Added
- Added the test for `DeepGCNLayer` ([5704](https://github.com/pyg-team/pytorch_geometric/pull/5704))
- Allow `.` in `ModuleDict` key names ([5227](https://github.com/pyg-team/pytorch_geometric/pull/5227))
- Added `edge_label_time` argument to `LinkNeighborLoader` ([5137](https://github.com/pyg-team/pytorch_geometric/pull/5137), [#5173](https://github.com/pyg-team/pytorch_geometric/pull/5173))
- Let `ImbalancedSampler` accept `torch.Tensor` as input ([5138](https://github.com/pyg-team/pytorch_geometric/pull/5138))
- Added `flow` argument to `gcn_norm` to correctly normalize the adjacency matrix in `GCNConv` ([5149](https://github.com/pyg-team/pytorch_geometric/pull/5149))
- `NeighborSampler` supports graphs without edges ([5072](https://github.com/pyg-team/pytorch_geometric/pull/5072))
- Added the `MeanSubtractionNorm` layer ([5068](https://github.com/pyg-team/pytorch_geometric/pull/5068))
- Added `pyg_lib.segment_matmul` integration within `RGCNConv` ([5052](https://github.com/pyg-team/pytorch_geometric/pull/5052), [#5096](https://github.com/pyg-team/pytorch_geometric/pull/5096))
- Support `SparseTensor` as edge label in `LightGCN` ([5046](https://github.com/pyg-team/pytorch_geometric/issues/5046))
- Added support for `BasicGNN` models within `to_hetero` ([5091](https://github.com/pyg-team/pytorch_geometric/pull/5091))
- Added support for computing weighted metapaths in `AddMetapaths` ([5049](https://github.com/pyg-team/pytorch_geometric/pull/5049))
- Added inference benchmark suite ([4915](https://github.com/pyg-team/pytorch_geometric/pull/4915))
- Added a dynamically sized batch sampler for filling a mini-batch with a variable number of samples up to a maximum size ([4972](https://github.com/pyg-team/pytorch_geometric/pull/4972))
- Added fine grained options for setting `bias` and `dropout` per layer in the `MLP` model ([4981](https://github.com/pyg-team/pytorch_geometric/pull/4981))
- Added `EdgeCNN` model ([4991](https://github.com/pyg-team/pytorch_geometric/pull/4991))
- Added scalable `inference` mode in `BasicGNN` with layer-wise neighbor loading ([4977](https://github.com/pyg-team/pytorch_geometric/pull/4977))
- Added inference benchmarks ([4892](https://github.com/pyg-team/pytorch_geometric/pull/4892), [#5107](https://github.com/pyg-team/pytorch_geometric/pull/5107))
- Added PyTorch 1.12 support ([4975](https://github.com/pyg-team/pytorch_geometric/pull/4975))
- Added `unbatch_edge_index` functionality for splitting an `edge_index` tensor according to a `batch` vector ([4903](https://github.com/pyg-team/pytorch_geometric/pull/4903))
- Added node-wise normalization mode in `LayerNorm` ([4944](https://github.com/pyg-team/pytorch_geometric/pull/4944))
- Added support for `normalization_resolver` ([4926](https://github.com/pyg-team/pytorch_geometric/pull/4926), [#4951](https://github.com/pyg-team/pytorch_geometric/pull/4951), [#4958](https://github.com/pyg-team/pytorch_geometric/pull/4958), [#4959](https://github.com/pyg-team/pytorch_geometric/pull/4959))
- Added notebook tutorial for `torch_geometric.nn.aggr` package to documentation ([4927](https://github.com/pyg-team/pytorch_geometric/pull/4927))
- Added support for `follow_batch` for lists or dictionaries of tensors ([4837](https://github.com/pyg-team/pytorch_geometric/pull/4837))
- Added `Data.validate()` and `HeteroData.validate()` functionality ([4885](https://github.com/pyg-team/pytorch_geometric/pull/4885))
