Given that Core ML does not utilize the GPU or ANE when using flexible shapes, it becomes important to allow fixed shapes to be enforced for some use cases. One yet-to-be-tested idea is to use enumerated shapes that cover the range of the target model.
Given that Core ML does not utilize the GPU or ANE when using flexible shapes, it becomes important to allow fixed shapes to be enforced for some use cases. One yet-to-be-tested idea is to use enumerated shapes that cover the range of the target model.