aclmdlGetInputDimsV2

Applicability

Product	Supported
Atlas 350 Accelerator Card	√
Atlas A3 training product/Atlas A3 inference product	√
Atlas A2 training product/Atlas A2 inference product	√
Atlas 200I/500 A2 inference product	√
Atlas inference product	√
Atlas training product	√

Function Usage

Obtains the shape of an input tensor based on the model description.

If the model contains static AIPP configuration information, call aclmdlGetInputDims or aclmdlGetInputDimsV2 to obtain the shape as required. The differences between the two APIs are as follows:

aclmdlGetInputDims: The obtained shape values are consistent with those of the input image. For details, see Table 1.
aclmdlGetInputDimsV2: The obtained size of the H dimension is that of the input image multiplied by a coefficient, and the size of the C dimension varies according to the image format. For details, see aclmdlGetInputDimsV2 in aclmdlGetInputDimsV2. The product of the dimensions is consistent with that obtained by the aclmdlGetInputSizeByIndex call.

Prototype

aclError aclmdlGetInputDimsV2(const aclmdlDesc *modelDesc, size_t index, aclmdlIODims *dims)

Parameters

Parameter	Input/Output	Description
modelDesc	Input	Pointer to data of the aclmdlDesc type. Call aclmdlCreateDesc to create data of the aclmdlDesc type in advance.
index	Input	Index of the input whose dimension is to be obtained, indexed from 0.
dims	Output	Pointer to the input dimension information. For details about the type definition, see aclmdlIODims. In dynamic batch/image size scenarios, if the batch size is -1 or the width and height are -1, the tensor shape is dynamic. For example, if the format of the input tensor is NCHW, then the tensor shape [–1,3,224,224] indicates that the batch size is dynamic, and the tensor shape [1,3,–1,–1] indicates that the image size is dynamic. The italic numbers are examples only. If the length of a tensor name exceeds 127, the API replaces it with the format acl_modelId_${id}_input_${index}_${random string} when outputting dims.name. If the converted name conflicts with an existing tensor name, a random string (_${random string}) is appended; otherwise, no random string is added. The API also maintains a mapping between the converted and original names. You can call aclmdlGetTensorRealName to retrieve the original name by passing the converted one (or simply pass the original name to get it back unchanged). For a tensor name that has a length of 127 or fewer, dims.name directly reflects the original tensor name. In the static AIPP scenario, this API applies only to the NHWC format. Table 1 describes the mapping between the supported input image formats and dims shapes.

Parameter

Input/Output

Description

modelDesc

Input

Pointer to data of the aclmdlDesc type.

Call aclmdlCreateDesc to create data of the aclmdlDesc type in advance.

index

Input

Index of the input whose dimension is to be obtained, indexed from 0.

dims

Output

Pointer to the input dimension information. For details about the type definition, see aclmdlIODims.

In dynamic batch/image size scenarios, if the batch size is -1 or the width and height are -1, the tensor shape is dynamic. For example, if the format of the input tensor is NCHW, then the tensor shape [–1,3,224,224] indicates that the batch size is dynamic, and the tensor shape [1,3,–1,–1] indicates that the image size is dynamic. The italic numbers are examples only.

If the length of a tensor name exceeds 127, the API replaces it with the format acl_modelId_${id}_input_${index}_${random string} when outputting dims.name. If the converted name conflicts with an existing tensor name, a random string (_${random string}) is appended; otherwise, no random string is added. The API also maintains a mapping between the converted and original names. You can call aclmdlGetTensorRealName to retrieve the original name by passing the converted one (or simply pass the original name to get it back unchanged). For a tensor name that has a length of 127 or fewer, dims.name directly reflects the original tensor name.

In the static AIPP scenario, this API applies only to the NHWC format. Table 1 describes the mapping between the supported input image formats and dims shapes.

**Table 1** Mapping between the supported image formats and dims shapes with static AIPP
Image Format	Format	Shape
YUV420SP_U8	NHWC	n,h*1.5,w,1
XRGB8888_U8	NHWC	n,h,w,4
RGB888_U8	NHWC	n,h,w,3
YUV400_U8	NHWC	n,h,w,1
ARGB8888_U8	NHWC	n,h,w,4
YUYV_U8	NHWC	n,h,w,2
YUV422SP_U8	NHWC	n,h*2,w,1
AYUV444_U8	NHWC	n,h,w,4

Returns

0 on success; else, failure. For details, see aclError.

Parent topic: aclmdlDesc