- Added `LinkNeighborLoader` support to `LightningDataModule` ([4868](https://github.com/pyg-team/pytorch_geometric/pull/4868))
- Added `predict()` support to the `LightningNodeData` module ([4884](https://github.com/pyg-team/pytorch_geometric/pull/4884))
- Added `time_attr` argument to `LinkNeighborLoader` ([4877](https://github.com/pyg-team/pytorch_geometric/pull/4877), [#4908](https://github.com/pyg-team/pytorch_geometric/pull/4908))
- Added a `filter_per_worker` argument to data loaders to allow filtering of data within sub-processes ([4873](https://github.com/pyg-team/pytorch_geometric/pull/4873))
- Added a `NeighborLoader` benchmark script ([4815](https://github.com/pyg-team/pytorch_geometric/pull/4815), [#4862](https://github.com/pyg-team/pytorch_geometric/pull/4862/files))
- Added support for `FeatureStore` and `GraphStore` in `NeighborLoader` ([4817](https://github.com/pyg-team/pytorch_geometric/pull/4817), [#4851](https://github.com/pyg-team/pytorch_geometric/pull/4851), [#4854](https://github.com/pyg-team/pytorch_geometric/pull/4854), [#4856](https://github.com/pyg-team/pytorch_geometric/pull/4856), [#4857](https://github.com/pyg-team/pytorch_geometric/pull/4857), [#4882](https://github.com/pyg-team/pytorch_geometric/pull/4882), [#4883](https://github.com/pyg-team/pytorch_geometric/pull/4883), [#4929](https://github.com/pyg-team/pytorch_geometric/pull/4929), [#4992](https://github.com/pyg-team/pytorch_geometric/pull/4922), [#4962](https://github.com/pyg-team/pytorch_geometric/pull/4962), [#4968](https://github.com/pyg-team/pytorch_geometric/pull/4968), [#5037](https://github.com/pyg-team/pytorch_geometric/pull/5037), [#5088](https://github.com/pyg-team/pytorch_geometric/pull/5088), [#5270](https://github.com/pyg-team/pytorch_geometric/pull/5270), [#5307](https://github.com/pyg-team/pytorch_geometric/pull/5307), [#5318](https://github.com/pyg-team/pytorch_geometric/pull/5318))
- Added a `normalize` parameter to `dense_diff_pool` ([4847](https://github.com/pyg-team/pytorch_geometric/pull/4847))
- Added `size=None` explanation to jittable `MessagePassing` modules in the documentation ([4850](https://github.com/pyg-team/pytorch_geometric/pull/4850))
- Added documentation to the `DataLoaderIterator` class ([4838](https://github.com/pyg-team/pytorch_geometric/pull/4838))
- Added `GraphStore` support to `Data` and `HeteroData` ([4816](https://github.com/pyg-team/pytorch_geometric/pull/4816))
- Added `FeatureStore` support to `Data` and `HeteroData` ([4807](https://github.com/pyg-team/pytorch_geometric/pull/4807), [#4853](https://github.com/pyg-team/pytorch_geometric/pull/4853))
- Added `FeatureStore` and `GraphStore` abstractions ([4534](https://github.com/pyg-team/pytorch_geometric/pull/4534), [#4568](https://github.com/pyg-team/pytorch_geometric/pull/4568), [#5120](https://github.com/pyg-team/pytorch_geometric/pull/5120))
- Added support for dense aggregations in `global_*_pool` ([4827](https://github.com/pyg-team/pytorch_geometric/pull/4827))
- Added Python version requirement ([4825](https://github.com/pyg-team/pytorch_geometric/pull/4825))
- Added TorchScript support to `JumpingKnowledge` module ([4805](https://github.com/pyg-team/pytorch_geometric/pull/4805))
- Added a `max_sample` argument to `AddMetaPaths` in order to tackle very dense metapath edges ([4750](https://github.com/pyg-team/pytorch_geometric/pull/4750))
- Test `HANConv` with empty tensors ([4756](https://github.com/pyg-team/pytorch_geometric/pull/4756), [#4841](https://github.com/pyg-team/pytorch_geometric/pull/4841))
