Skip to content

Releases: THUDM/CogDL

CogDL v0.6

27 Apr 10:57
bf094be
Compare
Choose a tag to compare

The new v0.6 release updates the tutorials and adds more examples, such as GraphMAE, GraphMAE2, and BGRL.

What's Changed

New Contributors

Full Changelog: v0.5.3...v0.6

CogDL v0.5.3

01 Jun 12:10
398f931
Compare
Choose a tag to compare

Release 0.5.3

The CogDL v0.5.3 release supports mixed-precision training by setting fp16=True and provides a basic example written by Jittor. It also updates the tutorial in the document, fixes downloading links of some datasets, and fixes potential bugs of operators.

What's Changed

Full Changelog: v0.5.2...v0.5.3

CogDL v0.5.2

16 Dec 05:37
dc61b6e
Compare
Choose a tag to compare

Release 0.5.2

The CogDL 0.5.2 release adds a GNN example for ogbn-products and updates geom datasets. It also fixes some potential bugs including setting devices, using cpu for inference, etc.

What's Changed

Full Changelog: v0.5.1...v0.5.2

CogDL v0.5.1

01 Dec 16:10
d181881
Compare
Choose a tag to compare

Release 0.5.1

The CogDL 0.5.1 release adds fast operators including SpMM (cpu version) and scatter_max (cuda version). It also adds lots of datasets for node classification which can be found in this link.

What's Changed

Full Changelog: v0.5.0...v0.5.1

CogDL v0.5.0

20 Nov 15:37
c9c6dde
Compare
Choose a tag to compare

Release 0.5.0

The CogDL 0.5.0 release focuses on modular design and ease of use. It designs and implements a unified training loop for GNN, which introduces DataWrapper to help prepare the training/validation/test data and ModelWrapper to define the training/validation/test steps.

What's Changed

New Contributors

Full Changelog: 0.4.1...v0.5.0

CogDL v0.5.0-alpha1

06 Nov 17:43
096f36b
Compare
Choose a tag to compare
CogDL v0.5.0-alpha1 Pre-release
Pre-release

Release 0.5.0-alpha1

The CogDL 0.5.0 release focuses on modular design and ease of use. It designs and implements a unified training loop for GNN, which introduces DataWrapper to help prepare the training/validation/test data and ModelWrapper to define the training/validation/test steps.

CogDL v0.5.0-alpha0

27 Oct 16:58
e0e8f3d
Compare
Choose a tag to compare
CogDL v0.5.0-alpha0 Pre-release
Pre-release

Release 0.5.0-alpha0

The CogDL 0.5.0 release focuses on modular design and ease of use. It designs and implements a unified training loop for GNN, which introduces DataWrapper to help prepare the training/validation/test data and ModelWrapper to define the training/validation/test steps.

CogDL v0.4.1

13 Aug 14:07
481c08f
Compare
Choose a tag to compare

A new release! 🎉🎉🎉
In the new v0.4.1 release, CogDL implements multiple deepgnn models and we also give a analysis of deepgnn in Chinese. Now CogDL. supports both reversible and actnn for memory efficiency to help build super deep GNNs. Come and have a try. BTW, we are glad to announce that we will give a tutorial on KDD 2021 in August. Please see this link for more details. 🎉

New Features

  • #230 Add new tasks for OAGBert, including zero-shot inference and supervised classification
  • #243 #251 Add new pipelines of GenerateEmbeddingPipeline
  • #248 Add recommendation task
  • #249 Separate layers from models for users to build custom models more conveniently.
  • #256 Add message-passing base framework.
  • #262 #263 #266 Supports actnn in graph neural networks
  • #266 Add message-passing ops implemented in Python

New Models

  • #258 Add c&s(correct and smooth) and SAGN
  • #260 #261 Add RevGNN wrappers and models (revgcn, revgat, revgen)

New Datasets

  • #230 Add datasest for OAGBert: l0fos, aff30, arxivvenue.

New Examples

  • #265 Implements HGNN using CogDL.

Bug Fixes

  • #237 #240 Fix bugs in calling ge-spmm and using Graph
  • #238 Modify examples of gnns to adapt to cogdl.Graph.
  • #257 Fix bugs in ogb datasets and moe-gcn
  • #259 Fix bugs in calling cusparse API.

Docs

  • #242 Add a brief tutorial for CogDL.

CogDL v0.4.0

30 May 16:29
3206099
Compare
Choose a tag to compare

A new major release! 🎉🎉🎉
The new v0.4.0 release refactors the data storage (from Data to Graph) and provides more fast operators to speed up GNN training. It also includes many self-supervised learning methods on graphs. BTW, we are glad to announce that we will give a tutorial on KDD 2021 in August. Please see this link for more details. 🎉

New Features

  • Reformat Data Storage (from Data to Graph), edge_index from torch.Tensor to tuple(Tensor, Tensor). The inputs of each GNN are unified as one parameter graph.
  • #205 #210 #212 Add SDDMM operator
  • #234 Add multi-head SpMM operator and speed up edge_softmax.
  • #211 #222 Support distributed training

New Models

New Datasets

  • #226 Add ogbn-mag dataset

New Examples

  • #233 Add Simple-HGN model

Bug Fixes

  • #209 Fix STPGNN and heterogeneous task
  • #225 Fix TUDataset

CogDL v0.3.0

03 Mar 12:02
9833a4b
Compare
Choose a tag to compare

A new major release! 🎉🎉🎉
It provides a fast spmm operator to speed up GNN training. We also release the first version of CogDL paper in arXiv. In the paper, we introduce the design, the characteristics, the features, and the reproducible leaderboards.
Welcome to join our slack!

New Features

New Models

  • #193 Add ClusterGCN model
  • #194 Add GraphSAINT model

New Datasets

  • #167 Add Reddit dataset
  • #175 Add PPI dataset

New Examples

  • #173 Add usages of customized models
  • #174 Add usages of customized datasets

Miscellaneous

  • #170 Remove PyG dependency of several models
  • #169 #174 #182 Remove PyG dependency of datasets