kubasienki

It happens also for me - I wasn't even able to connect via ssh to my server. Running acrarium as a docker container. But I have 80 reports only. Interactions...

Just found out that if I stop training when loss in relatively high it detects objects but a lot of boxes at one object

Probably. My guess is that yolo architecture works on grid for which my objects are too small. But that's just my guess.

Does this PR include the model?

It's not working for me `RuntimeError: CUDA out of memory. Tried to allocate 512.00 MiB (GPU 0; 10.91 GiB total capacity; 8.79 GiB already allocated; 141.25 MiB free; 9.31 GiB...

@ZeroCool22 I've already tried Shivam's repo - do you use any additional memory optimization?

It seems the conda has a too-old version of the transformers. After updating transformers from git and installing a proper version of xformers it works, but it works slower than...