- Added the `bias` vector to the `GCN` model definition in the "Create Message Passing Networks" tutorial ([4755](https://github.com/pyg-team/pytorch_geometric/pull/4755))
- Added `transforms.RootedSubgraph` interface with two implementations: `RootedEgoNets` and `RootedRWSubgraph` ([3926](https://github.com/pyg-team/pytorch_geometric/pull/3926))
- Added `ptr` vectors for `follow_batch` attributes within `Batch.from_data_list` ([4723](https://github.com/pyg-team/pytorch_geometric/pull/4723))
- Added `torch_geometric.nn.aggr` package ([4687](https://github.com/pyg-team/pytorch_geometric/pull/4687), [#4721](https://github.com/pyg-team/pytorch_geometric/pull/4721), [#4731](https://github.com/pyg-team/pytorch_geometric/pull/4731), [#4762](https://github.com/pyg-team/pytorch_geometric/pull/4762), [#4749](https://github.com/pyg-team/pytorch_geometric/pull/4749), [#4779](https://github.com/pyg-team/pytorch_geometric/pull/4779), [#4863](https://github.com/pyg-team/pytorch_geometric/pull/4863), [#4864](https://github.com/pyg-team/pytorch_geometric/pull/4864), [#4865](https://github.com/pyg-team/pytorch_geometric/pull/4865), [#4866](https://github.com/pyg-team/pytorch_geometric/pull/4866), [#4872](https://github.com/pyg-team/pytorch_geometric/pull/4872), [#4934](https://github.com/pyg-team/pytorch_geometric/pull/4934), [#4935](https://github.com/pyg-team/pytorch_geometric/pull/4935), [#4957](https://github.com/pyg-team/pytorch_geometric/pull/4957), [#4973](https://github.com/pyg-team/pytorch_geometric/pull/4973), [#4973](https://github.com/pyg-team/pytorch_geometric/pull/4973), [#4986](https://github.com/pyg-team/pytorch_geometric/pull/4986), [#4995](https://github.com/pyg-team/pytorch_geometric/pull/4995), [#5000](https://github.com/pyg-team/pytorch_geometric/pull/5000), [#5034](https://github.com/pyg-team/pytorch_geometric/pull/5034), [#5036](https://github.com/pyg-team/pytorch_geometric/pull/5036), [#5039](https://github.com/pyg-team/pytorch_geometric/issues/5039), [#4522](https://github.com/pyg-team/pytorch_geometric/pull/4522), [#5033](https://github.com/pyg-team/pytorch_geometric/pull/5033), [#5085](https://github.com/pyg-team/pytorch_geometric/pull/5085), [#5097](https://github.com/pyg-team/pytorch_geometric/pull/5097), [#5099](https://github.com/pyg-team/pytorch_geometric/pull/5099), [#5104](https://github.com/pyg-team/pytorch_geometric/pull/5104), [#5113](https://github.com/pyg-team/pytorch_geometric/pull/5113), [#5130](https://github.com/pyg-team/pytorch_geometric/pull/5130), [#5098](https://github.com/pyg-team/pytorch_geometric/pull/5098), [#5191](https://github.com/pyg-team/pytorch_geometric/pull/5191))
- Added the `DimeNet++` model ([4432](https://github.com/pyg-team/pytorch_geometric/pull/4432), [#4699](https://github.com/pyg-team/pytorch_geometric/pull/4699), [#4700](https://github.com/pyg-team/pytorch_geometric/pull/4700), [#4800](https://github.com/pyg-team/pytorch_geometric/pull/4800))
- Added an example of using PyG with PyTorch Ignite ([4487](https://github.com/pyg-team/pytorch_geometric/pull/4487))
- Added `GroupAddRev` module with support for reducing training GPU memory ([4671](https://github.com/pyg-team/pytorch_geometric/pull/4671), [#4701](https://github.com/pyg-team/pytorch_geometric/pull/4701), [#4715](https://github.com/pyg-team/pytorch_geometric/pull/4715), [#4730](https://github.com/pyg-team/pytorch_geometric/pull/4730))
