1.3.1
This is a minor release which is mostly distributed for official PyTorch 1.2 support. In addition, it provides minor bugfixes and the following new features:
Modules
- Non-normalized
ChebConv
in combination with a largest eigenvalue transform TAGCN
- Graph U-Net
Node2Vec
EdgePooling
- Alternative
GMMConv
formulation with separate kernels - Alternative Top-K pooling formulation based on thresholds with examples on synthetic
COLORS
andTRIANGLES
datasets
Datasets
- Pascal VOC 2011 with Berkeley keypoint annotations (
PascalVOCKeypoints
) DBP15K
datasetWILLOWObjectClass
dataset
Please also update related external packages via, e.g.:
$ pip install --upgrade torch-cluster