PF4Public@users.noreply.github.com Niko1221/Strata https://github.com/Niko1221/Strata/issues https://github.com/Niko1221/Strata/tree/main/docs Build the engine for NVIDIA GPUs (CUDA backend, sm_75 through sm_120: RTX 20/30/40/50) Build the engine for AMD GPUs (ROCm/HIP backend, gfx1100 and gfx1201: RDNA3/RDNA4) Link the ggml installed by sci-ml/ggml instead of building the pinned llama.cpp commit's ggml; strata's own ggml-cuda MMQ sources are still compiled, from ggml's installed source tree