- Added benchmarks via [`wandb`](https://wandb.ai/site) ([#4656](https://github.com/pyg-team/pytorch_geometric/pull/4656), [#4672](https://github.com/pyg-team/pytorch_geometric/pull/4672), [#4676](https://github.com/pyg-team/pytorch_geometric/pull/4676))
- Added `unbatch` functionality ([4628](https://github.com/pyg-team/pytorch_geometric/pull/4628))
- Confirm that `to_hetero()` works with custom functions, _e.g._, `dropout_adj` ([4653](https://github.com/pyg-team/pytorch_geometric/pull/4653))
- Added the `MLP.plain_last=False` option ([4652](https://github.com/pyg-team/pytorch_geometric/pull/4652))
- Added a check in `HeteroConv` and `to_hetero()` to ensure that `MessagePassing.add_self_loops` is disabled ([4647](https://github.com/pyg-team/pytorch_geometric/pull/4647))
- Added `HeteroData.subgraph()`, `HeteroData.node_type_subgraph()` and `HeteroData.edge_type_subgraph()` support ([4635](https://github.com/pyg-team/pytorch_geometric/pull/4635))
- Added the `AQSOL` dataset ([4626](https://github.com/pyg-team/pytorch_geometric/pull/4626))
- Added `HeteroData.node_items()` and `HeteroData.edge_items()` functionality ([4644](https://github.com/pyg-team/pytorch_geometric/pull/4644))
- Added PyTorch Lightning support in GraphGym ([4511](https://github.com/pyg-team/pytorch_geometric/pull/4511), [#4516](https://github.com/pyg-team/pytorch_geometric/pull/4516) [#4531](https://github.com/pyg-team/pytorch_geometric/pull/4531), [#4689](https://github.com/pyg-team/pytorch_geometric/pull/4689), [#4843](https://github.com/pyg-team/pytorch_geometric/pull/4843))
- Added support for returning embeddings in `MLP` models ([4625](https://github.com/pyg-team/pytorch_geometric/pull/4625))
- Added faster initialization of `NeighborLoader` in case edge indices are already sorted (via `is_sorted=True`) ([4620](https://github.com/pyg-team/pytorch_geometric/pull/4620), [#4702](https://github.com/pyg-team/pytorch_geometric/pull/4702))
- Added `AddPositionalEncoding` transform ([4521](https://github.com/pyg-team/pytorch_geometric/pull/4521))
- Added `HeteroData.is_undirected()` support ([4604](https://github.com/pyg-team/pytorch_geometric/pull/4604))
- Added the `Genius` and `Wiki` datasets to `nn.datasets.LINKXDataset` ([4570](https://github.com/pyg-team/pytorch_geometric/pull/4570), [#4600](https://github.com/pyg-team/pytorch_geometric/pull/4600))
- Added `nn.aggr.EquilibrumAggregation` implicit global layer ([4522](https://github.com/pyg-team/pytorch_geometric/pull/4522))
- Added support for graph-level outputs in `to_hetero` ([4582](https://github.com/pyg-team/pytorch_geometric/pull/4582))
- Added `CHANGELOG.md` ([4581](https://github.com/pyg-team/pytorch_geometric/pull/4581))
- Added `HeteroData` support to the `RemoveIsolatedNodes` transform ([4479](https://github.com/pyg-team/pytorch_geometric/pull/4479))
- Added `HeteroData.num_features` functionality ([4504](https://github.com/pyg-team/pytorch_geometric/pull/4504))
- Added support for projecting features before propagation in `SAGEConv` ([4437](https://github.com/pyg-team/pytorch_geometric/pull/4437))
- Added `Geom-GCN` splits to the `Planetoid` datasets ([4442](https://github.com/pyg-team/pytorch_geometric/pull/4442))
- Added a `LinkNeighborLoader` for training scalable link predictions models [4396](https://github.com/pyg-team/pytorch_geometric/pull/4396), [#4439](https://github.com/pyg-team/pytorch_geometric/pull/4439), [#4441](https://github.com/pyg-team/pytorch_geometric/pull/4441), [#4446](https://github.com/pyg-team/pytorch_geometric/pull/4446), [#4508](https://github.com/pyg-team/pytorch_geometric/pull/4508), [#4509](https://github.com/pyg-team/pytorch_geometric/pull/4509))
