gpu issue on qiskit - aer method , docker - cuquantum-appliance:23.10

[e-dinesh]

Hi, I'm facing an issue with my AER-State Vector simulation.

I'm using the docker image <nvcr.io/nvidia/cuquantum-appliance:23.10> and run this file [https://github.com/qiskit-community/qiskit-community-tutorials/blob/master/aer/qv_cuStateVec.ipynb]

If I use < sim = AerSimulator(method='statevector', device='GPU') >, I can see the utilization of GPUs. On the other hand, if I modify <sim = Aer.get_backend('statevector_simulator', device='GPU')> it points to cpu, NOT the GPUs.

Any idea, why is it? @sam-stanwyck

[mtjrider]

Hi @e-dinesh

Thanks for posting. Sorry to hear you're having some trouble.

To diagnose, we'll need more information. Can you provide us with some output? See below.

1. By using your container's Python:

import qiskit_aer
devices = qiskit_aer.backends.aer_simulator.AerSimulator().available_devices()
print(devices)

2. What command did you launch to create the container?

##############↓↓↓↓↓↓↓↓↓↓↓↓
docker run -it --gpus all nvcr.io/nvidia/cuquantum-appliance:23.10
##############↑↑↑↑↑↑↑↑↑↑↑↑

If --gpus is not specified, the container falls back to a runtime configuration without visibility to the NVIDIA driver. In that situation, only the CPU is detectable.