dgl
dgl copied to clipboard
dmlc
Python package built to ease deep learning on graph, on top of existing DL frameworks.
## Description Rework Dataloader cpu affinitization to affinitize compute threads as well by default, making it a context manager helper method The feature enables dataloading worker threads affinitization and separates...
## Description In order to support canonical etypes for DistDGL, let's enable it in graph partition book first. What tricky here is to support both etypes and canonical. In current...
## Description I've locally successfully tested `test_chunk_graph` and `test_partition`. The test for `test_dispatch` raised an error and I'm not sure if this is due to an inappropriate ip config file....
## 🐛 Bug I am having some problems using the RelGraphConv layer with one-hot encodings as node features. Apparently there is some problem eitther with the 'message' method of the...
import dgl import numpy as np import torch as th from dgl.nn import GATConv u = [0, 1, 0, 2, 1] v = [0, 1, 4, 3, 2] w =...
## Description A gpu_cache that can be used to cache vertex or edge feature storage is implemented. The core implementation comes from the HugeCTR repository and this PR is basically...
## 🐛 Bug I am seeing the same issue that was reported as fixed in #3841 in the latest `0.9.0` (and everything down to releases lower than `0.8.0`. See #3802...
## Description * Added `exclude_self` and `output_batch` options to `knn_graph` and `segmented_knn_graph`, and their wrappers, `dgl.nn.KNNGraph` and `dgl.nn.SegmentedKNNGraph` for Issue #4316 * Updated out-of-date comments on `remove_edges` and `remove_self_loop`, since...
## Description Hi, during my internship at NVIDIA Devtech AI division, I worked on exploring ways to accelarate minibatch training on GNNs. During this time, we noticed that if the...
Everything continues to work if we remove the cub submodule and the cuda_include_directory for cub from CMakeLists.txt https://github.com/dmlc/dgl/blob/master/CMakeLists.txt#L54. With the way things currently are, there are 2 cubs available, third_party/thrust/cub...