- Added an unsupervised `GraphSAGE` example on `PPI` ([4416](https://github.com/pyg-team/pytorch_geometric/pull/4416))
- Added support for `LSTM` aggregation in `SAGEConv` ([4379](https://github.com/pyg-team/pytorch_geometric/pull/4379))
- Added support for floating-point labels in `RandomLinkSplit` ([4311](https://github.com/pyg-team/pytorch_geometric/pull/4311), [#4383](https://github.com/pyg-team/pytorch_geometric/pull/4383))
- Added support for `torch.data` `DataPipes` ([4302](https://github.com/pyg-team/pytorch_geometric/pull/4302), [#4345](https://github.com/pyg-team/pytorch_geometric/pull/4345), [#4349](https://github.com/pyg-team/pytorch_geometric/pull/4349))
- Added support for the `cosine` argument in the `KNNGraph`/`RadiusGraph` transforms ([4344](https://github.com/pyg-team/pytorch_geometric/pull/4344))
- Added support graph-level attributes in `networkx` conversion ([4343](https://github.com/pyg-team/pytorch_geometric/pull/4343))
- Added support for renaming node types via `HeteroData.rename` ([4329](https://github.com/pyg-team/pytorch_geometric/pull/4329))
- Added an example to load a trained PyG model in C++ ([4307](https://github.com/pyg-team/pytorch_geometric/pull/4307))
- Added a `MessagePassing.explain_message` method to customize making explanations on messages ([4278](https://github.com/pyg-team/pytorch_geometric/pull/4278), [#4448](https://github.com/pyg-team/pytorch_geometric/pull/4448)))
- Added support for `GATv2Conv` in the `nn.models.GAT` model ([4357](https://github.com/pyg-team/pytorch_geometric/pull/4357))
- Added `HeteroData.subgraph` functionality ([4243](https://github.com/pyg-team/pytorch_geometric/pull/4243))
- Added the `MaskLabel` module and a corresponding masked label propagation example ([4197](https://github.com/pyg-team/pytorch_geometric/pull/4197))
- Added temporal sampling support to `NeighborLoader` ([4025](https://github.com/pyg-team/pytorch_geometric/pull/4025))
- Added an example for unsupervised heterogeneous graph learning based on "Deep Multiplex Graph Infomax" ([3189](https://github.com/pyg-team/pytorch_geometric/pull/3189))
Changed
- Changed docstring for `RandomLinkSplit` ([5190](https://github.com/pyg-team/pytorch_geometric/issues/5190))
- Switched to PyTorch `scatter_reduce` implementation - experimental feature ([5120](https://github.com/pyg-team/pytorch_geometric/pull/5120))
- Fixed `RGATConv` device mismatches for `f-scaled` mode ([5187](https://github.com/pyg-team/pytorch_geometric/pull/5187))
- Allow for multi-dimensional `edge_labels` in `LinkNeighborLoader` ([5186](https://github.com/pyg-team/pytorch_geometric/pull/5186))
- Fixed `GINEConv` bug with non-sequential input ([5154](https://github.com/pyg-team/pytorch_geometric/pull/5154))
- Improved error message ([5095](https://github.com/pyg-team/pytorch_geometric/pull/5095))
- Fixed `HGTLoader` bug which produced outputs with missing edge types ([5067](https://github.com/pyg-team/pytorch_geometric/pull/5067))
- Fixed dynamic inheritance issue in data batching ([5051](https://github.com/pyg-team/pytorch_geometric/pull/5051))
- Fixed `load_state_dict` in `Linear` with `strict=False` mode ([5094](https://github.com/pyg-team/pytorch_geometric/pull/5094))
- Fixed typo in `MaskLabel.ratio_mask` ([5093](https://github.com/pyg-team/pytorch_geometric/pull/5093))
- Fixed `data.num_node_features` computation for sparse matrices ([5089](https://github.com/pyg-team/pytorch_geometric/pull/5089))
