ggml-cuda: read sysfs VRAM usage on ROCm to fix over-reported free memory - #25123

ggml-cuda: read sysfs VRAM usage on ROCm to fix over-reported free memory#25123

liminfei-amd wants to merge 1 commit into

ggml-org:masterggml-org/llama.cpp:masterfrom

liminfei-amd:rocm-vram-sysfsliminfei-amd/llama.cpp:rocm-vram-sysfsCopy head branch name to clipboard