aclmdlBundleLoadModelWithMem

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

Loads a graph from the model based on the graph index. The model workspace is managed by the user.

Prototype

      
           aclError aclmdlBundleLoadModelWithMem(uint32_t bundleId, size_t index, void *workPtr, size_t workSize, void *weightPtr, size_t weightSize, uint32_t *modelId)

Parameters

Parameter	Input/Output	Description
bundleId	Input	Bundle ID returned after the model is successfully initialized by calling aclmdlBundleInitFromFile or aclmdlBundleInitFromMem.
index	Input	Index. After aclmdlBundleGetQueryModelNum is called to obtain the total number of graphs, the value range of the index is [0, total_number_of_graphs – 1].
workPtr	Input	Pointer to the address of the workspace (for storing temporary data during model execution) required by the model on the device. The workspace is managed by the user and cannot be freed during model execution. If a null pointer is passed to workPtr, the workspace is managed by the system. In the event where the memory is managed by the user, if multiple models are executed in serial, the models can share a workspace. However, you need to guarantee the serial execution sequence of the models and the workspace size (the same as the total size of the workspaces needed by all the models). Refer to the following description to ensure serial execution: For synchronous model execution, add a lock to ensure that tasks are executed in serial. For asynchronous model execution, use a single stream to ensure that tasks are executed in serial.
workSize	Input	Workspace size required for model execution, in bytes. Invalid when workPtr is a null pointer.
weightPtr	Input	Pointer to the address of the model weight memory (for storing weight data) on the device. The memory is managed by the user and cannot be freed during model execution. If a null pointer is passed to weightPtr, the memory is managed by the system. In the multithreaded scenario where the weight memory is managed by the user, if a model is loaded once in each thread, weightPtr can be shared because the weight memory is read-only at inference time. Note that weightPtr cannot be freed when the sharing is in progress.
weightSize	Input	Weight memory size required for model execution, in bytes. Invalid when weightPtr is a null pointer.
modelId	Output	ID of the executable graph.

Returns

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

Example

For details about the API call sequence and sample code, see Weight update (user-managed memory).

Parent topic: Model Loading and Unloading