DequantizeLinear - 10 vs 13

Next section compares an older to a newer version of the same operator after both definition are converted into markdown text. Green means an addition to the newer version, red means a deletion. Anything else is unchanged.

Files changed (1) hide show
  1. DequantizeLinear10 → DequantizeLinear13 +12 -5
DequantizeLinear10 → DequantizeLinear13 RENAMED
@@ -1 +1 @@
1
- The linear dequantization operator. It consumes a quantized tensor, a scale, a zero point to compute the full precision tensor.
1
+ The linear dequantization operator. It consumes a quantized tensor, a scale, and a zero point to compute the full precision tensor.
2
- The dequantization formula is y = (x - x_zero_point) * x_scale. 'x_scale' and 'x_zero_point' are both scalars.
2
+ The dequantization formula is y = (x - x_zero_point) * x_scale. x_scale and x_zero_point must have same shape, and can be either a scalar
3
+ for per-tensor / per layer quantization, or a 1-D tensor for per-axis quantization.
3
- 'x_zero_point' and 'x' must have same type. 'x' and 'y' must have same shape. In the case of dequantizing int32,
4
+ x_zero_point and x must have same type. x and y must have same shape. In the case of dequantizing int32,
4
5
  there's no zero point (zero point is supposed to be 0).
6
+
7
+ ### Attributes
8
+
9
+ * **axis - INT** (default is '1'):
10
+
11
+ (Optional) The axis of the dequantizing dimension of the input tensor. Ignored for per-tensor quantization. Negative value means counting dimensions from the back. Accepted range is [-r, r-1] where r = rank(input).
5
12
  ### Inputs
6
13
  Between 2 and 3 inputs.
7
14
  - **x** (heterogeneous) - **T**:
8
15
  N-D quantized input tensor to be de-quantized.
9
16
  - **x_scale** (heterogeneous) - **tensor(float)**:
10
- Scale for input 'x'. It's a scalar, which means a per-tensor/layer quantization.
17
+ Scale for input 'x'. It can be a scalar, which means a per-tensor/layer dequantization, or a 1-D tensor for per-axis dequantization.
11
18
  - **x_zero_point** (optional, heterogeneous) - **T**:
12
- Zero point for input 'x'. It's a scalar, which means a per-tensor/layer quantization. It's optional. 0 is the default value when it's not specified.
19
+ Zero point for input 'x'. Shape must match x_scale. It's optional. Zero point is 0 when it's not specified.
13
20
  ### Outputs
14
21
  - **y** (heterogeneous) - **tensor(float)**:
15
22
  N-D full precision output tensor. It has same shape as input 'x'.
16
23
  ### Type Constraints
17
24
  * **T** in ( tensor(int32), tensor(int8), tensor(uint8) ):
18
25
  Constrain 'x_zero_point' and 'x' to 8-bit/32-bit integer tensor.