Understanding of how sagemaker-debugger works

I have a question on basic understanding for which i do not find answer in the doc.

If i understand well, the documentation says that we do not have to anything if we use the AWS deep learning containers.
How is the association between the hook and the network is done?
Does it use the object returned by model_fn and expect to be a mxnet module, or pytorch or tentorflow model?

Actually my model_fn() function returns an object which encapsulates my mxnet module, will it work for me, or should i create especifically the hooks , etc.. ?

Thank you

Start with the SageMaker Debugger documentation covering AWS deep learning containers, model_fn, hooks, and supported frameworks. Determine how the hook is associated with the returned model and whether a model_fn result that encapsulates an MXNet module is supported; done means the documentation answers these questions and explains when custom hooks are required.

Written by the indexing model from the issue text.