[object Object]

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

• Description: Performs root mean square layer normalization (RMSNorm), a normalization method widely used in large language models. Unlike LayerNorm, RMSNorm omits mean subtraction. The AddRmsNorm operator fuses the preceding Add operator with RMSNorm to reduce data movement operations.
• Formula:$$x_i=x1Ref_i+x2Ref_i$$ $$\operatorname{RmsNorm}(x_i)=\frac{1}{\operatorname{Rms}(\mathbf{x})} * x_i * gamma_i, \quad \text { where } \operatorname{Rms}(\mathbf{x})=\sqrt{\frac{1}{n} \sum_{i=1}^n x_i^2+epsilon}$$

[object Object]

Each operator has calls. First, aclnnInplaceAddRmsNormGetWorkspaceSize is called to obtain the input parameters and compute the required workspace size based on the process. Then, aclnnInplaceAddRmsNorm is called to perform computation.

[object Object]

• Parameters

  [object Object]
  • [object Object]Atlas inference products[object Object]
    • The data types of x1Ref, x2Ref, and gamma cannot be BFLOAT16.
    • The parameter [object Object] is invalid in current product usage scenarios.

• Returns

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

  The first-phase API implements input parameter validation. The following error codes may be returned.

  [object Object]

[object Object]

• Parameters

  [object Object]

• Returns

  aclnnStatus: status code. For details, see .

[object Object]

• Boundary value scenarios:
  • When the input is [object Object], the output is [object Object].
  • When the input is [object Object], the output is [object Object].
• Deterministic computation:
  • aclnnInplaceAddRmsNorm defaults to a deterministic implementation.

[object Object]

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