I’m not familiar with PyTorch Geometric, but in case you can precompute the targets and each sample has one corresponding target class, you could take a look at this code snippet, which explains the usage of the WeightedRandomSampler.
I’m not familiar with PyTorch Geometric, but in case you can precompute the targets and each sample has one corresponding target class, you could take a look at this code snippet, which explains the usage of the WeightedRandomSampler.