Hi!
I was trying to test SALIENT++ library on my own dataset on a local machine with 1 single GPU.
I read both works on SALIENT and SALIENT++ with specific indications in the repositories, and I have installed SALIENT++ library as suggested. Now, my question is on how to introduce a fast sampler in my code where I am currently using the PyG NeighborLoader, to try to appreciate a speed-up on training and inference on a large graph dataset.
To ensure to overcome to the PyG bottlenecks, should I simply load my datatset using the from_pyg method of FastDataset class and the recall a model from the ones defined in driver/models.py?
Thank you
Hi!
I was trying to test SALIENT++ library on my own dataset on a local machine with 1 single GPU.
I read both works on SALIENT and SALIENT++ with specific indications in the repositories, and I have installed SALIENT++ library as suggested. Now, my question is on how to introduce a fast sampler in my code where I am currently using the PyG
NeighborLoader, to try to appreciate a speed-up on training and inference on a large graph dataset.To ensure to overcome to the PyG bottlenecks, should I simply load my datatset using the
from_pygmethod ofFastDatasetclass and the recall a model from the ones defined indriver/models.py?Thank you