dequantize

aimet_torch.quantization.affine.dequantize(tensor, scale, offset, block_size=None)

Applies dequantization to the input.

Precisely,

\[out = (input + offset) * scale\]

If block size \(B = \begin{pmatrix} B_0 & B_1 & \cdots & B_{D-1} \end{pmatrix}\) is specified, this equation will be further generalized as

\[ \begin{align}\begin{aligned}out_{j_0 \cdots j_{D-1}} & = (input_{j_0 \cdots j_{D-1}} + offset_{i_0 \cdots i_{D-1}}) * scale_{i_0 \cdots i_{D-1}}\\\text{where} \quad \forall_{0 \leq d < D} \quad i_d = \left\lfloor \frac{j_d}{B_d} \right\rfloor\end{aligned}\end{align} \]

Parameters:

• tensor (Tensor) – Tensor to dequantize

• scale (Tensor) – Scale for dequantization

• offset (Tensor) – Offset for dequantization

• block_size (Tuple[int, ...], optional) – Block size