Do not copy params in get_params.

[mcarbajo]

Summary

For compatibility with the scikit-learn clone function get_params has to return the same parameters as given in the init function. Fix issue #13586

Related Issues

#13586

PR Overview

• This PR requires new unit tests [y/n] (make sure tests are included)
• This PR requires to update the documentation [y/n] (make sure the docs are up-to-date)
• This PR is backwards compatible [y/n]
• This PR changes the current API [y/n] (all API changes need to be approved by fchollet)

[zachmayer]

One test on python 3.6/TF 2.0 failed:
[gw1] [ 94%] FAILED tests/keras/utils/data_utils_test.py::test_ordered_enqueuer_timeout_threads

===== 1 failed, 1126 passed, 114 skipped, 49 warnings in 498.19s (0:08:18) =====
Using TensorFlow backend.
The command "if [[ "$MODE" == "INTEGRATION_TESTS" ]]; then PYTHONPATH=$PWD:$PYTHONPATH py.test tests/integration_tests; elif [[ "$MODE" == "PEP8_DOC" ]]; then PYTHONPATH=$PWD:$PYTHONPATH py.test --pep8 -m pep8 -n0 && py.test tests/docs; elif [[ "$MODE" == "API" ]]; then PYTHONPATH=$PWD:$PYTHONPATH pip install git+git://www.github.com/keras-team/keras.git && python update_api.py && pip install -e .[tests] --progress-bar off && py.test tests/test_api.py; elif [[ "$RUN_ONLY_BACKEND_TESTS" == "1" ]]; then PYTHONPATH=$PWD:$PYTHONPATH py.test  tests/keras/backend/; else PYTHONPATH=$PWD:$PYTHONPATH py.test tests/ --ignore=tests/integration_tests --ignore=tests/docs --ignore=tests/keras/legacy/layers_test.py --ignore=tests/test_api.py --cov-config .coveragerc --cov=keras tests/; fi" exited with 1.
cache.2
store build cache
Done. Your build exited with 1.

Maybe it's a flake and re-running will pass?

[mcarbajo]

It could. Clearly the error is not related to the small change made

[zachmayer]

@fchollet How do we re-run the tests on this PR? It seems like the test failure is a flake, but we are not sure.

[amueller]

Any news on this?

[zachmayer]

@amueller I think we just need to get someone from keras to re-run tests on this PR.

@fchollet Can you re-trigger travis on this PR, so we can get passing tests?

[zachmayer]

This is a silly question, but how would I checkout this branch locally?

[cmarmo]

Hi all, sorry for stepping in, I was investigating the scikit-learn issue 17022 and that led me here. I have checked out this branch: for me it fixes scikit-learn 17022 and scikit-learn issue 15722 opened by @mcarbajo himself.

This is a silly question, but how would I checkout this branch locally?

@zachmayer, here the command line procedure I used (I beg your pardon if I misunderstood your question)

$ git remote add mcarbajo https://github.com/mcarbajo/keras.git
$ git fetch mcarbajo
$ git checkout -b mcarbajopatch -t mcarbajo/patch-1 

Hope it could help to finalize this. Thanks for listening.

[zachmayer]

@cmarmo thanks! I'm gonna try to make a new PR sometime and see if I can get this test to pass.

[johannesjung]

Any updates on this? @zachmayer, did you create a new PR?

[zachmayer]

I still can't make my own PR:

Mayer-2:keras zach2$ git checkout -b mcarbajopatch -t mcarbajo/patch-1 
fatal: 'mcarbajo/patch-1' is not a commit and a branch 'mcarbajopatch' cannot be created from it

[cmarmo]

@zachmayer do you mind checking that you have run git fetch mcarbajo first? Thanks.

[zachmayer]

I did run git fetch mcarbajo first. I just tried again, same error. gh pr checkout 13598 worked for me though

[zachmayer]

ok, the conflict in keras/wrappers/scikit_learn.py is gonna be hard to fix lol. The whole file is now just from tensorflow.keras.wrappers.scikit_learn import *

@fchollet Should we stop making PRs to the keras repo and try to fix this in the tensorflow repo?

[zachmayer]

https://github.com/keras-team/keras/blob/master/keras/wrappers/scikit_learn.py

[zachmayer]

I'll try making a PR here: https://github.com/tensorflow/tensorflow/blob/master/tensorflow/python/keras/wrappers/scikit_learn.py

[zachmayer]

I made a PR to tensorflow here: tensorflow/tensorflow#41341