Defined in File custom.h
Get the custom version.
For a custom backend that doesn’t define this entry point, the inference server will assume the backend version is 1. The currently supported versions are defined below, returning any other version is an error:
Version 1: Input and output tensors must be communicated via system memory (i.e. CPU memory). The CustomExecute function must be defined and CustomGetNextInputFn_t and CustomGetOutputFn_t define the function signature for the input and output callbacks.
Version 2: Input and output tensors may be communicated by both system memory and GPU memory. The CustomExecuteV2 function must be defined and CustomGetNextInputV2Fn_t and CustomGetOutputV2Fn_t define the function signature for the input and output callbacks.
the custom version.