- Fixed `torch.fx` bug with `torch.nn.aggr` package ([5021](https://github.com/pyg-team/pytorch_geometric/pull/5021)))
- Fixed `GenConv` test ([4993](https://github.com/pyg-team/pytorch_geometric/pull/4993))
- Fixed packaging tests for Python 3.10 ([4982](https://github.com/pyg-team/pytorch_geometric/pull/4982))
- Changed `act_dict` (part of `graphgym`) to create individual instances instead of reusing the same ones everywhere ([4978](https://github.com/pyg-team/pytorch_geometric/pull/4978))
- Fixed issue where one-hot tensors were passed to `F.one_hot` ([4970](https://github.com/pyg-team/pytorch_geometric/pull/4970))
- Fixed `bool` arugments in `argparse` in `benchmark/` ([4967](https://github.com/pyg-team/pytorch_geometric/pull/4967))
- Fixed `BasicGNN` for `num_layers=1`, which now respects a desired number of `out_channels` ([4943](https://github.com/pyg-team/pytorch_geometric/pull/4943))
- `len(batch)` will now return the number of graphs inside the batch, not the number of attributes ([4931](https://github.com/pyg-team/pytorch_geometric/pull/4931))
- Fixed `data.subgraph` generation for 0-dim tensors ([4932](https://github.com/pyg-team/pytorch_geometric/pull/4932))
- Removed unnecssary inclusion of self-loops when sampling negative edges ([4880](https://github.com/pyg-team/pytorch_geometric/pull/4880))
- Fixed `InMemoryDataset` inferring wrong `len` for lists of tensors ([4837](https://github.com/pyg-team/pytorch_geometric/pull/4837))
- Fixed `Batch.separate` when using it for lists of tensors ([4837](https://github.com/pyg-team/pytorch_geometric/pull/4837))
- Correct docstring for SAGEConv ([4852](https://github.com/pyg-team/pytorch_geometric/pull/4852))
- Fixed a bug in `TUDataset` where `pre_filter` was not applied whenever `pre_transform` was present
- Renamed `RandomTranslate` to `RandomJitter` - the usage of `RandomTranslate` is now deprecated ([4828](https://github.com/pyg-team/pytorch_geometric/pull/4828))
- Do not allow accessing edge types in `HeteroData` with two node types when there exists multiple relations between these types ([4782](https://github.com/pyg-team/pytorch_geometric/pull/4782))
- Allow `edge_type == rev_edge_type` argument in `RandomLinkSplit` ([4757](https://github.com/pyg-team/pytorch_geometric/pull/4757), [#5221](https://github.com/pyg-team/pytorch_geometric/pull/5221))
- Fixed a numerical instability in the `GeneralConv` and `neighbor_sample` tests ([4754](https://github.com/pyg-team/pytorch_geometric/pull/4754))
- Fixed a bug in `HANConv` in which destination node features rather than source node features were propagated ([4753](https://github.com/pyg-team/pytorch_geometric/pull/4753))
- Fixed versions of `checkout` and `setup-python` in CI ([4751](https://github.com/pyg-team/pytorch_geometric/pull/4751))
- Fixed `protobuf` version ([4719](https://github.com/pyg-team/pytorch_geometric/pull/4719))
- Fixed the ranking protocol bug in the RGCN link prediction example ([4688](https://github.com/pyg-team/pytorch_geometric/pull/4688))
- Math support in Markdown ([4683](https://github.com/pyg-team/pytorch_geometric/pull/4683))
- Allow for `setter` properties in `Data` ([4682](https://github.com/pyg-team/pytorch_geometric/pull/4682), [#4686](https://github.com/pyg-team/pytorch_geometric/pull/4686))
- Allow for optional `edge_weight` in `GCN2Conv` ([4670](https://github.com/pyg-team/pytorch_geometric/pull/4670))
