Gui Martins

Just did the update to 1.60.0 and it's still running into the same issue. ~~This permission denied issue makes me think it might be the permission of the downloaded files....

Same happening with v1.61.0: ```bash zsh ⌁ kgp NAME READY STATUS RESTARTS AGE immich-machine-learning-6978bfdbdf-z8www 0/1 CrashLoopBackOff 2 (25s ago) 70s ``` ```python Traceback (most recent call last): File "/opt/venv/lib/python3.11/site-packages/insightface/__init__.py", line...

Interesting.. Freshly built container image for machine learning from the main branch: ``` zsh ⌁ docker run -it test Traceback (most recent call last): File "/opt/venv/lib/python3.11/site-packages/insightface/__init__.py", line 8, in import...

Same error building with a simple pip install of the requirements. This is the container image I'm using: ```dockerfile FROM python:3.11.4-bullseye@sha256:5b401676aff858495a5c9c726c60b8b73fe52833e9e16eccdb59e93d52741727 ENV NODE_ENV=production \ TRANSFORMERS_CACHE=/cache \ PYTHONDONTWRITEBYTECODE=1 \ PYTHONUNBUFFERED=1 \...

Only thing that I can se affecting this now is SELinux on the host running the container runtime for k3s. Makes me wonder what exactly is this package trying to...

Error is slightly different now from the new version thanks to the update from #2951 ```bash zsh ⌁ klf immich-machine-learning-668757f9b6-jgmsq Traceback (most recent call last): File "", line 198, in...

Kinda? I mean, the files are labeled as such in the container by default: ``` root@code-658d97b879-j2f6g:/usr/src/app# ls -Z /opt/venv/lib/python3.11/site-packages/onnxruntime/capi/ system_u:object_r:var_lib_t:s0 __init__.py system_u:object_r:var_lib_t:s0 onnxruntime_inference_collection.py system_u:object_r:var_lib_t:s0 _ld_preload.py system_u:object_r:var_lib_t:s0 onnxruntime_pybind11_state.cpython-311-x86_64-linux-gnu.so system_u:object_r:var_lib_t:s0 _pybind_state.py system_u:object_r:var_lib_t:s0...

Aha! That seems to be k3s that didn't enable selinux integration by default: ``` system_u:system_r:unconfined_service_t:s0 30095 ? 00:01:23 longhorn ``` All pods are coming up as `unconfined_service_t` Seems to be...

I can confirm, adding a proper label to the kubernetes containers allowed the execution to work properly. My instance is now running just fine for all machine learning tasks: ![CleanShot...

Same issue here running on Kubernetes. v1.48.1 running on the server and v1.48.0 mobile on iOS 16.4.