SetSingleOutputShape

Applicability

Product	Supported
Atlas A3 training products/Atlas A3 inference products	√
Atlas A2 training products/Atlas A2 inference products	√
Atlas 200I/500 A2 inference products	x
Atlas inference product's AI Core	x
Atlas inference product's Vector Core	x
Atlas training products	x

Function

Sets the shape of the result matrix Output on a single core.

The high-level Conv3D API currently supports the M-axis merge mode for output. In the M-axis merge mode, the Conv3D API internally processes Wout and Hout as the same axis. The output is first performed along the Wout direction. After an entire row of Wout is output, the next row of Wout is output.

Figure 1 M-axis merge mode diagram

Prototype

__aicore__ inline void SetSingleOutputShape(uint64_t singleCo, uint64_t singleDo, uint64_t singleM)

Parameters

Parameter	Input/Output	Description
singleCo	Input	Output C dimension size on a single core.
singleDo	Input	Output D dimension size on a single core.
singleM	Input	Output M dimension size on a single core, that is, the product of the H dimension size and the W dimension size.

Returns

None

Restrictions

Currently, this API supports only the setting of the C, D, and M dimensions (that is, the dimension obtained after the H axis and W axis are merged) of the output. The size of the original output cannot be set.

Example

conv3dApi.SetSingleOutputShape(singleCoreCout, singleCoreDout, singleCoreM);

Parent topic: Conv3D Kernel APIs