Closed WrongFerret closed 4 months ago
Seems like that error is in PyTorch itself. Can you try downgrading to a stable release? You are using a nightly build because a stable build for ROCm 6.1 doesn't exist yet.
Activate the venv:
source venv/bin/activate
pip install torch==2.3.0+rocm6.0 torchvision==0.18.0+rocm6.0 --index-url https://download.pytorch.org/whl/rocm6.0
Seems like that error is in PyTorch itself. Can you try downgrading to a stable release? You are using a nightly build because a stable build for ROCm 6.1 doesn't exist yet.
Activate the venv:
source venv/bin/activate
pip install torch==2.3.0+rocm6.0 torchvision==0.18.0+rocm6.0 --index-url https://download.pytorch.org/whl/rocm6.0
That was the issue. Thank you very much!
NOTE: Issue is still there for 2.4.1. Might be fixed in nightly.
Fix is to uninstall pynvml pip uninstall pynvml nvidia-ml-py -y
Issue Description
Hello, I just updated to the latest release of SD.Next and upgraded to ROCM 6.1.1 as I saw the newest version supported 6.1. I am using Linux Mint 21.3 on an AMD 7800XT. I am now having issues launching it, with the program not detecting AMDSMI. I've pasted the output of
./webui.sh --debug
below. I've found a couple people using Automatic1111 and ComfyUI having similar issues here and here. I've recompiled amd_smi according to these directions and tried a fresh install of SD.Next but I am still getting this error. Any help would be greatly appreciatedRecompile I tried:
Version Platform Description
inxi -b
output:Relevant .bashrc lines
Relevant log output
Backend
Diffusers
Branch
Master
Model
Other
Acknowledgements