DequantizeLinear - 10 vs 13

Files changed (1) hide show
  1. DequantizeLinear10 → DequantizeLinear13 +16 -7
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
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**:
10
+ (Optional) The axis of the dequantizing dimension of the input
11
+ tensor. Ignored for per-tensor quantization. Negative value means
12
+ counting dimensions from the back. Accepted range is [-r, r-1] where
13
+ r = rank(input).
5
14
  **Inputs**
6
15
  Between 2 and 3 inputs.
7
16
  * **x** (heterogeneous) - **T**:
8
17
  N-D quantized input tensor to be de-quantized.
9
18
  * **x_scale** (heterogeneous) - **tensor(float)**:
10
- Scale for input 'x'. It's a scalar, which means a per-tensor/layer
19
+ Scale for input 'x'. It can be a scalar, which means a per-
20
+ tensor/layer dequantization, or a 1-D tensor for per-axis
11
- quantization.
21
+ dequantization.
12
22
  * **x_zero_point** (optional, heterogeneous) - **T**:
23
+ Zero point for input 'x'. Shape must match x_scale. It's optional.
13
- Zero point for input 'x'. It's a scalar, which means a per-
14
- tensor/layer quantization. It's optional. 0 is the default value
15
- when it's not specified.
24
+ Zero point is 0 when it's not specified.
16
25
  **Outputs**
17
26
  * **y** (heterogeneous) - **tensor(float)**:
18
27
  N-D full precision output tensor. It has same shape as input 'x'.
19
28
  **Type Constraints**
20
29
  * **T** in (
21
30
  tensor(int32),
22
31
  tensor(int8),
23
32
  tensor(uint8)
24
33
  ):
25
34
  Constrain 'x_zero_point' and 'x' to 8-bit/32-bit integer tensor.