klyn.math.nn.quantization

Package klyn.math.nn.quantization

Contains 6 type(s).

Types

Classes

DeviceQuantizedMatrix - Resident CPU or NVIDIA weights, loaded in bounded blocks.
GpuQuantizedMatrix - NVIDIA-only specialization of DeviceQuantizedMatrix.
QuantizationOptions - Options for deterministic, row-local symmetric Q8 conversion.
QuantizationReport - Summary of a successfully published Q8 checkpoint; no tensor data is retained.
QuantizedMatrix - A packed Klyn Q8 matrix with row-local FP32 scales.
Quantizer - Bounded-memory, deterministic weight-only Q8 checkpoint conversion.