forked from third-party-mirrors/ollama
If we try to use mmap when the model is larger than the system free space, loading is slower than the no-mmap approach.
If we try to use mmap when the model is larger than the system free space, loading is slower than the no-mmap approach.