[object Object]

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

• Description: Perform the backward pass of the single-path rotary position encoding .

• Formula:

  In the forward pass of rotary position encoding, the axis list for broadcasting is [object Object], and the computation formula can be expressed as follows:

  • [object Object]Atlas A3 training products/Atlas A3 inference products[object Object], [object Object]Atlas A2 training products/Atlas A2 inference products[object Object]:

  (1) Half mode (mode equals 0):

  $$dy1, dy2 = chunk(dy, chunks=2, dim=-1)$$ $$cos1, cos2 = chunk(cos, chunks=2, dim=-1)$$ $$sin1, sin2 = chunk(sin, chunks=2, dim=-1)$$ $$x1, x2 = chunk(x, chunks=2, dim=-1)$$ $$dx = cat((cos1 * dy1 + sin2 * dy2, cos2 * dy2 - sin1 * dy1), dim=-1)$$ $$dcos = sum(dy * x, dims)$$ $$dsin = sum(dy * cat((-x2, x1), dim=-1), dims)$$

  (2) Interleave mode (mode equals 1):

  $$dy1, dy2 = dy[..., :: 2], dy[..., 1 :: 2]$$ $$cos1, cos2 = cos[..., :: 2], cos[..., 1 :: 2]$$ $$sin1, sin2 = sin[..., :: 2], sin[..., 1 :: 2]$$ $$x1, x2 = x[..., :: 2], x[..., 1 :: 2]$$ $$dx = stack((cos1 * dy1 + sin2 * dy2, cos2 * dy2 - sin1 * dy1), dim=-1).reshape(dy.shape)$$ $$dcos = sum(dy * x, dims)$$ $$dsin = sum(dy * stack((-x2, x1), dim=-1).reshape(dy.shape), dims)$$

  (3) Quarter mode (mode equals 2):

  $$dy1, dy2, dy3, dy4 = chunk(dy, chunks=4, dim=-1)$$ $$cos1, cos2, cos3, cos4 = chunk(cos, chunks=4, dim=-1)$$ $$sin1, sin2, sin3, sin4 = chunk(sin, chunks=4, dim=-1)$$ $$x1, x2, x3, x4 = chunk(x, chunks=4, dim=-1)$$ $$dx = cat((cos1 * dy1 + sin2 * dy2, cos2 * dy2 - sin1 * dy1, cos3 * dy3 + sin4 * dy4, cos4 * dy4 - sin3 * dy3), dim=-1)$$ $$dcos = sum(dy * x, dims)$$ $$dsin = sum(dy * cat((-x2, x1, -x4, x3), dim=-1), dims)$$

  (4) Interleave-half mode (mode equals 3):

  $$dy1, dy2 = chunk(dy, chunks=2, dim=-1)$$ $$cos1, cos2 = chunk(cos, chunks=2, dim=-1)$$ $$sin1, sin2 = chunk(sin, chunks=2, dim=-1)$$ $$x1, x2 = x[..., :: 2], x[..., 1 :: 2]$$ $$dx = stack((cos1 * dy1 + sin2 * dy2, cos2 * dy2 - sin1 * dy1), dim=-1).reshape(dy.shape)$$ $$dcos = sum(dy * cat((x1, x2), dim=-1), dims)$$ $$dsin = sum(dy * cat((-x2, x1), dim=-1), dims)$$

[object Object]

Each operator has calls. First call [object Object] to obtain the input parameters and calculate the required workspace size according to the process, and then call [object Object] to perform the computation.

[object Object]

• Parameters

  [object Object]
  • Parameter mode constraints:
    • [object Object]Atlas A3 training products/Atlas A3 inference products[object Object] and [object Object]Atlas A2 training products/Atlas A2 inference products[object Object]: 0=half, 1=interleave.

• 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 Computation:

  • [object Object] defaults to deterministic implementation.

  • [object Object]Atlas A3 training products/Atlas A3 inference products[object Object] and [object Object]Atlas A2 training products/Atlas A2 inference products[object Object]:

    • The input tensor [object Object] supports BNSD, BSND, SBND, and TND layouts.
    • The D dimensions of the input tensors [object Object], [object Object], [object Object], [object Object], and the output tensors [object Object], [object Object], [object Object] must be the same, satisfy D < 896, and must be a multiple of 2.
    • The shapes of the input tensors [object Object], [object Object], and the output tensor [object Object] must be exactly the same.
    • The shapes of the input tensors [object Object], [object Object] and the output tensors [object Object], [object Object] must be exactly the same, and the shapes of [object Object] and [object Object] must also be exactly the same.
    • Half mode:
      • B, N < 1000; When calculating [object Object] and [object Object], B * N <= 1024.
      • When [object Object] is BNSD, [object Object] and [object Object] support 11SD, B1SD, and BNSD; when [object Object] and [object Object] are B1SD, the condition B < S must be satisfied.
      • When [object Object] is BSND, [object Object] and [object Object] support 1S1D, BS1D, BSND; when [object Object] and [object Object] are BS1D, B < S must be satisfied.
      • When [object Object] is SBND, [object Object] and [object Object] support S11D, SB1D, SBND.
      • When [object Object] is TND, [object Object] and [object Object] support T1D and TND.
    • Interleave mode:
      • B * N < 1000 (N < 1000 when [object Object] is TND).
      • When [object Object] is BNSD, [object Object] and [object Object] support 11SD.
      • When [object Object] is BSND, [object Object] and [object Object] support 1S1D.
      • When [object Object] is SBND, [object Object] and [object Object] support S11D.
      • When [object Object] is TND, [object Object] and [object Object] support T1D.

[object Object]

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