1

I am trying to run this ballon segmentation model that I found on GitHub: https://github.com/matterport/Mask_RCNN/tree/master/samples/balloon .

It runs on my linux virtual machine but very very slowly, so I stop it after few iterations and installed tensorflow-gpu on my local windows OS, using: https://www.codingforentrepreneurs.com/blog/install-tensorflow-gpu-windows-cuda-cudnn/

everything worked correctly.

But, unfortunately, I got this weird error, for which I can not find any solution on the internet: 'could not create cudnn handle: CUBLAS_STATUS_ALLOC_FAILED

ANY suggestion would be of great great help! I am now blocked since two days and dont know what to try now...

miki
  • 639
  • 2
  • 6
  • 16
  • Possible duplicate of [Tensorflow crashes with CUBLAS\_STATUS\_ALLOC\_FAILED](https://stackoverflow.com/questions/41117740/tensorflow-crashes-with-cublas-status-alloc-failed) – kett May 08 '19 at 08:45

1 Answers1

1

I find where it comes from. I actually had added two different paths in the environment variable (one for cuda 8 and one for cuda 9). deleting the one for cuda 8 remove this error!

miki
  • 639
  • 2
  • 6
  • 16