aclmdlSetDatasetTensorDesc
Applicability
|
Product |
Supported |
|---|---|
|
Atlas 350 Accelerator Card |
√ |
|
|
√ |
|
|
√ |
|
|
x |
|
|
√ |
|
|
√ |
Function Usage
Sets the tensor description of a dynamic-shape input or output before the model is executed.
Prototype
1
|
aclError aclmdlSetDatasetTensorDesc(aclmdlDataset *dataset, aclTensorDesc *tensorDesc, size_t index) |
Parameters
|
Parameter |
Input/Output |
Description |
|---|---|---|
|
dataset |
Output |
Pointer to the address of aclmdlDataset where aclTensorDesc is to be added, indicating the input or output data structure for model execution. Call aclmdlCreateDataset to create data of the aclmdlDataset type in advance, and then call aclmdlAddDatasetBuffer to add aclDataBuffer to aclmdlDataset. |
|
tensorDesc |
Input |
Pointer to the address of aclTensorDesc, indicating the input or output tensor description for model execution. Call aclCreateTensorDesc to create data of the aclTensorDesc type in advance. Currently, you can set only the dimension information in the model input and output tensor description (namely the numDims parameter that indicates the number of dimensions and the dims parameter that indicates the dimension size in the aclCreateTensorDesc API). You cannot set the data type or format. The number and size of dimensions must be within the range of the input shape set during model build. |
|
index |
Input |
Input or output index. When a model has multiple inputs and outputs, you can call aclmdlGetInputNameByIndex and aclmdlGetOutputNameByIndex to obtain the input and output names and set the names based on their indexes to avoid index errors. |
Returns
0 on success; else, failure. For details, see aclError.
Restrictions
For a single model, only one of the aclmdlSetDynamicBatchSize, aclmdlSetDynamicHWSize, aclmdlSetInputDynamicDims, and aclmdlSetDatasetTensorDesc APIs can be called.