Can edge_softmax be computed on out-neighbor?

https://docs.dgl.ai/api/python/nn.pytorch.html?highlight=edge_softmax#dgl.nn.pytorch.softmax.edge_softmax
Hi, I want to do softmax based on out-neighbors. But the doc says that it is based on in-neighbors. Any ideas to do edge_softmax based on out-neighbors given an existed graph?

Currently edge_softmax do not have such functionalities, one workaround is to use reverse api to get a graph which reverse the order of source/destination node of each edge in the original graph, the apply edge_softmax on this graph.

We will provide reduction on source nodes and edge softmax on outer neighbors in the next release.

Another workaround is using group_apply functions. https://docs.dgl.ai/generated/dgl.DGLGraph.group_apply_edges.html?highlight=group_apply#dgl.DGLGraph.group_apply_edges

However this is not efficient as the built-in edge_softmax function.