RunTimeError: Training Faster RCNN

evangelos 2019-02-02 21:40:30 UTC #1

Hi,

The training stopped in half of the potential time, raising a GPU related error. Any help?
Is it worth spending another 5-6h training it?

Cheers,
Evan

redeipirati 2019-02-04 09:52:55 UTC #2

Hi Evan,

The training stopped in half of the potential time, raising a GPU related error.

This doesn't seem a machine or GPU error, but a bug in the code you are running when handling multi-threading. Did you check if this error was reported in the Github repo?

Hope that helps.

evangelos 2019-02-05 10:08:45 UTC #3

Nothing similar on Github. It seems that I get this error when I use a version of VGG16, which is pruned and trained on my own dataset. I do not really know how it is related to the error.
Thank you for your answer.

Cheers,
Evan