Andrei Ivanov

Fixing one of the `captum` tests.

2

The current PR provides the fix for the following 9 failing tests. ``` FAILED test/explain/algorithm/test_captum_explainer.py::test_captum_explainer_multiclass_classification[index1-ModelTaskLevel.node-MaskType.object-MaskType.attributes-ShapleyValueSampling] - RuntimeError: shape '[-1, 2, 1, 1]' is invalid for input of size 3 FAILED...

[Test] Fixing issues related to `SampledHeteroCSC` and `StaticHeteroCSC` deprecation.

14

## Description As we can see [here](https://github.com/pyg-team/pytorch_geometric/issues/8570): ``` test/nn/conv/cugraph/test_cugraph_rgcn_conv.py: 72 warnings /usr/local/lib/python3.10/dist-packages/pylibcugraphops/pytorch/graph.py:302: DeprecationWarning: SampledHeteroCSC is deprecated with the 23.08 release and will be removed in 23.10, use HeteroCSC instead. warnings.warn(...

[UnitTest] Fixing AttributeError: module 'numpy' has no attribute `exceptions`

5

## Description This PR fixes a bug: ``` Line 3180: =================================== FAILURES =================================== Line 3181: _________________ test_global_uniform_negative_sampling[int32] _________________ Line 3247: _________________ test_global_uniform_negative_sampling[int64] _________________ Line 3255: tests/python/common/sampling/test_sampling.py:1777: Line 3256: _ _...

[DeprecationWarning] Resolving warnings caused by deprecated function usage.

8

## Description This PR provides a fix for the following warnings: ``` /opt/dgl/dgl-source/tools/distgraphlaunch.py:133: DeprecationWarning: setDaemon() is deprecated, set the daemon attribute instead thread.setDaemon(True) ``` ## Checklist Please feel free to...

[FutureWarning] Resolving warnings caused by deprecated function usage.

14

## Description This PR provides a fix for the following warnings: ``` /usr/local/lib/python3.10/dist-packages/dgl/backend/pytorch/sparse.py:148: FutureWarning: `torch.cuda.amp.autocast(args...)` is deprecated. Please use `torch.amp.autocast('cuda', args...)` instead. return th.cuda.amp.autocast(enabled=False) /usr/local/lib/python3.10/dist-packages/dgl/backend/pytorch/sparse.py:157: FutureWarning: `torch.cuda.amp.autocast_mode._cast(value, dtype)` is deprecated....

[FutureWarning] Remove the usage of the deprecated parameter.

13

## Description By removing the deprecated `multi_class` parameter in `LogisticRegression(...)` calls, as indicated by following warnings, ``` /usr/local/lib/python3.10/dist-packages/sklearn/linear_model/_logistic.py:1247: FutureWarning: 'multi_class' was deprecated in version 1.5 and will be removed in...

[GRAPHBOLT] Activation of several `graphbolt` tests for CPU

10

## Description The few existing `graphbolt` have been modified to run on the CPU as well. ## Checklist Please feel free to remove inapplicable items for your PR. - [x]...

[Bug] Preventing side effects in distributed tests by destroying process group …

1

in `test_molecule_gpt_datase`. This PR addresses a test isolation issue by explicitly destroying the distributed process group at the end of the `test_molecule_gpt_dataset`. This is necessary because: The test initializes a...

This PR fixes a DDP runtime error triggered on some machines during multi-GPU training of the GITMol model, caused by conditionally unused parameters in the model's forward pass. 🧠 Root...

[Bug] Restoring the `TensorAttr.fully_specify` method.

5

The `FeatureStore.__setitem__` method was modified in [PR#9782](https://github.com/pyg-team/pytorch_geometric/pull/9782) Previously, it used `key.fully_specify()` : ``` def __setitem__(self, key: TensorAttr, value: FeatureTensorType): key = self._tensor_attr_cls.cast(key) # We need to fully-specify the key for...