burn_tensor::quantization

Type Alias QuantizationParameters

Source 
pub type QuantizationParameters<B> = QParams<Tensor<B, 1>, Tensor<B, 1, Int>>;
Expand description

The tensor quantization parameters.

Aliased Type§

struct QuantizationParameters<B> {
    pub scale: Tensor<B, 1>,
    pub offset: Option<Tensor<B, 1, Int>>,
}

Fields§

§scale: Tensor<B, 1>

The scaling factor.

§offset: Option<Tensor<B, 1, Int>>

The zero-point offset.