Skip to main content

quantized_matmul

mlx_rs::ops

Function quantized_matmul 

Source 
pub fn quantized_matmul<'a>(
    x: impl AsRef<Array>,
    w: impl AsRef<Array>,
    scales: impl AsRef<Array>,
    biases: impl Into<Option<&'a Array>>,
    options: QuantizedMatmulOptions,
) -> Result<Array, QuantizationError>
Expand description

Multiply x by packed weights. Pass biases for affine weights and None for floating-point modes; options must match the weights’ quantization.