CUDA: replace GGML_CUDA_F16 with CUDA arch checks#15433

JohannesGaessler merged 1 commit intoggml-org:masterggml-org/llama.cpp:masterfrom

JohannesGaessler:cuda-remove-f16JohannesGaessler/llama.cpp:cuda-remove-f16Copy head branch name to clipboard