Jianyuan Wang comments

Results 238 comments of


                                            Jianyuan Wang

Training loss first decrease then increase

1. The loss curve largely depends on your training set. While, I would suggest visualizing using a higher smoothness value in tensorboard. 2. Yeah the loss should be negative, this...

CUDA out of memory on demo_colmap.py

Hi I heard this issue before but I cannot help debug this problem as I don't have access to an 4090

CUDA out of memory on demo_colmap.py

Hi @XLR-man Thanks @murlock1000 for sharing the solution. Does this solve your problem? https://github.com/facebookresearch/vggt/pull/253

CUDA out of memory on demo_colmap.py

I guess this is a problem specific to 4090

How to limit dataset usage

Hi for this purpose I would suggest setting len_train for each dataset manually

pycolmap DLL initialization failed on RTX 5090 with CUDA 12.8

It can be solved by: https://github.com/colmap/pycolmap/issues/288

padding without mask

Hey it should not, as we also do so during training.

Is there a planning updating DINOv2 to DINOv3?

It will come with our next release.