[pytorch] Advanced indexing that supports all indexing features on PyTorch

[KexinFeng]

Description

Full support of pytorch indexing.

This PR enables all indexing features in pytorch. The indexing behaviour is guaranteed to be consistent, especially when mixed indices are passed.

Demo example

See the following code for demo

djl/integration/src/main/java/ai/djl/integration/tests/ndarray/NDIndexTest.java

Lines 153 to 178 in f040c38

	// get from integer array (higher rank included) or float array
	original = manager.arange(1, 7f).reshape(-1, 2);
	NDArray index = manager.create(new long[] {0, 0, 1, 2}, new Shape(2, 2));
	NDArray indexFloat = manager.create(new float[] {0, 0, 1, 2}, new Shape(2, 2));
	NDArray actual = original.get(index);
	NDArray actual2 = original.get(indexFloat);
	expected = manager.create(new float[] {1, 2, 1, 2, 3, 4, 5, 6}, new Shape(2, 2, 2));
	Assert.assertEquals(actual, expected);
	Assert.assertEquals(actual2, expected);
	// indexing with boolean, slice, and integer array (higher rank included) or float array
	original = manager.arange(3 * 3 * 3 * 3).reshape(3, 3, 3, 3);
	NDArray bool1 = manager.create(new boolean[] {true, false, true});
	NDArray index1 = manager.create(new long[] {2, 2}, new Shape(1, 2));
	NDArray index2 = manager.create(new float[] {0, 1}, new Shape(1, 2));
	actual = original.get(":{}, {}, {}, {}", 2, index1, bool1, index2);
	expected = manager.create(new int[] {18, 25, 45, 52}, new Shape(2, 1, 2));
	Assert.assertEquals(actual, expected);
	// indexing with null, slice and integer array (higher rank included) or float array
	original = manager.arange(3 * 3 * 3).reshape(3, 3, 3);
	index1 = manager.create(new float[] {0, 1}, new Shape(2));
	index2 = manager.create(new long[] {0, 0, 2, 1}, new Shape(2, 2));
	actual = original.get(":{}, {}, {}, {}", 2, index1, index2, null);
	expected = manager.create(new int[] {0, 3, 2, 4, 9, 12, 11, 13}, new Shape(2, 2, 2, 1));
	Assert.assertEquals(actual, expected);

The pytorch doc:

C indexing API
source

The relevant preceeding PRs:

Support of take from pytorch #1627
Add support of take on MXNet engine #1649

[codecov-commenter]

Codecov Report

Merging #1719 (3f0c4e8) into master (bb5073f) will decrease coverage by 1.46%.
The diff coverage is 70.41%.

@@             Coverage Diff              @@
##             master    #1719      +/-   ##
============================================
- Coverage     72.08%   70.62%   -1.47%     
- Complexity     5126     5562     +436     
============================================
  Files           473      527      +54     
  Lines         21970    24691    +2721     
  Branches       2351     2680     +329     
============================================
+ Hits          15838    17438    +1600     
- Misses         4925     5922     +997     
- Partials       1207     1331     +124     

Impacted Files	Coverage Δ
api/src/main/java/ai/djl/modality/cv/Image.java	69.23% <ø> (-4.11%)	⬇️
...rc/main/java/ai/djl/modality/cv/MultiBoxPrior.java	76.00% <ø> (ø)
...rc/main/java/ai/djl/modality/cv/output/Joints.java	71.42% <ø> (ø)
.../main/java/ai/djl/modality/cv/output/Landmark.java	100.00% <ø> (ø)
...main/java/ai/djl/modality/cv/output/Rectangle.java	72.41% <0.00%> (ø)
...i/djl/modality/cv/translator/BigGANTranslator.java	21.42% <ø> (-5.24%)	⬇️
...odality/cv/translator/BigGANTranslatorFactory.java	33.33% <0.00%> (+8.33%)	⬆️
...nslator/InstanceSegmentationTranslatorFactory.java	14.28% <0.00%> (-3.90%)	⬇️
.../cv/translator/StyleTransferTranslatorFactory.java	40.00% <ø> (ø)
.../ai/djl/modality/cv/translator/YoloTranslator.java	8.33% <0.00%> (-0.50%)	⬇️
... and 410 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 a53acb2...3f0c4e8. Read the comment docs.