pyg-team / pytorch_geometric

Graph Neural Network Library for PyTorch

https://pyg.org

pyg-team/pytorch_geometric Issues

torch_geometric InMemoryDataset doesn't load data
Closed 11 hours ago2
'import torch_geometric' fails: File already exists in database: onnx/onnx-ml.proto
Closed 11 hours ago2
pyproject.toml doesn't list as dependencies modules imported at runtime: dgl, torch_sparse
Closed a day ago2
Graph loader (NeighborLoader, HGTLoader, etc) mysteriously crash on RCDD dataset
Updated a day ago3
The link to the Karate Club paper is broken
Updated a day ago
pyproject.toml doesn't list 'torch' as a dependency
Closed 2 days ago1
Returning map_index while relabelling the nodes
Closed 3 days ago3
torch.jit.frontend.NotSupportedError: Compiled functions can't take variable number of arguments or use keyword-only arguments with defaults
Closed 4 days ago10
Trying to install pyg through conda downgrades torch to CPU
Updated 5 days ago2
loss.backward() is too slow
Updated 5 days ago1
User defined heterogeneous graphs not moving to GPU properly.
Closed 5 days ago
DataLoader Combines HeteroData into Single Graph
Closed 19 days ago11
Wrong return values for group_argsort
Updated 6 days ago
loss.backward() is too slow
Closed 7 days ago
ToSparseTensor() leads to GPU memory leakage when dealing with data generated from RandomLinkSplit()
Closed 8 days ago1
Device number causes a "CUDA error: an illegal memory access was encountered" in PNAConv
Closed 8 days ago5
torch::jit::load error
Closed 8 days ago3
enforce single quotes in `ruff` and `pre-commit`
Closed 10 days ago3
PyG Dependencies (torch-scatter, torch-sparse, etc.) are incompatible with conda-forge distribution of PyTorch
Updated 11 days ago2
SNAPDataset ram usage
Closed 11 days ago2
to_dgl does not convert graph with isolated nodes or zero edges graph
Closed 12 days ago1
Add typing_extensions support for python 3.7
Updated 12 days ago1
Batching Graphs only increments edge_index not node_index
Closed 13 days ago1
[Roadmap] Heterogeneous Graphs Explainability Support
Updated 15 days ago4
Gradient does not propagate during network training
Updated 15 days ago1
TypeError: BasicGNN.forward() got an unexpected keyword argument 'return_attention_weights'
Updated 15 days ago1
schnet from_pretrained crashed
Updated 16 days ago1
Metapath2vec index out of bounds error
Closed 19 days ago1
Sparse version of `nn.dense.dense_mincut_pool`
Updated 20 days ago3
Request for Enhancement: Support for Weighted Graphs in ClusterData
Updated 20 days ago2
Heterogenous (HGT) Explainer RuntimeError: mat1 and mat2 shapes cannot be multiplied (10752x512 and 1638x512)
Closed 20 days ago1
Add edge_attr reduction argument to torch_geometric.nn.pool.pool.pool_edge
Closed a month ago1
Dataset penn94 dimension of node features seems to large, 4814 instead of 5
Updated a month ago1
Custom `propagate` is not taken into account by the jit compiler in PyG 2.5.0 and 2.5.1
Closed a month ago11
After installing the dependency libraries, installing torch_geometric reported an error.
Updated a month ago4
Hetero Data Dimension Mismatch
Updated a month ago3
Cat uncoallesced sparse_coo tensor.
Closed a month ago1
QM9 Dataset failed to download
Updated a month ago1
MessagePassing: propagate method
Closed a month ago5
Typo (?) in Aggr.Aggregation's description
Closed a month ago
Add dataset - 1M synthetic random graphs
Updated a month ago5
Creating a graph with `torch_geometric.nn.pool.radius` using `max_num_neighbors` behaves different on GPU than it does on CPU
Updated a month ago3
Example regression GNN architecture for homogeneous graph and node level prediction
Updated a month ago1
Convolutional module which inherits from MessagePassing throws OSError when reloading from read-only disk
Closed 2 months ago2
Empty feature generated when a graph is empty and `trim_to_layer` is enabled
Closed 2 months ago
ModuleNotFoundError: No module named 'torch'
Updated 2 months ago3
Possible overwriting scenario with Jinja
Closed 2 months ago3
Import MessagePassing raise TypeError: metaclass conflict
Closed 2 months ago2
Behavior of `EdgeIndex.get_csr` for directed networks with zero-outdegree nodes
Updated 2 months ago1
Calling Transforms Directly
Updated 2 months ago1