This seems to be an upstream issue with WGPU, but I figured I'd start by reporting it here (lmk if I should report to wgpu-py instead/as well). In a clean conda environment with ONLY fastplotlib and dependencies installed, fastplotlib imports with minor issues (see below). Even after 4+ hours of debugging, I could not resolve these entirely. They seem relatively harmless, although I have not confirmed that fastplotlib actually works.

Unable to find extension: VK_EXT_physical_device_drm
No config found!
Max vertex attribute stride unknown. Assuming it is 2048
Max vertex attribute stride unknown. Assuming it is 2048

However, even in a relatively minimal conda environment with the scipy stack (including matplotlib + seaborn), sklearn, and pytorch, I get the more catastrophic failure RuntimeWarning: WGPU could not enumerate any adapters, fastplotlib will not work. Running vulkaninfo --summary outside any environment and in the working environment shows two adapters: the GPU and the CPU. As soon as I load the minimal environment the GPU disappears. This appears to be related to vulkan runtimes installed by conda (as diagnosed by ls $CONDA_PREFIX/lib | grep -E 'vulkan|EGL|GLX|gbm|drm'). These seem to prevent the working GPU runtime from loading correctly. As I understand it, WGPU will try and fall back on other backends, but those obviously fail as well, and this is a bit over my head. This seems like a pretty obvious configuration issue on my end, but if I want to use fastplotlib I need it to work inside my analysis environment.

System:
Ubuntu 24 LTS
Nvidia GPU (driver version 580)
Vulkan 1.3.275
wgpu-py 0.29
fastplotlib 0.6.1
glfw 2.10.0 (clean environment)
pygfx 0.15.3 (minimal env)