Xiao Han

Hi, When checking a traceback, I expect that I can click on the any file path, which leads me to the actual file. For instance, in the following buffer, when...

# Summary I cannot use `elpy-goto-definition` for my own modules when editing Jupyter notebooks. # Steps to reproduce Say I am in Python project (you may clone it from https://github.com/xiaohan2012/test-goto-def)...

MinCut pooling for sparse adjacency matrix

1

# Motivation The current `nn.dense.mincut_pool` requires the input adjacency matrix to be a dense matrix. This requirement does not scale to large graphs # What This PR considers a sparse...

# What and why - `nn.dense.DMoNPooling` requires the input adjacency matrix to be dense, which does not scale to large graphs. - This PR adds the "sparse" version of `DMoNPooling`

Sparse version of `nn.dense.dense_mincut_pool`

6

### 🚀 The feature, motivation and pitch The [nn.dense.dense_mincut_pool](https://pytorch-geometric.readthedocs.io/en/latest/generated/torch_geometric.nn.dense.dense_mincut_pool.html#torch_geometric.nn.dense.dense_mincut_pool) operator requires the input adjacency matrix to be dense. This requirement does not scale well as the matrix size grows. We...

pickling models

1

Hi, I'm trying to use napkinXC under [ray](https://www.ray.io/), which relies `pickle` for data serialization. It seems that napkinXC models cannot be pickled. For instance, ```python import pickle from napkinxc.models import...

The PR fixes the bug described in https://github.com/pyg-team/pytorch_geometric/issues/10148

### 🐛 Describe the bug In [DMoNPooling](https://pytorch-geometric.readthedocs.io/en/2.6.0/generated/torch_geometric.nn.dense.DMoNPooling.html), calculating the cluster loss (the $$\mathcal{L}_c$$ term in the documentation) requires calculating the graph sizes for a batch of graphs. The current implementation,...