Module quantized_matrixfree 

Matrix-free operations for quantized tensors

This module provides matrix-free operations for quantized tensors, enabling efficient memory usage and computation for large models. It combines the benefits of quantization (reduced memory footprint) with matrix-free operations (no need to materialize large matrices).

Structs

QuantizedMatrixFreeOp

A matrix-free operator that represents a quantized matrix

Functions

quantized_to_linear_operator

Convert a QuantizedMatrixFreeOp to a generic LinearOperator

Type Aliases

MatVecFn

Type alias for the matrix-vector product function