Closed mocleiri closed 2 years ago
I just noticed that the interpreter is being recreated on each invoke call. We should be able to keep the initialized instance around and store the tensorflow c++ interpreter as an mp_obj_t on the microlite_interpreter_t object.
This has been merged. We probably have to revisit the destructor case.
I just noticed that the interpreter is being recreated on each invoke call. We should be able to keep the initialized instance around and store the tensorflow c++ interpreter as an mp_obj_t on the microlite_interpreter_t object.