PyG Nightly (PyTorch Geometric Nightly)

2.8.0.dev20260501 verified Fri May 01 auth: no python

Nightly build of PyTorch Geometric (PyG), a library for graph neural networks built on PyTorch. Current version 2.8.0.dev20260501. It tracks the latest development, includes experimental features and unreleased fixes, and is updated daily. Requires Python >=3.10 and PyTorch.

pip install pyg-nightly

Common errors

error ModuleNotFoundError: No module named 'pyg' ↓

cause Incorrect import due to name mismatch between install and import.

fix

Use import torch_geometric.

error RuntimeError: torch_sparse is not installed. Please install it. ↓

cause Missing optional dependency for sparse operations.

fix

Install it: pip install torch-sparse-nightly or use pip install pyg-nightly[full].

Warnings

breaking Nightly builds may introduce API-breaking changes without notice. Code that works with stable PyG may break with pyg-nightly. ↓

fix Pin to a specific nightly version if reproducibility is required, or use the stable release.

gotcha Import as `import torch_geometric`, not `import pyg`. The pip package name is different from the import name. ↓

fix Use `import torch_geometric`.

deprecated Some modules like `torch_geometric.nn.dense` are deprecated in favor of newer modules. ↓

fix Consult the changelog and migrate to the recommended replacements.

Install

pip install pyg-nightly -f https://data.pyg.org/whl/torch-2.5.0+cu121.html

Imports

torch_geometric
wrong
```
import pyg
```
correct
```
import torch_geometric
```
The package installs as pyg-nightly but imports as torch_geometric.

Quickstart

Create a simple graph with two edges.

import torch
import torch_geometric
from torch_geometric.data import Data

edge_index = torch.tensor([[0, 1, 1, 2],
                           [1, 0, 2, 1]], dtype=torch.long)
x = torch.tensor([[-1], [0], [1]], dtype=torch.float)
data = Data(x=x, edge_index=edge_index)
print(data)