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

• Operator functions: When TP domain communication is involved, ReduceScatterV is performed first, followed by AllToAllV, and finally the received data is aggregated (multiplied by weights and then summed). When TP domain communication is not involved, AllToAllV is performed, after which the received data is aggregated (multiplied by weights and then summed), followed by fused Add + RMSNorm. Operators support dynamic scale-in. After a communication domain is created, if a faulty rank is detected, it can be removed from the communication domain, allowing operators to continue executing properly without recompilation. Special expert scenarios are also supported.

  Compared with the [object Object] API, this API has the following changes:

  • Dynamic scale-in support: The operator can run properly without recompilation after faulty ranks are removed after the created communication domain. Enable this feature by passing the [object Object] parameter.

  • Special expert scenarios are supported:

  • zeroExpert ≠ 0: Enabled by setting the [object Object] parameter to a value greater than 0.

    $$Moe(oriXOptional) = 0$$

  • copyExpert ≠ 0: Enabled by setting the [object Object] parameter to a value greater than 0 and setting a valid value for the [object Object] parameter.

    $$Moe(oriXOptional) = oriXOptional$$

  • constExpert ≠ 0: Enabled by setting the [object Object] parameter to a value greater than 0 and setting valid values for the [object Object], [object Object], [object Object], and [object Object] parameters.

    $$Moe(oriXOptional) = constExpertAlpha1Optional * oriXOptional + constExpertAlpha2Optional * constExpertVOptional$$

• Formula:

$$rsOut = ReduceScatterV(expandX)\\ ataOut = AllToAllV(rsOut)\\ combineOut = Sum(expertScales * ataOut + expertScales * sharedExpertX)\\ x = combineOut + residualX\\ y = \frac{x}{RMS(x)} * gamma,\quad\text{where}\ RMS(x) = \sqrt{\frac{1}{H}\sum_{i=1}^{H}x_{i}^{2}+normEps}$$

[object Object]

[object Object]

Each operator has calls. First, [object Object] is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, [object Object] 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

  aclnnStatus status code. For details, see .

[object Object]

• Deterministic computing:

  • [object Object]aclnnMoeDistributeCombineAddRmsNormV2[object Object] defaults to a deterministic implementation.

• [object Object]aclnnMoeDistributeDispatchV3[object Object] and [object Object]aclnnMoeDistributeCombineAddRmsNormV2[object Object] must be used together. For details, see .

• The values of expertIds, xActiveMaskOptional, elasticInfoOptional, groupEp, epWorldSize, moeExpertNum, groupTp, tpWorldSize, expertShardType, sharedExpertNum, sharedExpertRankNum, globalBs, commAlg, zeroExpertNum, copyExpertNum, constExpertNum, and HCCL_BUFFSIZE used during API calling must be the same for all ranks, at all network layers, and the same as those of DispatchV3.

• The deployment information after dynamic scale-in is transferred to the operator through the [object Object] parameter. Other parameters do not need to be modified. After dynamic scale-in, the number of MoE experts deployed on the current rank must be the same as that before scale-in.

• [object Object]Atlas A3 training products/Atlas A3 inference products[object Object]: In this scenario, a single rank contains dual dies. Therefore, the "rank" in the parameter description indicates a single die.

• moeExpertNum + zeroExpertNum + copyExpertNum + constExpertNum < MAX_INT32, where the value of MAX_INT32 is 2147483647.

• The dynamic scale-in feature cannot be enabled in the tensor parallelism scenario.

• The shape format is described as follows:

  • A: Maximum number of tokens to be distributed by the current rank. The value range is as follows:

    • When globalBs is 0, A >= BS epWorldSize min(localExpertNum, K).
    • When globalBs is not 0, A >= globalBs * min(localExpertNum, K).

  • H: Hidden layer size, which must be in the range [1024, 8192].

  • Bs: Batch sequence size (number of tokens output by the rank), which must be in the range 0 < Bs ≤ 512.

  • K: Number of top K experts. The value range is 0 < K ≤ 16 and 0 < K ≤ moeExpertNum + zeroExpertNum + copyExpertNum + constExpertNum.

  • localExpertNum: Number of experts on the current rank.

    • For shared expert ranks, localExpertNum = 1.
    • For MoE expert ranks, localExpertNum = moeExpertNum/(epWorldSize - sharedExpertRankNum). If localExpertNum > 1, TP domain communication is not supported.

• HCCL_BUFFSIZE: Before calling this API, check whether the value of the [object Object] environment variable is proper. This environment variable indicates the buffer size occupied by a single communication domain, in MB. If this environment variable is not set, the default value 200 MB is used.

  • Within an EP communication domain: The value must be greater than or equal to 2 and satisfy 1024 ^ 2 (HCCL_BUFFSIZE - 2) / 2 >= BS 2 (H + 128) (epWorldSize * localExpertNum + K + 1). [object Object] must be the number of experts assigned to the current rank when using MoE.
  • Within a TP communication domain: The value must satisfy >= (A * Align512(Align32(h * 2) + 44) + A * Align512(h * 2)) * 2.

• Constraints on the use of communication domains:

  • [object Object] and [object Object] in a model support only the same EP communication domain, and no other operators are allowed in the communication domain.
  • [object Object] and [object Object] in a model support only the same TP communication domain or both do not support a TP communication domain. If a TP communication domain is supported, no other operators are allowed in the communication domain.
  • [object Object]Atlas A3 training products/Atlas A3 inference products[object Object]: Nodes in a communication domain must be in the same SuperPoD. Cross-SuperPoD nodes are not supported.

[object Object]

Take the [object Object]Atlas A3 training products/Atlas A3 inference products[object Object] as an example. Call the [object Object] and [object Object] APIs.

• The example code is as follows:

  [object Object]