When running a model on gpu, there will be a transform from .so model file into executable opencl instructions. So 1st running will be very slow. Now I want to save such opencl instructions into a local .bin file, and I just need to read from it next time. Is there any mechanism to achieve that?
