Creating Dataloader for Custom Dataset | Pytorch Geometric

In this video I show a simple method for creating dataloader for custom graph data in pytorch geometric!

ig: @abhirajkanse

[NasimMahmud]

You have given an answer, and at same time you haven’t. Can you update the collab with some cells with 3/nodes, their node attr, edge attr and then form Data(). Will be real help for newbie. Thanks.

[comp-it]

HI, thanks for the video. I have my dataset of 3D point clouds but in txt files. Total there are 95 files . its a lot of point cloud files. please let me know, how i can make it possible for pytorch geometric i.e making nodes and its features.... etc...

[hoseinmousavi]

HI Abhiraj,
I got an error that hope you could help me.
I am doing the graph classification with PyG.
The edge_index shows something wrong:
1. How can I make sure that the edge_index point the the valid node indices?
2. There are some negative indices in the edge_index.
I think both of these problems comes from a single mistake!
Thanks!

[keshavraghuwanshi]

please make a same video for SMILES notations, i really need it.🙌

[cloverduyen]

Hi, thanks for the video. I also made the code like this. I made a list of data [(data1), (data2)]. And use DalaLoader to load the datalist. But it seems I failed to do so. I guess the problem is the edge array shape which doesn't match the format of DataLoader. It requires the size of [2, number of edges] and must be coo format. Could you tell me how to make this edges array? Thank you

[ronjarieger]

is it possible to read the colab code?
thank you for your work :)

[fasalshah]

Please add some example code for reference.