search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.73k stars 7.96k forks source link

Cuda malloc failed: cannot allocate memory #3284

Open yutaizhou opened 5 years ago

yutaizhou commented 5 years ago

I have trained yolov3-tiny successfully, but I want additional modification to detect bigger objects better since my data set consists of objects of various sizes. I added additional downsampling in the feature extraction stage and appended additional upsampling along with yolo detection layer to the original yolov3-tiny. If it's out of memory, it should tell you something along the line of "out of memory", but what does Cuda malloc failed: cannot allocate memory even mean?

I managed to get past this issue by turning off random for the yolo layers, but this results in significantly reduced mAP.

AlexeyAB commented 5 years ago

Cuda malloc failed: cannot allocate memory

Can you show screenshot?