aclnnRmsNorm

📄 View Source Code

Applicable Products

Function

• Description: The RmsNorm operator is a normalization operation commonly used in large models. Compared with the LayerNorm operator, it removes the part of subtracting the mean.

• Formula:

  $$\operatorname{RmsNorm}(x_i)=\frac{1}{\operatorname{Rms}(\mathbf{x})} * x_i * gamma_i, \quad \text { where } \operatorname{Rms}(\mathbf{x})=\sqrt{\frac{1}{n} \sum_{i=1}^n x_i^2+epsilon}$$

Function Prototype

Each operator uses a two-phase API. You must first call the aclnnRmsNormGetWorkspaceSize API to obtain the input parameters and calculate the required workspace size based on the computation process, and then call the aclnnRmsNorm API to perform the computation.

aclnnStatus aclnnRmsNormGetWorkspaceSize(
  const aclTensor *x,
  const aclTensor *gamma,
  double           epsilon,
  const aclTensor *yOut,
  const aclTensor *rstdOut,
  uint64_t        *workspaceSize,
  aclOpExecutor  **executor)

aclnnStatus aclnnRmsNorm(
  void          *workspace,
  uint64_t       workspaceSize,
  aclOpExecutor *executor,
  aclrtStream    stream)

aclnnRmsNormGetWorkspaceSize

• Parameters

  Parameter	Input/Output	Description	Instruction	Data Type	Data Format	Dimension (Shape)	Non-Contiguous Tensor
  x (aclTensor*)	Input	Indicates the input for normalization computation, corresponding to `x` in the formula.	Supports null tensor.	FLOAT32, FLOAT16, BFLOAT16	ND	1-8	√
  gamma (aclTensor*)	Input	Indicates the scaling factor (weight) for normalization computation, corresponding to `gamma` in the formula.	Supports null Tensor. The shape must satisfy gamma_shape = x_shape[n:], n < x_shape.dims().	FLOAT32, FLOAT16, BFLOAT16	ND	1-8	√
  epsilon (double)	Input	Indicates the value added to the variance to avoid division by zero. Corresponds to `epsilon` in the formula.	The default value is 1e-6.	-	-	-	-
  yOut (aclTensor*)	Output	Indicates the final output after normalization, corresponding to `RmsNorm(x)` in the formula.	Supports null tensor. The shape is consistent with the shape of the input `x`.	FLOAT32, FLOAT16, BFLOAT16	ND	1-8	√
  rstdOut (aclTensor*)	Output	Indicates the reciprocal of the standard deviation after normalization, used for the normalization operation, corresponding to the reciprocal of `Rms(x)` in the formula.	Supports null tensor. The shape is consistent with the first several dimensions of the input parameter `x`. The first several dimensions refer to the dimensions of `x` minus the dimensions of `gamma`, indicating the dimensions that do not require normalization.	FLOAT32	ND	1-8	√
  workspaceSize (uint64_t*)	Output	Returns the workspace size that the user needs to apply on the Device side.	-	-	-	-	-
  executor (aclOpExecutor**)	Output	Returns the op executor, which contains the operator computation flow.	-	-	-	-	-

  • Atlas inference products: The data types of parameters x, gamma, and yOut do not support BFLOAT16.

• Return Value

  aclnnStatus: return code. For details, see aclnn Return Code.

  The first-phase API performs input parameter validation and returns an error in the following scenarios:

  Return Code	Error Code	Description
  ACLNN_ERR_PARAM_NULLPTR	161001	If the passed parameter is a required input, output, or required attribute and is a null pointer, 161001 is returned.
  ACLNN_ERR_PARAM_INVALID	161002	The data type of the input or output is not within the supported range.
  ACLNN_ERR_INNER_TILING_ERROR	561002	The parameter does not meet the requirements in Parameters.

aclnnRmsNorm

• Parameters

  Parameter	Input/Output	Description
  workspace	Input	Workspace memory address applied on the Device side.
  workspaceSize	Input	Workspace size applied on the Device side, obtained through the first-phase API aclnnRmsNormGetWorkspaceSize.
  executor	Input	Operator executor, which contains the operator computation process.
  stream	Input	Specifies the stream for task execution.

• Return Value

  aclnnStatus: return code. For details, see aclnn Return Code.

Constraints

• Atlas inference products: The last axis length of the x and gamma inputs must be greater than or equal to 32 bytes.

