Unable to enable CUDA processing for ML inference

Passing NVIDIA card to immich with parameters specified in the documentation makes Immich see the NVidia card as expected. However, Immich doesn't seem to have "CUDAExecutionProvider" in its onnxruntime package. It only loads ['AzureExecutionProvider', 'CPUExecutionProvider']. For example, run this 'docker exec -it immich bash -c "python3 -c \"import onnxruntime as ort; print(ort.get_available_providers())\""' As a result, Immich doesn't leverage CUDA at all and ML processing falls on CPU.