It is a data class that defines the input and output data structures of model inference.

The tensor class involves resource allocation on the device, which conflicts with the scope of MxDeInit. Therefore, the scope of the tensor class must be smaller than the scope of MxDeInit.

Supported Models

Table 1 lists supported hardware. The meanings of the symbols are as follows:

√: supported
x: not supported

**Table 1** Supported hardware
API	Atlas 200I/500 A2 inference product	Atlas inference product	Atlas 800I A2 inference product	Atlas 200/300/500 inference product
BatchConcat	√	√	√	√
Clone	√	√	√	x
GetByteSize	√	√	√	√
GetData	√	√	√	√
GetDataType	√	√	√	√
GetDeviceId	√	√	√	√
GetMemoryType	√	√	√	x
GetReferRect	√	√	√	x
GetShape	√	√	√	√
GetValidRoi	√	√	√	x
IsEmpty	√	√	√	x
IsWithMargin	√	√	√	x
Malloc	√	√	√	√
operator =	√	√	√	x
operator ==	√	√	√	x
SetShape	√	√	√	x
Tensor	√	√	√	√
~Tensor	√	√	√	√
TensorFree	√	√	√	x
TensorMalloc	√	√	√	√
ToDevice	√	√	√	√
ToDvpp	√	√	√	x
ToHost	√	√	√	√
SetReferRect	x	√	x	x
SetTensorValue	√	√	√	x
SetValidRoi	√	√	√	x
Transpose	√	√	√	x

Class Description

Supported Models