• Boundary value scenario description:

  • Atlas inference products: The input does not support containing Inf and NaN.

  • Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products, Ascend 950PR/Ascend 950DT: When the input is Inf, the output is Inf. When the input is NaN, the output is NaN.

• Supported data types by platform:

  • Atlas A2 training products/Atlas A2 inference products, Atlas A3 training products/Atlas A3 inference products, Ascend 950PR/Ascend 950DT:

    x Data Type	gamma Data Type	yOut Data Type	rstdOut Data Type
    FLOAT16	FLOAT32	FLOAT16	FLOAT32
    BFLOAT16	FLOAT32	BFLOAT16	FLOAT32
    FLOAT16	FLOAT16	FLOAT16	FLOAT32
    BFLOAT16	BFLOAT16	BFLOAT16	FLOAT32
    FLOAT32	FLOAT32	FLOAT32	FLOAT32

  • Atlas inference products:

    x Data Type	gamma Data Type	yOut Data Type	rstdOut Data Type
    FLOAT16	FLOAT16	FLOAT16	FLOAT32
    FLOAT32	FLOAT32	FLOAT32	FLOAT32

• deterministic computation:

  • aclnnRmsNorm defaults to a deterministic implementation.

Example

The following provides example code for reference only. For details about compilation and running, see Compile and Run Samples.

#include <iostream>
#include <vector>
#include "acl/acl.h"
#include "aclnnop/aclnn_rms_norm.h"

#define CHECK_RET(cond, return_expr) \
    do {                             \
        if (!(cond)) {               \
            return_expr;             \
        }                            \
    } while (0)

#define LOG_PRINT(message, ...)         \
    do {                                \
        printf(message, ##__VA_ARGS__); \
    } while (0)

int64_t GetShapeSize(const std::vector<int64_t>& shape)
{
    int64_t shape_size = 1;
    for (auto i : shape) {
        shape_size *= i;
    }
    return shape_size;
}

int Init(int32_t deviceId, aclrtStream* stream)
{
    // Boilerplate, resource initialization.
    auto ret = aclInit(nullptr);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclInit failed. ERROR: %d\n", ret); return ret);
    ret = aclrtSetDevice(deviceId);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSetDevice failed. ERROR: %d\n", ret); return ret);
    ret = aclrtCreateStream(stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtCreateStream failed. ERROR: %d\n", ret); return ret);

    return 0;
}

template <typename T>
int CreateAclTensor(
    const std::vector<T>& hostData, const std::vector<int64_t>& shape, void** deviceAddr, aclDataType dataType,
    aclTensor** tensor)
{
    auto size = GetShapeSize(shape) * sizeof(T);
    // Call aclrtMalloc to apply for device-side memory.
    auto ret = aclrtMalloc(deviceAddr, size, ACL_MEM_MALLOC_HUGE_FIRST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMalloc failed. ERROR: %d\n", ret); return ret);

    // Call aclrtMemcpy to copy host-side data to device-side memory.
    ret = aclrtMemcpy(*deviceAddr, size, hostData.data(), size, ACL_MEMCPY_HOST_TO_DEVICE);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtMemcpy failed. ERROR: %d\n", ret); return ret);

    // Calculate the strides of a contiguous tensor.
    std::vector<int64_t> strides(shape.size(), 1);
    for (int64_t i = shape.size() - 2; i >= 0; i--) {
        strides[i] = shape[i + 1] * strides[i + 1];
    }

    // Call aclCreateTensor to create an aclTensor.
    *tensor = aclCreateTensor(
        shape.data(), shape.size(), dataType, strides.data(), 0, aclFormat::ACL_FORMAT_ND, shape.data(), shape.size(),
        *deviceAddr);
    return 0;
}

