Search code examples
google-colaboratoryyolodarknet

Why am I getting a CUDA Error when calculating mAP in darknet YOLOv2 in Google Colab?


Code that used to work a month or two ago now posts the following error:

 calculation mAP (mean average precision)...
 Detection layer: 15 - type = 27 
4CUDA status Error: file: ./src/dark_cuda.c : () : line: 598 : build time: Sep 15 2021 - 19:35:53 

 CUDA Error: an illegal memory access was encountered
Darknet error location: ./src/dark_cuda.c, check_error, line #70
CUDA Error: an illegal memory access was encountered: File exists

Below is a link to a notebook created by tech zizou that I used (I just modified the weights and config for YOLOv2).

https://colab.research.google.com/drive/1zqRb08ljHvIIMR4fgAXeNy1kUtjDU85B?usp=sharing

Any idea on what has changed recently that would cause such an error?


Solution

  • It appears that Google Colab has reduced the memory allowed for free notebooks.

    Changing the config file from batch=64 and subdivisions=2 to batch=32 and subdivisions=8 solved this problem for me.