- Description: Traverses different sampling points of feature maps of different sizes by using parameters such as the sample location, attention weights, mapped value feature, start index location of a multi-scale feature, and spatial size of a multi-scale feature map (which facilitates changing a sampling location from a normalized value to an absolute location). A reverse operator calculates the gradient corresponding to the input based on the contribution of the forward input to the output and an initial gradient.
- Formula:
Each operator has calls. First, aclnnMultiScaleDeformableAttentionGradGetWorkspaceSize is called to obtain the workspace size required for computation and the executor that contains the operator computation process. Then, aclnnMultiScaleDeformableAttentionGrad is called to perform computation.
[object Object]
[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]
Deterministic compute
- aclnnMultiScaleDeformableAttentionGrad defaults to a non-deterministic implementation and does not support deterministic implementation currently.
channels%8 = 0, and channels ≤ 256
num_queries < 500000
num_levels ≤ 16
num_heads ≤ 16
num_points ≤ 16
The following example is for reference only. For details, see .
[object Object]