[object Object][object Object][object Object]undefined[object Object]

• Description: Scales and masks the input x, and then performs the softmax operation to obtain the output.
• Formula:$$y = Softmax((scale * x) * mask, dim = -1)$$ $$Softmax(X_i) ={e^{X_i - max(X, dim=-1)} \over \sum{e^{X_i - max(X, dim=-1)}}}$$

[object Object]

Each operator has calls. First, aclnnScaledMaskedSoftmaxGetWorkspaceSize is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, aclnnScaledMaskedSoftmax is called to perform computation.

[object Object]

• Parameters:

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

  The first-phase API implements input parameter verification. The following errors may be thrown:

  [object Object]

[object Object]

• Parameters:

  [object Object]

• Returns:

  [object Object]: status code. For details, see .

[object Object]

• Deterministic compute:

  • aclnnScaledMaskedSoftmax defaults to a deterministic implementation.

• The fourth dimension of x must be in the range of $(0, 4096]$.

• The first two dimensions in the shape of mask and those of x can be different, but must meet the .

[object Object]

The following example is for reference only. For details, see .