SJTU-IPADS / gnnlab

A Factored System for Sample-based GNN Training over GPUs

SJTU-IPADS/gnnlab Issues

Cannot re-initialize CUDA in forked subprocess. To use CUDA with multiprocessing, you must use the 'spawn' start method
Closed 10 months ago5
How does gnnlab implement neighbor sampling for GCN models in CUDA?
Closed a year ago4
Failed to run in A100
Closed a year ago7
Reading file error when generating cache rank table
Closed a year ago26
Run gnnlab with smaller batch size
Closed 2 years ago6
/graph-learning/utility/data-process/common/options.h:21:10: fatal error: CLI/App.hpp: No such file or directory
Closed a year ago2
Check failed: (data) != ((void *)-1)
Closed a year ago2
Java class can not be find when preparing the dataset.
Closed 2 years ago4
CUDA: invalid device function for single GPU training
Closed 2 years ago7
Some questions about implementation.
Closed 2 years ago7
Is the hot cached embeddings in each GPU the same?
Closed 2 years ago2
A question about paper results
Closed 2 years ago8