- Fixed the interplay between `TUDataset` and `pre_transform` that modify node features ([4669](https://github.com/pyg-team/pytorch_geometric/pull/4669))
- Make use of the `pyg_sphinx_theme` documentation template ([4664](https://github.com/pyg-team/pyg-lib/pull/4664), [#4667](https://github.com/pyg-team/pyg-lib/pull/4667))
- Refactored reading molecular positions from sdf file for qm9 datasets ([4654](https://github.com/pyg-team/pytorch_geometric/pull/4654))
- Fixed `MLP.jittable()` bug in case `return_emb=True` ([4645](https://github.com/pyg-team/pytorch_geometric/pull/4645), [#4648](https://github.com/pyg-team/pytorch_geometric/pull/4648))
- The generated node features of `StochasticBlockModelDataset` are now ordered with respect to their labels ([4617](https://github.com/pyg-team/pytorch_geometric/pull/4617))
- Fixed typos in the documentation ([4616](https://github.com/pyg-team/pytorch_geometric/pull/4616), [#4824](https://github.com/pyg-team/pytorch_geometric/pull/4824), [#4895](https://github.com/pyg-team/pytorch_geometric/pull/4895), [#5161](https://github.com/pyg-team/pytorch_geometric/pull/5161))
- The `bias` argument in `TAGConv` is now actually applied ([4597](https://github.com/pyg-team/pytorch_geometric/pull/4597))
- Fixed subclass behavior of `process` and `download` in `Datsaet` ([4586](https://github.com/pyg-team/pytorch_geometric/pull/4586))
- Fixed filtering of attributes for loaders in case `__cat_dim__ != 0` ([4629](https://github.com/pyg-team/pytorch_geometric/pull/4629))
- Fixed `SparseTensor` support in `NeighborLoader` ([4320](https://github.com/pyg-team/pytorch_geometric/pull/4320))
- Fixed average degree handling in `PNAConv` ([4312](https://github.com/pyg-team/pytorch_geometric/pull/4312))
- Fixed a bug in `from_networkx` in case some attributes are PyTorch tensors ([4486](https://github.com/pyg-team/pytorch_geometric/pull/4486))
- Added a missing clamp in `DimeNet` ([4506](https://github.com/pyg-team/pytorch_geometric/pull/4506), [#4562](https://github.com/pyg-team/pytorch_geometric/pull/4562))
- Fixed the download link in `DBP15K` ([4428](https://github.com/pyg-team/pytorch_geometric/pull/4428))
- Fixed an autograd bug in `DimeNet` when resetting parameters ([4424](https://github.com/pyg-team/pytorch_geometric/pull/4424))
- Fixed bipartite message passing in case `flow="target_to_source"` ([4418](https://github.com/pyg-team/pytorch_geometric/pull/4418))
- Fixed a bug in which `num_nodes` was not properly updated in the `FixedPoints` transform ([4394](https://github.com/pyg-team/pytorch_geometric/pull/4394))
- PyTorch Lightning >= 1.6 support ([4377](https://github.com/pyg-team/pytorch_geometric/pull/4377))
- Fixed a bug in which `GATConv` was not jittable ([4347](https://github.com/pyg-team/pytorch_geometric/pull/4347))
- Fixed a bug in which the GraphGym config was not stored in each specific experiment directory ([4338](https://github.com/pyg-team/pytorch_geometric/pull/4338))
- Fixed a bug in which `nn.models.GAT` did not produce `out_channels`-many output channels ([4299](https://github.com/pyg-team/pytorch_geometric/pull/4299))
- Fixed mini-batching with empty lists as attributes ([4293](https://github.com/pyg-team/pytorch_geometric/pull/4293))
- Fixed a bug in which `GCNConv` could not be combined with `to_hetero` on heterogeneous graphs with one node type ([4279](https://github.com/pyg-team/pytorch_geometric/pull/4279))
Removed
- Remove internal metrics in favor of `torchmetrics` ([4287](https://github.com/pyg-team/pytorch_geometric/pull/4287))