As i understand, the AllOpsResolver has been replaced by MicroMutableOpResolver in the latest version of Tensorflow lite micro. And now it is required to manually add each of the required operation in code.
However in my project i am trying to build a device which is able to do predictions on a variable tsfl model without any changes in the firmware code and the need to recompile. I am able to update the flat buffer over wifi and would like to instantiate (as a first step) an OpResolver which includes all operations so that i have the worst case covered. Later i might instantiate the OpResolver based on the actually required operations by my model.
My question is if there is still a way to do this in one like it used to be by using AllOpsResolver in previous versions ? Right now i have to call the
op_resolver.AddXYZOperation() for the full list of operation to achieve this.
I understand that the memory footprint for this approach would be bad, but since i currently dont know which operations will be required in future by my changeable model, its a valid approach.