Long training time and batch size #6
Description
Thank you for the impressive job!
However, when I try to reproduce the results, I find the out of memory issue. I only train on one RTX 4090 GPU.
I have reduce the size of dataset to solve. But another question occurred, it require very long time to finish the training. I would like to increase the batch_size, but the whole code seemed only wrote based on batch_size=1. Changing the batch_size would make the whole model not work properly.
P.S. did the SE(3)-transformer applied the faster version contributed by NVIDIA?