QIFEIDKN/STAGATE_pyG

Instability of training results

Opened this issue · 5 comments

zzdj11 commented

Hello, when I run with the pytorch version, I found that even after setting the random seed, I still can't make the result stable, I don't know what's the reason, I hope to get your help, thanks!

forechoandlook commented

你好，我用pytorch版本运行的时候，发现即使设置了随机种子，还是不能让结果稳定，不知道是什么原因，希望能得到你的帮助，谢谢！

Hello, what is your result with this repo? My reproducation result is ARI 0.29 in 151676.

forechoandlook commented

The result has been advanced by this repo https://github.com/QIFEIDKN/STAGATE_pyG-Sigmoid

zzdj11 commented

你好，我用pytorch版本运行的时候，发现即使设置了随机种子，还是不能让结果稳定，不知道是什么原因，希望能得到你的帮助，谢谢！

Hello, what is your result with this repo? My reproducation result is ARI 0.29 in 151676.

Hello! My replication results are very unstable, the ARI results on the 151767 dataset are between 0.50-0.65 and I still can't make the final results stable by random seeding.

forechoandlook commented

你好，我用pytorch版本运行的时候，发现即使设置了随机种子，还是不能让结果稳定，不知道是什么原因，希望能得到你的帮助，谢谢！

Hello, what is your result with this repo? My reproducation result is ARI 0.29 in 151676.

Hello! My replication results are very unstable, the ARI results on the 151767 dataset are between 0.50-0.65 and I still can't make the final results stable by random seeding.

Can you show your code by google drive?

zhaofangyuan98 commented

Hello, could you tell me your env version(python, cuda, pytorch, torch-geometric, torch-scatter, torch-sparse and so on)? Thank you very much!