Glenn Jocher b5de52c4cd torch.cuda.amp bug fix (#2750) ...

PR https://github.com/ultralytics/yolov5/pull/2725 introduced a very specific bug that only affects multi-GPU trainings. Apparently the cause was using the torch.cuda.amp decorator in the autoShape forward method. I've implemented amp more traditionally in this PR, and the bug is resolved.

2021-04-09 18:19:49 +02:00

..

hub

Add TransformerLayer, TransformerBlock, C3TR modules (#2333)

2021-04-01 17:26:53 +02:00

__init__.py

add "__init__.py" file in <models> folder

2020-06-15 14:37:07 +08:00

common.py

torch.cuda.amp bug fix (#2750)

2021-04-09 18:19:49 +02:00

experimental.py

Update detections() self.t = tuple() (#2617)

2021-03-26 13:23:11 +01:00

export.py

GPU export options (#2297)

2021-03-06 12:02:10 -08:00

yolo.py

Add TransformerLayer, TransformerBlock, C3TR modules (#2333)

2021-04-01 17:26:53 +02:00

yolov5l.yaml

YOLOv5 v4.0 Release (#1837)

2021-01-04 19:54:09 -08:00

yolov5m.yaml

YOLOv5 v4.0 Release (#1837)

2021-01-04 19:54:09 -08:00

yolov5s.yaml

YOLOv5 v4.0 Release (#1837)

2021-01-04 19:54:09 -08:00

yolov5x.yaml

YOLOv5 v4.0 Release (#1837)

2021-01-04 19:54:09 -08:00