Refactoring deep learning models to work with any explainer

[amit-sharma]

[READY TO MERGE] Refactors deep learning model and explainers so that they are compatible with the sklearn explainers. sklearn explainers do not need access to gradient, so they should also work for deep learning models.

• DL and sklearn models need to have the same output
• DL methods employ user-provided encoding of features
• model needs to handle the conversion from raw to one hot encoded data
• Remove any native calls to DL model's methods
• Updated tests, added torch as a test dependency

Possible fix to #246

[github-actions[bot]]

Unit Test Results

261 tests  +14   256 :heavy_check_mark: +14   19m 21s :stopwatch: + 8m 27s     1 suites ±  0       5 :zzz: ±  0      1 files   ±  0       0 :x: ±  0 

Results for commit d0127b47. ± Comparison against base commit 2e9d093c.

This pull request removes 29 and adds 43 tests. Note that renamed tests count towards both.

tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_features_to_vary[1-2-features_to_vary0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_features_to_vary[1-2-features_to_vary1-random]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_invalid_query_instance[features_to_vary0-permitted_range0-inverse_mad]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_maxiter[0-7-kdtree-0]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_maxiter[0-7-random-0]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_no_cfs[7-3]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range[1-2-all-permitted_range0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range[1-2-all-permitted_range1-random]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range_categorical[1-2-all-permitted_range0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range_categorical[1-2-all-permitted_range1-random]
…

tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_features_to_vary[sklearn-1-2-features_to_vary0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_features_to_vary[sklearn-1-2-features_to_vary1-random]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_invalid_query_instance[sklearn-features_to_vary0-permitted_range0-inverse_mad]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_maxiter[sklearn-0-7-kdtree-0]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_maxiter[sklearn-0-7-random-0]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_no_cfs[sklearn-7-3]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range[sklearn-1-2-all-permitted_range0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range[sklearn-1-2-all-permitted_range1-random]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range_categorical[sklearn-1-2-all-permitted_range0-kdtree]
tests.test_dice_interface.test_dice_genetic.TestDiceGeneticBinaryClassificationMethods ‑ test_permitted_range_categorical[sklearn-1-2-all-permitted_range1-random]
…

:recycle: This comment has been updated with latest results.

[codecov-commenter]

Codecov Report

Merging #261 (28dcb9e) into master (48ef38c) will increase coverage by 0.38%. The diff coverage is 87.50%.

@@            Coverage Diff             @@
##           master     #261      +/-   ##
==========================================
+ Coverage   71.51%   71.90%   +0.38%     
==========================================
  Files          26       27       +1     
  Lines        3468     3541      +73     
==========================================
+ Hits         2480     2546      +66     
- Misses        988      995       +7     

Flag	Coverage Δ
unittests	71.90% <87.50%> (+0.38%)	:arrow_up:

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
dice_ml/diverse_counterfactuals.py	84.35% <ø> (ø)
dice_ml/explainer_interfaces/dice_random.py	91.72% <ø> (-1.56%)	:arrow_down:
dice_ml/explainer_interfaces/dice_tensorflow1.py	0.00% <0.00%> (ø)
dice_ml/explainer_interfaces/dice_genetic.py	95.45% <81.81%> (-0.53%)	:arrow_down:
dice_ml/data_interfaces/private_data_interface.py	80.53% <82.75%> (+0.24%)	:arrow_up:
dice_ml/dice.py	80.95% <86.36%> (-1.55%)	:arrow_down:
dice_ml/data_interfaces/public_data_interface.py	87.03% <87.50%> (-1.90%)	:arrow_down:
dice_ml/model_interfaces/pytorch_model.py	89.65% <89.65%> (+0.18%)	:arrow_up:
dice_ml/data_interfaces/base_data_interface.py	93.33% <96.42%> (+5.09%)	:arrow_up:
dice_ml/explainer_interfaces/explainer_base.py	89.94% <96.42%> (+0.19%)	:arrow_up:
... and 13 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more Δ = absolute <relative> (impact), ø = not affected, ? = missing data Powered by Codecov. Last update 48ef38c...28dcb9e. Read the comment docs.