|
liblloyal 1.0.0
Composable primitives for llama.cpp inference
|
Model cache key combining file path and GPU configuration. More...
#include <model_registry.hpp>
Public Member Functions | |
| bool | operator== (const ModelKey &o) const |
Public Attributes | |
| std::string | canonPath |
| Normalized file path (file:// prefix removed) | |
| int | n_gpu_layers |
| Number of layers offloaded to GPU (-1 = all) | |
| bool | use_mmap |
| Whether to use memory mapping. | |
Model cache key combining file path and GPU configuration.
Used as key in ModelRegistry cache to uniquely identify model instances. Different GPU configurations of the same model are cached separately.
Definition at line 35 of file model_registry.hpp.
|
inline |
Definition at line 40 of file model_registry.hpp.
| std::string lloyal::ModelKey::canonPath |
Normalized file path (file:// prefix removed)
Definition at line 36 of file model_registry.hpp.
| int lloyal::ModelKey::n_gpu_layers |
Number of layers offloaded to GPU (-1 = all)
Definition at line 37 of file model_registry.hpp.
| bool lloyal::ModelKey::use_mmap |
Whether to use memory mapping.
Definition at line 38 of file model_registry.hpp.