Training speed
aaawww222 opened this issue · comments
Hi,
Thanks for the great repo! I’m trying to overfit LGM with a single object in Gojaverse dataset. One interesting thing I found is that it takes around 10 seconds per iteration at the first 20 or 30 iterations and it gradually gets slower to a few minutes or even 10 minutes per iteration. Do you think this is normal?
Thanks!
@aaawww222 Hi, no, the training speed should be stable during the training progress. Have you tried training other models and see if they are getting slower too?