I get:
"WARNING! No available CUDA detected."
on windows 64 bit, GTX1080, i7-6700, 16GB RAM. Only one gpu (no onboard). Tried running in admin too but it didn't help. Any ideas?
Hi Joe,
Welcome to the community.
First of all, I would make sure that you have the latest graphics drivers for your system.
DainApp is bundled with the necessary CUDA binaries required for it to run (DAIN uses CUDA 9) so you shouldn't need to install anything extra.
You will want to also double-check the GPU that DainApp has selected for use, I know you said you only have one, but if DainApp has'nt selected it by default that could point to an issue.
For reference, what version of DainApp are you using?
Edit: I've been doing some of my own work and I just came across the same error you've been experiencing.
With further testing, I have narrowed it down to the Experimental Algorithm. For some reason it's not recognizing that the PC has a CUDA enabled GPU. The error occurs before any work is sent to the GPU so its not an issue with RAM spikes. It's not a settings problem either as I have tried toggling Half Precision Float, Don't use PyTorch Benchmark and interpolation multipliers in different combinations with no success.
I would advise avoiding the Experimental Algorithm entirely in 0.41
Hi Theo,
Thanks for answering,
I'm using v0.4 and I also tried with 0.22, same problem. What's weird is that 0.22 worked for me before. The only change to the system was driver and os updates. I have the latest NVIDIA drivers and latest update for Windows at the moment.
How do I change algo and see what gpu Dain is using? Is there a config file to edit? I've just been running the exe.
btw: crash log below:
App Crash Log Traceback (most recent call last):
File "my_design.py", line 61, in run
File "my_DAIN_class.py", line 1524, in RenderVideo
File "my_DAIN_class.py", line 846, in Configure
File "networks\DAIN.py", line 53, in __init__
File "site-packages\torch\nn\modules\module.py", line 307, in cuda
File "site-packages\torch\nn\modules\module.py", line 203, in _apply
File "site-packages\torch\nn\modules\module.py", line 203, in _apply
File "site-packages\torch\nn\modules\module.py", line 225, in _apply
File "site-packages\torch\nn\modules\module.py", line 307, in <lambda>
File "site-packages\torch\cuda\__init__.py", line 153, in _lazy_init
RuntimeError: cuda runtime error (100) : no CUDA-capable device is detected at ..\aten\src\THC\THCGeneral.cpp:47
Hmmm,
Have you tried running DainApp as admin?
And are you using the Default algorithm or are you trying to use the Experimental algorithm? I managed to reproduce a "no CUDA capable device error" when using the Experimental algorithm so I would avoid trying to use that.
Its odd that 0.22 has stopped working though. Are you using a Beta, Game Ready or Studio Driver? Looking at Nvidia's descriptions of the drivers, I would use the Studio Driver as it seems to be targeted at maintaining compatibility and stability rather than outright cutting edge (& potentially buggy) performance.
0.41 should be coming out for everybody outside of Patreon in the next month or so (I think) but GRisk has mentioned to others that if they are having GPU driver issues to try and run 0.38 as it may be a bug related to the GPU detection & selection functionality added in 0.40. If the steps above have no effect I'd try 0.38
I hope this helps!
Theo