int main()
{
    // 1. (Boilerplate) Initialize the device/stream. See the acl API manual.
    // Fill in the deviceId based on the actual device.
    int32_t deviceId = 0;
    aclrtStream stream;
    auto ret = Init(deviceId, &stream);
    // check Handle as needed.
    CHECK_RET(ret == 0, LOG_PRINT("Init acl failed. ERROR: %d\n", ret); return ret);
    // 2. Construct the input and output. Customize the construction based on the API.
    std::vector<int64_t> xShape = {2, 16};
    std::vector<int64_t> gammaShape = {16};
    std::vector<int64_t> yShape = {2, 16};
    std::vector<int64_t> rstdShape = {2, 1};
    void* xDeviceAddr = nullptr;
    void* gammaDeviceAddr = nullptr;
    void* yDeviceAddr = nullptr;
    void* rstdDeviceAddr = nullptr;
    aclTensor* x = nullptr;
    aclTensor* gamma = nullptr;
    aclTensor* y = nullptr;
    aclTensor* rstd = nullptr;
    std::vector<float> xHostData = {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7,
                                    0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7};
    std::vector<float> gammaHostData = {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7};
    std::vector<float> yHostData = {0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7,
                                    0, 1, 2, 3, 4, 5, 6, 7, 0, 1, 2, 3, 4, 5, 6, 7};
    std::vector<float> rstdHostData = {1, 2};
    double epsilon = 1e-6;

    // Create the x aclTensor.
    ret = CreateAclTensor(xHostData, xShape, &xDeviceAddr, aclDataType::ACL_FLOAT, &x);
    CHECK_RET(ret == ACL_SUCCESS, return ret);
    // Create the gamma aclTensor.
    ret = CreateAclTensor(gammaHostData, gammaShape, &gammaDeviceAddr, aclDataType::ACL_FLOAT, &gamma);
    CHECK_RET(ret == ACL_SUCCESS, return ret);
    // Create y aclTensor.
    ret = CreateAclTensor(yHostData, yShape, &yDeviceAddr, aclDataType::ACL_FLOAT, &y);
    CHECK_RET(ret == ACL_SUCCESS, return ret);
    // Create rstd aclTensor.
    ret = CreateAclTensor(rstdHostData, rstdShape, &rstdDeviceAddr, aclDataType::ACL_FLOAT, &rstd);
    CHECK_RET(ret == ACL_SUCCESS, return ret);

    // 3. Call the CANN operator library API. Replace with the specific API.
    uint64_t workspaceSize = 0;
    aclOpExecutor* executor;
    // Call the first-phase API of aclnnRmsNorm.
    ret = aclnnRmsNormGetWorkspaceSize(x, gamma, epsilon, y, rstd, &workspaceSize, &executor);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnRmsNormGetWorkspaceSize failed. ERROR: %d\n", ret); return ret);
    // Apply for device memory based on the workspaceSize calculated by the first-phase API.
    void* workspaceAddr = nullptr;
    if (workspaceSize > 0) {
        ret = aclrtMalloc(&workspaceAddr, workspaceSize, ACL_MEM_MALLOC_HUGE_FIRST);
        CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("allocate workspace failed. ERROR: %d\n", ret); return ret;);
    }
    // Call the second-phase API of aclnnRmsNorm.
    ret = aclnnRmsNorm(workspaceAddr, workspaceSize, executor, stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclnnRmsNorm failed. ERROR: %d\n", ret); return ret);
    // 4. (Boilerplate) Synchronously wait for task execution to complete.
    ret = aclrtSynchronizeStream(stream);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("aclrtSynchronizeStream failed. ERROR: %d\n", ret); return ret);
    // 5. Obtain the output value. Copy the result from the device memory to the host. Modify based on the specific API definition.
    auto size = GetShapeSize(yShape);
    std::vector<float> resultData(size, 0);
    ret = aclrtMemcpy(
        resultData.data(), resultData.size() * sizeof(resultData[0]), yDeviceAddr, size * sizeof(float),
        ACL_MEMCPY_DEVICE_TO_HOST);
    CHECK_RET(ret == ACL_SUCCESS, LOG_PRINT("copy result from device to host failed. ERROR: %d\n", ret); return ret);
    for (int64_t i = 0; i < size; i++) {
        LOG_PRINT("result[%ld] is: %f\n", i, resultData[i]);
    }

    // 6. Release aclTensor and aclScalar. Modify based on the specific API definition.
    aclDestroyTensor(x);
    aclDestroyTensor(gamma);
    aclDestroyTensor(y);
    aclDestroyTensor(rstd);

    // 7. Release device resources. Modify based on the specific API definition.
    aclrtFree(xDeviceAddr);
    aclrtFree(gammaDeviceAddr);
    aclrtFree(yDeviceAddr);
    aclrtFree(rstdDeviceAddr);
    if (workspaceSize > 0) {
        aclrtFree(workspaceAddr);
    }
    aclrtDestroyStream(stream);
    aclrtResetDevice(deviceId);
    aclFinalize();
    return 0;
}