pytorch_geometric · Issues· 1337 open
Open on GitHubLocally synced open issues (discussions stay on GitHub)
- #10815
unbatch_edge_index drops trailing edgeless graphs when batch_size is inferred
Updated Sep 16, 2026 - #8257
Request code to implement XENet paper
featuregood first issuehelp wantedUpdated Sep 10, 2026 - #9660
27 tests fail
buggood first issuetestUpdated Sep 8, 2026 - #10801
`Batch.from_data_list` only store node/edge types of first element for `HeteroData`
documentationUpdated Sep 7, 2026 - #10720
Notebook `5. Point Cloud Classification` still uses `torch-cluster` instead of `pyg-lib`
bugUpdated Sep 6, 2026 - #10783
APPNP applies edge dropout cumulatively across propagation steps
bugUpdated Aug 28, 2026 - #8258
Add `EGT` to `nn.conv`
featuregood first issuehelp wantedUpdated Aug 26, 2026 - #10772
Avoid importing the `llm` subtree (and pandas) on `import torch_geometric`
Updated Aug 16, 2026 - #8330
Add `channel_list` to `BasicGNN`
good first issuehelp wantedrefactorUpdated Aug 10, 2026 - #10768
HeteroData.subgraph() crashes for edge attributes with non-zero concatenation dimensions
Updated Aug 9, 2026 - #5773
Dataset `TOSCA` url broken
bugUpdated Aug 4, 2026 - #10204
SetTransformerAggregation raises errors in eval mode on CUDA (but not CPU)
bugUpdated Jul 30, 2026 - #10517
qm9_nn_conv.py yields very large and inconsistent results
bugUpdated Jul 27, 2026 - #7499
Laplacian Eigenvector PE is not deterministic
bughelp wantedtransformdeterminismUpdated Jul 22, 2026 - #10597
HeteroData silently accepts numpy arrays. Then fails during training with cryptic message
bugUpdated Jun 26, 2026