefill and fits multi-token prediction for GSQ quantizations on AMD gfx1151 hardware.
