Module fusion

Module fusion 

Source
Expand description

Fusion of consecutive quantized operations

This module provides optimized implementations for fusing multiple quantized operations, avoiding the overhead of intermediate dequantization and requantization steps in performance-critical code paths.

Functionsยง

fused_quantized_matmul_chain
Fused quantized matrix multiplication chain
fused_quantized_matvec_sequence
Fused quantized matrix-vector multiplication sequence