Expand description
Backend capability declarations.
A capability record declares the exact model architectures, quantization formats, KV-cache encodings, maximum batch sizes, and graph acceleration modes supported by a given prepared backend.
Structs§
- Backend
Capabilities - Explicit capability declaration for a prepared backend.
Enums§
- KvEncoding
- Supported KV cache encoding formats.
- Supported
Architecture - Supported neural model architecture families.