Memory Issue when running CODEX_SpleenDataset

[Luchixiang]

Hi! Thanks for the great work. I'm trying to run the CODEX_SpleenDataset using your code. But I find it hard to load a (80000, 80000) matrix into the GPU. Training the model on the CPU is really slow. I'm wondering how you trained on this dataset. Did you try sparse tensor and GCN instead of DenseGCN?

[huBioinfo]

Hi, I'm sorry for the very late reply. We didn't use sparse tensor due to the usage of the mincutpool function, which is dependent on the dense tensor implementation.

But after one year's effort, we just released an extended version CytoCommunity+, which can handle datasets with millions of cells while using limited (graphics) memory. If you're still interested, please try https://github.com/huBioinfo/CytoCommunity-plus Thanks!