Mask_rcnn: why I got "floating point exception(core dumped)" error when traning

[heiheiya]

When I training Mask RCNN use
python3 samples/coco/coco.py train --dataset=mypath --model=coco
after train 7th epoch, I got an error:
floating point exception(core dumped)

[engineer1109]

You need to purchase Titan or Tesla

[heiheiya]

My GPU is GeForce GTX 1080 Ti, and I set IMAGES_PER_GPU=1, it isn't ok? The "floating point exception" is due to hardware?

[engineer1109]

@heiheiya Update tensorflow to 1.9

[008karan]

I am also facing same error in 1st epoch itself. using tensorflow 1.8. have you solved it. @heiheiya

[008karan]

changing the cuda version worked for me. now using cuda 9 with cudnn 7 on tensorflow 1.9.

[heiheiya]

@008karan No, it still isn't work for me.

[heiheiya]

@engineer1109 I update tensorflow to 1.9.0, cuda to 9.0, cudnn to 7.0.5, but is still isn't work.

[heiheiya]

May be there are some installation didn't take effect, after I restarted my machine, it worked now. Thank you very much @engineer1109 @008karan