Hi there!
I moved my codebase to newest version of DGL (0.8.1), because I thought that it supports (as mentioned in title) sampling edges for heterogenous graphs on GPU. Unfortunately I can’t find any docs mentioning how to actually do it. I tried passing train_eid_dict (with values moved on gpu), but there comes error that everything should be on GPU. Could you tell me if the feature is ready and if so, where can I find how to update dgl.dataloading.DataLoader initialization to work as expected?
Cheers!
PR for the fix is in [Bug] Fix multi-GPU edge classification crashing with pure GPU sampling by BarclayII · Pull Request #3946 · dmlc/dgl · GitHub.
This topic was automatically closed 30 days after the last reply. New replies are no longer allowed.