neve33210
neve33210
same here
Same here. convnext in a detection model is prone to trt fp16. For convnext v2, the problem is even worse due to GRN. With that, trt fp16 can't detection anything!
same here. VIT is bad except for ImageNet. that's funny
I'm doing this. no luck so far.
> @oscarnevarezleal @cian0 @cloneofsimo @CombainerA19 @anotherjesse Hi, dears. I have add some code to support the Automatic111 WebUI. What you should do is just runing 'train_lora_dreambooth.py', then you will find...
Maybe it's my mistaking. The DAT model is doing upscaling, just not really doing well. It changes the raw image a little and the result is far from a LDSR...
the exact same issue is happening at yolov8s.pt
same here, stuck for almost 30s between epochs or between train and valid epoch just use workers=0 and fix it for now. There are certainly something to be fixed in...
> @sipie800 I have found the issue cause. It is caused by this line: > > https://github.com/ultralytics/ultralytics/blob/a5410ed79e6ff30ee56d359c43f3c49d6d840ced/ultralytics/yolo/engine/validator.py#L135 > > > > > The code block in this loop takes 0.35s...
> How can apply a temporary fix for this, is using worker_num=0 enough? Yes. But a fix is essential to get it to full speed.