Spaces:
Running
on
Zero
Been getting this error today, so I thought I'd pass it on
error
I2I Inference Error: NVML_SUCCESS == r INTERNAL ASSERT FAILED at "../c10/cuda/CUDACachingAllocator.cpp":838, please report a bug to PyTorch.
Thank you for reporting. Same here.
ControlNet seemed to be malfunctioning, so I omitted it and rebooted. Some library versions make this functionality unstable...
Haven't had that problem since on here, thanks! Too bad ControlNet is one of the things I'm interested in, but it seems to be having similar issues even in places where it used to work most of the time (like the original DiffuseCraft...)
I'm glad it's not a deep-seated bug!
Right now, with the release of the SD 3.5 model, the work to support quantization, and the renewal of the Zero GPU space, HF and Diffusers are under construction as a whole...
This ControlNet issue is a compounding issue, but we believe the main reason is a bug in the renewal of the Zero GPU space.
After discussing the DiffuseCraft Flux Inference error with the original author of DiffuseCraft, we have concluded that this is most likely the case.