w013nad
w013nad
Short update. When I did this, my error rates jumped way up. It's still deforming correctly but my RMSE of desired output to actual output jumped form 0.03 to 0.04....
It's not super complex. We have a version working on our end. You just chip out images to the size you want, inference them, offset the results, and run non-max-suppression...
@JustasBart @AyushExel We need to talk with legal but I'd love to help implement it. Regrettably, I'm a bit bound by my NDA. I was also able to implement the...
@protestToViolence Sahi is exactly what was being proposed. Yolo is not capable of estimating the resolution for you. You have to resize the image to match the resolution yourself. If...
How many targets is the model detecting? The segmentation models are very memory hungry when you have lots of targets.
Are you running COCO or a custom dataset. If it's a custom set, how many targets are in the images?
@PureHing Going through those annotations, at least one of those images has over 460 targets. Try removing any images with more than 50 targets and it should work better. YOLOv8...
I got a similar error. It seems to come from trying to put too many tokens into the model. I was putting 5k words into the model. ```Traceback (most recent...
> > Hi! Is there or will there be support for the OpenAI Batch API ? > > vLLM currently has partial support for this (#4794). This requires a completely...
Wait, is this just the error when you're OOM? I thought you had another error message for that? Also, is fp8 still loaded in all at once and then quantized?...