RX 6600 is not used

[jilek-josef]

Hi I am on bare metal ubuntu and I succesfully managed to setup SD Web UI aside of one thing my CPU does all the hard job. I use this command to run TORCH_COMMAND='pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/rocm5.1.1' HSA_OVERRIDE_GFX_VERSION=10.3.0 python launch.py --no-half --skip-torch-cuda-test however my GPU RX 6600 is not used. Please help.
Log:
(venv) pc@pc:~/stable-diffusion-webui$ TORCH_COMMAND='pip install torch torchvision --extra-index-url https://download.pytorch.org/whl/rocm5.1.1' HSA_OVERRIDE_GFX_VERSION=10.3.0 python launch.py --no-half --skip-torch-cuda-test
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
Commit hash: e672cfb
Installing requirements for Web UI
Launching Web UI with arguments: --no-half
/home/pc/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/cuda/init.py:88: UserWarning: HIP initialization: Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice (Triggered internally at ../c10/hip/HIPFunctions.cpp:110.)
return torch._C._cuda_getDeviceCount() > 0
Warning: caught exception 'Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice', memory monitor disabled
No module 'xformers'. Proceeding without it.
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
Loading weights [3e1a125f] from /home/pc/stable-diffusion-webui/models/Stable-diffusion/wd-v1-3-full-opt.ckpt
Applying cross attention optimization (InvokeAI).
Model loaded.
Textual inversion embeddings loaded(0):
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().

[denliner]

I have encountered the same issue.
You need to re-install torch and torchvision with "--index-url" instead of "--extra-index-url".
Using "--extra-index-url" for that kind of thing is bad idea because of pip's sketchy priority order pypa/pip#8606.
Pip probably installed the CUDA version instead of the ROCm version.

[jilek-josef]

Thanks for reply, I uninstalled torch and torchvision and run the edited command but it didn't work so I deleted venv folder and try to setup it from 0 but it's kinda confused now and still looks like it installs the CUDA version
Log:
(venv) pc@pc:~/stable-diffusion-webui$ TORCH_COMMAND='pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm5.1.1' HSA_OVERRIDE_GFX_VERSION=10.3.0 python launch.py --no-half --skip-torch-cuda-test
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
Commit hash: e672cfb
Installing torch and torchvision
Traceback (most recent call last):
File "/home/pc/stable-diffusion-webui/launch.py", line 294, in
prepare_environment()
File "/home/pc/stable-diffusion-webui/launch.py", line 206, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch")
File "/home/pc/stable-diffusion-webui/launch.py", line 49, in run
raise RuntimeError(message)
RuntimeError: Couldn't install torch.
Command: "/home/pc/stable-diffusion-webui/venv/bin/python" -m pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm5.1.1
Error code: 1
stdout: Looking in indexes: https://download.pytorch.org/whl/rocm5.1.1
Collecting torch
Using cached https://download.pytorch.org/whl/rocm5.1.1/torch-1.13.1%2Brocm5.1.1-cp310-cp310-linux_x86_64.whl (1358.4 MB)
Collecting torchvision
Using cached https://download.pytorch.org/whl/rocm5.1.1/torchvision-0.14.1%2Brocm5.1.1-cp310-cp310-linux_x86_64.whl (70.5 MB)
Collecting torch
Using cached https://download.pytorch.org/whl/rocm5.1.1/torch-1.13.0%2Brocm5.1.1-cp310-cp310-linux_x86_64.whl (1360.8 MB)
Using cached https://download.pytorch.org/whl/rocm5.1.1/torch-1.12.1%2Brocm5.1.1-cp310-cp310-linux_x86_64.whl (1354.7 MB)
Using cached https://download.pytorch.org/whl/rocm5.1.1/torch-1.12.0%2Brocm5.1.1-cp310-cp310-linux_x86_64.whl (1354.7 MB)

The conflict is caused by:
torch 1.13.1+rocm5.1.1 depends on typing-extensions
torch 1.13.0+rocm5.1.1 depends on typing-extensions
torch 1.12.1+rocm5.1.1 depends on typing-extensions
torch 1.12.0+rocm5.1.1 depends on typing-extensions

To fix this you could try to:

1. loosen the range of package versions you've specified
2. remove package versions to allow pip attempt to solve the dependency conflict

stderr: ERROR: Cannot install torch==1.12.0+rocm5.1.1, torch==1.12.1+rocm5.1.1, torch==1.13.0+rocm5.1.1 and torch==1.13.1+rocm5.1.1 because these package versions have conflicting dependencies.
ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts

(venv) pc@pc:/stable-diffusion-webui$ pip install typing-extensions
Collecting typing-extensions
Using cached typing_extensions-4.4.0-py3-none-any.whl (26 kB)
Installing collected packages: typing-extensions
Successfully installed typing-extensions-4.4.0
(venv) pc@pc:/stable-diffusion-webui$ TORCH_COMMAND='pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm5.1.1' HSA_OVERRIDE_GFX_VERSION=10.3.0 python launch.py --no-half --skip-torch-cuda-test
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
Commit hash: e672cfb
Installing torch and torchvision
Installing gfpgan
Installing clip
Installing open_clip
Installing requirements for CodeFormer
Installing requirements for Web UI
Launching Web UI with arguments: --no-half
/home/pc/stable-diffusion-webui/venv/lib/python3.10/site-packages/torchvision/io/image.py:13: UserWarning: Failed to load image Python extension: libc10_cuda.so: cannot open shared object file: No such file or directory
warn(f"Failed to load image Python extension: {e}")
/home/pc/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/cuda/init.py:88: UserWarning: HIP initialization: Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice (Triggered internally at ../c10/hip/HIPFunctions.cpp:110.)
return torch._C._cuda_getDeviceCount() > 0
Warning: caught exception 'Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice', memory monitor disabled
Segmentation fault (core dumped)

[fractal-fumbler]

try instead

--extra-index-url https://download.pytorch.org/whl/rocm5.2

[jilek-josef]

Tried reinstall again with rocm5.2 but I am back where I was
Log
(venv) pc@pc:~/stable-diffusion-webui$ TORCH_COMMAND='pip install torch torchvision --index-url https://download.pytorch.org/whl/rocm5.2' HSA_OVERRIDE_GFX_VERSION=10.3.0 python launch.py --no-half --skip-torch-cuda-test
Python 3.10.6 (main, Nov 14 2022, 16:10:14) [GCC 11.3.0]
Commit hash: e672cfb
Installing torch and torchvision
Installing requirements for Web UI
Launching Web UI with arguments: --no-half
/home/pc/stable-diffusion-webui/venv/lib/python3.10/site-packages/torch/cuda/init.py:88: UserWarning: HIP initialization: Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice (Triggered internally at ../c10/hip/HIPFunctions.cpp:110.)
return torch._C._cuda_getDeviceCount() > 0
Warning: caught exception 'Unexpected error from hipGetDeviceCount(). Did you run some cuda functions before calling NumHipDevices() that might have already set an error? Error 101: hipErrorInvalidDevice', memory monitor disabled
No module 'xformers'. Proceeding without it.
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
Loading weights [3e1a125f] from /home/pc/stable-diffusion-webui/models/Stable-diffusion/wd-v1-3-full-opt.ckpt
Applying cross attention optimization (InvokeAI).
Model loaded.
Textual inversion embeddings loaded(0):
Running on local URL: http://127.0.0.1:7860