…ment prediction computation for mnist (#273)

* feat: Add prediction service data pipeline to learner

* fix: borken tests

* chores: add prediction integration test using GRPCServer and ExampleGRPCLearnerClient

* fix: failing tests: use legacy.Adam - disable keras_mnist_diffpriv example

* chores: disable py3.8 tests - py3.7 should be enough for now

* fix: minor

* fix: remove ml frameworks depdencies from ml_interface

* feat: add mnist keras predicton service (#274)

* Added draft of mnist keras prediction mli.

* refactoring to more generic model input size.

* Added mnist image for testing prediction service.

* Added resize possible and more test images.

---------

Co-authored-by: Hanna Wagner <[email protected]>

---------

Co-authored-by: Hanna Wagner <[email protected]>

* Adding draft of generic prediction service.

* Added consistend naming and pred to grpc server.

* Fixed annotation error.

* Tmp fix for making the test work.

* Start with default pdl from tuple to dict.

* Finished basic generic prediction service.

* Update colearn_grpc/example_mli_factory.py

* Update colearn_grpc/example_mli_factory.py

---------

Co-authored-by: Lokman Rahmani <[email protected]>

* First draft of scania prediction implementation.

* Added second pred data loader.

* Init multiple metrics.

* Adapt input shape to match new metrics for scania.

* Updated keras learner and fixed some tets.

* Added multiple metrics for mnist.

* Adapted interface proto.

* Updated ProposedWeights in examples.

* Changed other learner classes to multiple metrics.

* test push.

* chores: enable github actions for prediction feature branch

* Ignore tf module pylint errors.

* depend fix.

* increase test timeout.

* fixing pytorch errors.

* fixed flake 8 and key error.

* fix vote criterion for pytorch learners.

* uncomment short pystest.

* Fix criterion in pytorch tests.

---------

Co-authored-by: lrahmani <[email protected]>