llama.cpp
ggml : add GGML_SCHED_NO_REALLOC option to disable reallocations in ggml_backend_sched
#17276
Merged
Merged