DynamicQuantizeLinear#

Domain: ai.onnx
Since version: 11

A Function to fuse calculation for Scale, Zero Point and FP32->8Bit conversion of FP32 Input data. Outputs Scale, ZeroPoint and Quantized Input for a given FP32 Input. Scale is calculated as:

y_scale = (maximum(0, max(x)) - minimum(0, min(x))) / (qmax - qmin)

where qmax and qmin are max and min values for quantization range i.e. [0, 255] in case of uint8
data range is adjusted to include 0.

Zero point is calculated as:

intermediate_zero_point = qmin - min(x)/y_scale
y_zero_point = cast(round(saturate(intermediate_zero_point)))

where qmax and qmin are max and min values for quantization range .i.e [0, 255] in case of uint8
for saturation, it saturates to [0, 255] if it’s uint8, or [-127, 127] if it’s int8. Right now only uint8 is supported.
rounding to nearest ties to even.

Data quantization formula is:

y = saturate (round (x / y_scale) + y_zero_point)

for saturation, it saturates to [0, 255] if it’s uint8, or [-127, 127] if it’s int8. Right now only uint8 is supported.
rounding to nearest ties to even.

Inputs

x (T1): Input tensor

Outputs

y (T2): Quantized output tensor
y_scale (tensor(float)): Output scale. It’s a scalar, which means a per-tensor/layer quantization.
y_zero_point (T2): Output zero point. It’s a scalar, which means a per-tensor/layer quantization.

Type Constraints

T1: Constrain ‘x’ to float tensor. Allowed types: tensor(float).
T2: Constrain ‘y_zero_point’ and ‘y’ to 8-bit unsigned integer tensor. Allowed types: tensor(uint8).

Examples#

test_dynamicquantizelinear

Node:
  DynamicQuantizeLinear(x) -> (y, y_scale, y_zero_point)

Inputs:
  x: shape=(6,), dtype=float32
    [ 0.  ,  2.  , -3.  , -2.5 ,  1.34,  0.5 ]

Outputs:
  y: shape=(6,), dtype=uint8
    [153, 255,   0,  26, 221, 179]
  y_scale: shape=(), dtype=float32
    0.01960784
  y_zero_point: shape=(), dtype=uint8
    153

test_dynamicquantizelinear_max_adjusted

Node:
  DynamicQuantizeLinear(x) -> (y, y_scale, y_zero_point)

Inputs:
  x: shape=(6,), dtype=float32
    [-1.  , -2.1 , -1.3 , -2.5 , -3.34, -4.  ]

Outputs:
  y: shape=(6,), dtype=uint8
    [191, 121, 172,  96,  42,   0]
  y_scale: shape=(), dtype=float32
    0.01568628
  y_zero_point: shape=(), dtype=uint8
    255

test_dynamicquantizelinear_min_adjusted

Node:
  DynamicQuantizeLinear(x) -> (y, y_scale, y_zero_point)

Inputs:
  x: shape=(3, 4), dtype=float32
    [[1.   , 2.1  , 1.3  , 2.5  ],
     [3.34 , 4.   , 1.5  , 2.6  ],
     [3.9  , 4.   , 3.   , 2.345]]

Outputs:
  y: shape=(3, 4), dtype=uint8
    [[ 64, 134,  83, 159],
     [213, 255,  96, 166],
     [249, 255, 191, 149]]
  y_scale: shape=(), dtype=float32
    0.01568628
  y_zero_point: shape=(), dtype=uint8
    0