yewzijian
commented
1 year ago The Minkowski engine is mainly used to perform the points subsampling in a faster way using GPU. Previous codes using KPConv does those operations in CPU during data loading. I found that to severely limit the training speed in my case, so I replaced that portion. Unfortunately, that led to many unforeseen issues. I suggest using those CPU operations if you can't get Minkowski to work.
Releasing a version without Minkowski dependencies is something I hope to do in the future. However, I currently don't have access to a GPU workstation and also am busy with other work, so I don't have a timeline for this.
keepgoing2996
The last release date of minkowski is in May 2021. The dependencies of it might not be easily met with new software and hardware. I found it impossible to make RegTR train on my machine because of a cuda memory problem to which I found no solution. Without Minkowski, I would have more freedom when choosing the versions of pytorch and everything, so that I cound have more chance to solve this problem. I am a slam/c++ veteran and deep learning/python newbie(starting learning deep learning 2 weeks ago), so its hard for me to modify it myself for now. I was wondering if you could be so kind to release a version of RegTR without Minkowski.