Softmax¶
General¶
Softmax operation applies following formula on every element of \(\src\) tensor (the variable names follow the standard Naming Conventions):
\[dst_i = \frac{exp(src_i)}{\sum_{j=1}^{C} exp(src_j)}\]
where \(C\) is a size of tensor along axis dimension.
Operation attributes¶
Attribute Name |
Descr |
|||
|---|---|---|---|---|
Represents the axis of which the Softmax is calculated. |
s64 |
Arbitrary s64 value ( |
Optional |
Execution arguments¶
The inputs and outputs must be provided according to below index order when constructing an operation.
Inputs¶
Index |
Argu |
|
|---|---|---|
0 |
|
Required |
Outputs¶
Index |
Argu |
|
|---|---|---|
0 |
|
Required |
Supported data types¶
Softmax operation supports the following data type combinations.
Src |
D |
|---|---|
f32 |
f32 |
bf16 |
bf16 |
f16 |
f16 |