search

cellmodeller / CellModeller

GPU-accelerated multicellular modelling framework
http://cellmodeller.github.io/CellModeller/
Other
56 stars 48 forks source link

"ValueError: cannot convert float NaN to integer" error for newer generations of NVIDIA GPUs? #28

Closed avimanyu786 closed 3 years ago

avimanyu786 commented 3 years ago

I'm getting this error on an RTX 3090 card. Since it has also been reported on an RTX 2080 SUPER as mentioned on issue https://github.com/HaseloffLab/CellModeller/issues/26#issuecomment-728847850, could it be an issue for RTX GPUs in particular?

I can run it just fine on a GTX 1050 Ti. But replicating the same thing with an RTX 3090 throws that error whenever I try to load and run an example simulation. The error is also accompanied with an Aborted (core dumped) message at random times.

Also, I would like to mention that the 1050 Ti is running on nvidia-driver-390, whereas the 3090 is running on nvidia-driver-460. I tried to install nvidia-driver-390 for 3090 to investigate this issue but it keeps giving a black screen giving no other alternative but to switch back to 460 again.

avimanyu786 commented 3 years ago

The above test was done on Ubuntu 20.04. When I tested this on Ubuntu 18.04, I got the same issue. There is no black screen on the display but the RTX 3090 refuses to work with nvidia-driver-390 giving only but a maximum resolution of 640 x 480 when I install this driver. The GPU only seems to recognize nvidia-driver-460 which is the most recent and available version on both OSs. The GTX 1050 Ti discussed above is on Ubuntu 18.04.

avimanyu786 commented 3 years ago

@danilexn Thanks a ton! Yet to test this on the 3090 but can confirm with the 1050 Ti that apparently, this commit also fixes #26..was able to run Tutorial_1a.py without any issue upto 10,000 cells.

avimanyu786 commented 3 years ago

@danilexn Just tested it on the 3090 and it's finally working :smile: :raised_hands: !!