Problem with GPU when training YoloV4 on workstation
1 次查看(过去 30 天)
显示 更早的评论
Hi everyone,
I am doing the video detection task on Matlab with YoloV4. I have followed all the stesps as the instructions on MATLAB (https://au.mathworks.com/help/vision/ug/object-detection-using-yolov4-deep-learning.html).
To increase the speed of training, i have change the training option to GPU with "ExcutionEnvironment". However, I am facing the problem with the gpu same like this:
Can anyone recommend the solution for me? Thank you so much.
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1106155/image.png)
My training dataset inclde 527 images and the training options and gpu properties are:
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1106160/image.png)
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1106165/image.png)
![](https://www.mathworks.com/matlabcentral/answers/uploaded_files/1106170/image.png)
1 个评论
Joss Knight
2022-9-1
It looks like a memory issue. You have a 32GiB card but only 110MB available. I suspect some basic underlying features can't work with that little memory. Trying resetting your device (reset(gpuDevice)) then put a breakpoint before the call to trainYOLOv4Detector and check that you have plenty of memory available. If that's okay we may have to check the size of your input data.
回答(0 个)
另请参阅
类别
在 Help Center 和 File Exchange 中查找有关 Image Data Workflows 的更多信息
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!