Guohao Li

Can you try the solution from this post? https://stackoverflow.com/questions/37337728/tensorflow-internalerror-blas-sgemm-launch-failed?noredirect=1&lq=1 By adding the following segment to your code ``` python if 'session' in locals() and session is not None: print('Close interactive...

@WeiLong-Zh Thanks for reporting this issue! I will take a look at it. It would be helpful if you can provide a script to reproduce this issue.

> This seems important, but I wonder if we could just define the equivalent aggregation on the aggregation class itself. > > ``` > class SomeAggregation: > @property > def...

Thanks @yanbing-j. LGTM!

Added some simple ones. `MaxAggr`, `MinAggr`, `SumAggr`, `SoftmaxAggr`, `PowermeanAggr`, `VarAggr`, `StdAggr`.

The `Dilated` class will only dilate the provided knn `edge_index`. You would first need to find the knn graph and then dilate it like [DilatedKnnGraph](https://github.com/lightaime/deep_gcns_torch/blob/751382aa2d25e25a2792c133cc99f8cfddae0657/gcn_lib/sparse/torch_edge.py#L32). The `Dilated` class itself wouldn’t...

We should first build a knn graph that has `k*d` neighbors for each nodes then use `[::d]` to get the dilated graphs. `edge_index = self.knn(x, self.k * self.dilation, batch)` So...

Thanks for the suggestion @zademn. That is definitely a good idea if we are dealing with a more complex case. But in our example, we always build knn graphs with...

Hi @SSamDav, sorry that I haven’t got to chance to clean the code of DEQ-GNN yet. I will work on it once I get some free time. Hopefully that would...

Hi @chaitjo. Thanks for your suggestion. We have not tried the proposed methods of DeeperGCN on 3D yet since for 3D point cloud tasks such as s3dis require longer training...