Burhan
Burhan
Updated to `time_sync()` if backward pass is successful or `float("nan")` otherwise
With regard to `mosaic` augmentation, as far as I'm aware, nothing has changed. You can see the full release notes here https://github.com/ultralytics/ultralytics/releases/tag/v8.3.0
Doesn't the `mamba-ssm` install require some complex CUDA setup/configurations? Maybe that's not what you're using @Tdzdele but I was looking at this the other day, and that seemed to be...
Similar to #8646
@glenn-jocher the error has to do with the fact that FastSAM doesn't support `val()` but there's a `pytest` for running `val()` with FastSAM with the changes. This was part of...
Found the previous PR https://github.com/ultralytics/ultralytics/pull/8762 including here for record keeping sake
@microsoft-github-policy-service agree company="Crexi"
@u-uzun see [my reply](https://community.ultralytics.com/t/ultralytics-actions-format-pull-request-target-failing/395/3?u=burhanq) to your post on the forums
> Is it possible to train without a validation set during yolov8? Yes ```py from ultralytics import YOLO model = YOLO("yolov8n.pt") result = model.train(data="coco8.yaml", val=False) ```
Maybe the x-axis coordinates are flipped? The result from `class 18` and `class 119` looks like it _might_ align if you flip the x-coordinates along the vertical axis.