Expand description
Real transformer forward passes for text generation.
Re-exports§
pub use backend::mac_gpu_support;pub use backend::total_system_ram_bytes;pub use backend::LlmBackendKind;pub use backend::MacGpuSupport;
Modules§
- backend
- Backend dispatch for native LLM forward passes.
- common
- Shared tensor ops for transformer forward passes.
Structs§
- Llama
Forward - Real Llama-architecture forward engine backed by safetensors weights.
- PhiForward
Enums§
- Forward
Engine - Architecture-specific inference engine with KV-cache support.