parameter to set the batch_size

[hzx123q]

Hello,Gal my GPU memory is 3GB that can't train the large point clouds(like lion),could you tell me which parameter is about setting the batch_size?

[galmetzer]

Hi, it depends on which part of the algorithm consumes too much memory.

If it's the dipole part, you can decrease max_pts.

if it's at the network part you can try increasing --number_parts