ml-stable-diffusion
ml-stable-diffusion copied to clipboard
Error computing NN outputs
Tried to run the framework on my iPad (6th gen)
and got the following error
[coreml] Error computing NN outputs -1
[coreml] Failure in -executePlan:error:.
Pipeline was initialised successfully however.
Tried using Unet chunks, reduced memory mode, added Extended Virtual Addressing and Increased Memory capabilities without luck.
Any advice would be welcomed!
When I set modelConfig.computeUnits = .cpuOnly
it seems to work at first but crashes later due to memory issues.
How can I know on which device I have to use .cpuOnly
or .cpuAndGPU
?
I've seen a similar error using my own models.
The only way I got around it was to initialize the model once using a singleton. For some reason iOS is fine initializing the model once, but subsequent attempts would result in that error IIRC.
I don't know if that is the case here, but just thought I'd mention it in case it may help.