Deprecate EvalModelTemplatein favor of BoringModel and another simple model that does actually learn

[Borda]

🚀 Feature

correct actual EvalModelTemplate to use new API unless it is testing other purposes or deprecated API

Motivation

better testing of the actual API

[williamFalcon]

we need to deprecate evalmodeltemplate.

[gianscarpe]

Hi @williamFalcon, I would like to start contributing to pytorch-lightning. Is this still a first good issue?

[gianscarpe]

Update:
This line is causing three tests in tests/backend/test_ddp.py to fail.
from tests.base import EvalModelTemplate in tests/backends/model_dpd.py

The error is \nModuleNotFoundError: No module named \'tests.base\'\n'. I think it could be related to this issue

[gianscarpe]

I fixed the issue related to my previous comment. I proceed with a PR, and start a new branch for implementing an EvalModel

[williamFalcon]

welcome to PL! excited to help you contribute.

Please do the deprecation only a few files at a time and make sure to ping @SeanNaren and @tchaton for guidance and advice!

[gianscarpe]

Hi @SeanNaren and @tchaton
I'm working on deprecating EvalModelTemplate and adopting BoringModel throught the test suite. And I have a couple of questions. I wrote a dispatcher function in develop_pipelines to call the right run_prediction if the model is of class BoringModel. Actually, BoringModel is just a perceptron with a hidden layer and build a build a random dataset. Do we need a more "advanced" test model?
At the moment I've refactored test_gpu and I'm working on test_cpu. I open a draft pull request to share my progress and my solution at #4623
Let me know if it sounds good to you and if you have any suggestions!

[gianscarpe]

[[> Hi @SeanNaren and @tchaton

I'm working on deprecating EvalModelTemplate and adopting BoringModel throught the test suite. And I have a couple of questions. I wrote a dispatcher function in develop_pipelines to call the right run_prediction if the model is of class BoringModel. Actually, BoringModel is just a perceptron with a hidden layer and build a build a random dataset. Do we need a more "advanced" test model?
At the moment I've refactored test_gpu and I'm working on test_cpu. I open a draft pull request to share my progress and my solution at #4623
Let me know if it sounds good to you and if you have any suggestions!

Hi @SeanNaren @tchaton @williamFalcon , I have an additional question: in test_cpu.py, are test_tbtt_cpu_model_result and test_tbptt_cpu_model_result_auto_reduce deprecated? They both use TrainResult, which should be deprecated by version 1.0