can't install local server on garuda linux, failed to fetch nvidia-cusparselt-cu12

[nidhoegg]

Hey everyone,
I am running into a little issue with the installation of the AI plugin.
I am running Garuda Linux, Krita Version 5.2.14 and newest Krita AI Diffusion Version (1.43.0). Or better to say, I am trying to run it.
Installing Krita and adding AI diffusion to it worked fine, then I am trying to do the local server setup with CUDA.
This is always raising this error:

Downloading nvidia-cufile-cu12 (1.1MiB)
 Downloading nvidia-cufile-cu12
  × Failed to download `nvidia-nvjitlink-cu12==12.8.93`
  ├─▶ Failed to fetch:
  │   `https://pypi.nvidia.com/nvidia-nvjitlink-cu12/nvidia_nvjitlink_cu12-12.8.93-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl`
  ├─▶ Request failed after 3 retries
  ├─▶ error sending request for url
  │   (https://pypi.nvidia.com/nvidia-nvjitlink-cu12/nvidia_nvjitlink_cu12-12.8.93-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl)
  ╰─▶ operation timed out
  help: `nvidia-nvjitlink-cu12` (v12.8.93) was included because `torch`
        (v2.8.0+cu128) depends on `nvidia-nvjitlink-cu12`
Traceback (most recent call last):
  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 350, in install
    await self._install(cb)
  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 177, in _install
    await try_install(comfy_dir, self._install_comfy, comfy_dir, network, cb)
  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 809, in try_install
    raise e
  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 804, in try_install
    await installer(*args)
  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 262, in _install_comfy
    await self._pip_install("PyTorch", ["-U"] + torch_args, cb)

[client.log](https://github.com/user-attachments/files/24272877/client.log)

  File "/home/web/.local/share/krita/pykrita/ai_diffusion/server.py", line 798, in _execute_process
    raise Exception(_("Error during installation") + f": {errlog}")
Exception: Error during installation: Using Python 3.12.11 environment at: venv

I tried installing Krita from the Garuda package manager and I am getting this error.
I tried running krita from Appimage, I am getting this error.
I tried to open the URL it says it can't fetch and it's working fine (it's just not 1.1MB in size like the logfile is saying), I can manually download the package.
So I am running out of ideas.
Any help would be greatly appreciated.
If you need any additional data, I will gladly provide it.
Many thanks in advance.

[nidhoegg]

Here is my full client.log:
client.log

[Acly]

It's using pypi.nvidia.com is that normal? Did you set that up via environment variables or somehow?

[nidhoegg]

I have no idea if it is normal, but I did not change or edit anything.
Like I said, I tried downloading krita from the distro repos and also tried the app image and this is what is happening straight out of the box when I open the AI Diffusion plugin, choose local managed server, cuda and click the install button.
If this is a wrong URL/setup, how can I change it for the plugin?
Than you very much in advance!

[Acly]

Looks like it's normal to download the nvidia packages from pypi.nvidia.com

I just had those downloads fail for me too (it's not always the same package, but always one of the nvidia cuda ones). I also tried directly from the command line and got the same errors. So it's not related to Krita or the plugin, looksl ike something is wrong with their servers at the moment...

I will try fewer parallel downloads, maybe it helps.

[Acly]

It worked with export UV_CONCURRENT_DOWNLOADS=1 - not sure if it really helped or was just luck though

[RomeoKNS]

I confirm that adding 'export UV_CONCURRENT_DOWNLOADS=1' to the krita.desktop file, 'exec' line, the error disappears and the server installs without problems