lightaime/deep_gcns_torch

error occurs in training modelnet40 after epoch 20

amiltonwong opened this issue · 2 comments

amiltonwong commented

Hi, @lightaime ,

I got the error RuntimeError: CUDA error: device-side assert triggered, after training modelnet40 in epoch 20 (details in log.txt)

Could you give me some hints on how to fix this issue?

Thanks~

guochengqian commented

hi @amiltonwong no worry, I will check it as soon as possible.

guochengqian commented

hi @amiltonwong. Please use the most recent code. I make the batch_index_select more stable. I train them just fine and easily reproduce the results.