burn_jit::kernel::quantization

Function quantize

Source 
pub fn quantize<R, F, I>(
    tensor: JitTensor<R>,
    scheme: &QuantizationScheme,
    scale: JitTensor<R>,
    offset: Option<JitTensor<R>>,
) -> JitTensor<R>
where
    R: JitRuntime,
    F: FloatElement,
    I: IntElement,
Expand description

Convert the tensor to a lower precision data type based on the quantization scheme and parameters.