getResolvedGpuLayers abstract method

Future<int?> getResolvedGpuLayers()

Returns resolved GPU layers used for the active model load.

Meaning is backend-specific. For llama.cpp this is the final layer count after device selection and fallback, including 0 for forced CPU fallback; mainGpu and per-device layer allocation determine where offload occurs. LiteRT-LM reports 0 for CPU and the requested maximum for other backends, rather than a measured per-device allocation.

Implementation

Future<int?> getResolvedGpuLayers();