Now that you've distributed a function, it is time to distribute a class object. At Neuro we focus on building an API on ML compute so it is important that you know how to distribute more complex objects such as ML models.
Modify/improve the previous challenge to allow distributing
across different workers the child class MyModel that inherits from the parent class NpuModel as
shown in the following examples: Pytorch example and Xgbost example.
Build on top of your previous API, the following:
- Build a function called
artifact, found inside init file that will be used as decorator to pass the required model to the parent class. - Build a parent class called
NpuModel, found inside objects.py that will allow handling artifacts and distribute the compute defined on the child class. - Build the pytorch and xgboost objects, called
PytorchModelandXgboostModelrespectively, found inside frameworks.py that will provide any required info as to what type of object is and how to serialise it for your distributing API. - If the above is completed, Pytorch example and Xgbost example should work as expected.