Deprecate torch_geometric.distributed (#10411)

Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com>
Co-authored-by: Akihiro Nitta <[email protected]>

Author: 3 people

.github/workflows/testing_dist.yml

@@ -87,6 +87,8 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/en/1.0.0/).
 
 ### Deprecated
 
+- Deprecated `torch_geometric.distributed` ([#10411](https://github.com/pyg-team/pytorch_geometric/pull/10411))
+
 ### Fixed
 
 - Fixed conversion to/from `cuGraph` graph objects by ensuring `cudf` column names are correctly specified ([#10343](https://github.com/pyg-team/pytorch_geometric/pull/10343))

CHANGELOG.md

@@ -1,3 +1,3 @@
-https://download.pytorch.org/whl/cpu/torch-1.13.0%2Bcpu-cp39-cp39-linux_x86_64.whl
+https://download.pytorch.org/whl/cpu/torch-2.8.0%2Bcpu-cp310-cp310-manylinux_2_28_x86_64.whl
 numpy>=1.19.5
 git+https://github.com/pyg-team/pyg_sphinx_theme.git

docs/requirements.txt

@@ -1,6 +1,12 @@
 torch_geometric.distributed
 ===========================
 
+.. warning::
+    ``torch_geometric.distributed`` has been deprecated since 2.7.0 and will
+    no longer be maintained. For distributed training, refer to :ref:`our
+    tutorials on distributed training <distributed_tutorials>` or `cuGraph
+    examples <https://github.com/rapidsai/cugraph-gnn/tree/main/python/cugraph-pyg/cugraph_pyg/examples>`_.
+
 .. currentmodule:: torch_geometric.distributed
 
 .. autosummary::

docs/source/modules/distributed.rst

@@ -1,9 +1,11 @@
+.. _distributed_tutorials:
+
 Distributed Training
 ====================
 
 .. nbgallery::
     :name: rst-gallery
 
-    distributed_pyg
     multi_gpu_vanilla
     multi_node_multi_gpu_vanilla
+    distributed_pyg

docs/source/tutorial/distributed.rst

@@ -1,6 +1,10 @@
 Distributed Training in PyG
 ===========================
 
+.. warning::
+    ``torch_geometric.distributed`` has been deprecated and will no longer be maintained.
+    For distributed training with cuGraph, refer to `cuGraph examples <https://github.com/rapidsai/cugraph-gnn/tree/main/python/cugraph-pyg/cugraph_pyg/examples>`_.
+
 .. figure:: ../_figures/intel_kumo.png
    :width: 400px
 

docs/source/tutorial/distributed_pyg.rst

@@ -1,6 +1,10 @@
 Multi-GPU Training in Pure PyTorch
 ==================================
 
+.. note::
+    For multi-GPU training with cuGraph, refer to `cuGraph examples <https://github.com/rapidsai/cugraph-gnn/tree/main/python/cugraph-pyg/cugraph_pyg/examples>`_.
+
+
 For many large scale, real-world datasets, it may be necessary to scale-up training across multiple GPUs.
 This tutorial goes over how to set up a multi-GPU training  pipeline in :pyg:`PyG` with :pytorch:`PyTorch` via :class:`torch.nn.parallel.DistributedDataParallel`, without the need for any other third-party libraries (such as :lightning:`PyTorch Lightning`).
 Note that this approach is based on data-parallelism.

docs/source/tutorial/multi_gpu_vanilla.rst

@@ -1,6 +1,10 @@
 Multi-Node Training using SLURM
 ===============================
 
+.. note::
+    For multi-GPU training with cuGraph, refer to `cuGraph examples <https://github.com/rapidsai/cugraph-gnn/tree/main/python/cugraph-pyg/cugraph_pyg/examples>`_.
+
+
 This tutorial introduces a skeleton on how to perform distributed training on multiple GPUs over multiple nodes using the `SLURM workload manager <https://slurm.schedmd.com/>`_ available at many supercomputing centers.
 The code is based on `our tutorial on single-node multi-GPU training <multi_gpu_vanilla.html>`_.
 Please go there first to understand the basics if you are unfamiliar with the concepts of distributed training in :pytorch:`PyTorch`.

docs/source/tutorial/multi_node_multi_gpu_vanilla.rst

@@ -3,6 +3,6 @@
 This directory contains examples for distributed graph learning.
 The examples are organized into two subdirectories:
 
-1. [`pyg`](./pyg): Distributed training via PyG's own `torch_geometric.distributed` package.
+1. [`pyg`](./pyg): Distributed training via PyG's own `torch_geometric.distributed` package (deprecated).
 1. [`graphlearn_for_pytorch`](./graphlearn_for_pytorch): Distributed training via the external [GraphLearn-for-PyTorch (GLT)](https://github.com/alibaba/graphlearn-for-pytorch) package.
 1. [`kuzu`](./kuzu): Remote backend via the [Kùzu](https://kuzudb.com/) graph database.

examples/distributed/README.md

@@ -1,6 +1,6 @@
 # Distributed Training with PyG
 
-**[`torch_geometric.distributed`](https://github.com/pyg-team/pytorch_geometric/tree/master/torch_geometric/distributed)** implements a scalable solution for distributed GNN training, built exclusively upon PyTorch and PyG.
+**[`torch_geometric.distributed`](https://github.com/pyg-team/pytorch_geometric/tree/master/torch_geometric/distributed)** (deprecated) implements a scalable solution for distributed GNN training, built exclusively upon PyTorch and PyG.
 
 Current application can be deployed on a cluster of arbitrary size using multiple CPUs.
 PyG native GPU application is under development and will be released soon.