王菲大片充满灵气 天后随性风格贯彻始终

Question

I'm trying to run quantum circuit simulations with Qiskit Aer using GPU acceleration (specifically an NVIDIA A100) in Google Colab Pro.

Here's a simplified example I'm running:

from qiskit import QuantumCircuit, transpile
from qiskit_aer import AerSimulator

qc = QuantumCircuit(2)
qc.h(0)
qc.cx(0, 1)
qc.save_statevector()

simulator = AerSimulator(method='statevector', device='GPU')
circ = transpile(qc, simulator)
result = simulator.run(circ).result()
state = result.get_statevector(0)
print(state)

This works when I use device='CPU', but when using device='GPU', I frequently get the error:

RuntimeError: Simulation device "GPU" is not supported on this system

I'm using Google Colab Pro with a confirmed A100 GPU available (!nvidia-smi confirms this). I have installed Qiskit Aer with GPU support using:

!pip install qiskit-aer-gpu

Still, I often lose GPU access after reconnecting or restarting the runtime.

---

Questions:

1. How can I ensure that Qiskit AerSimulator consistently uses the GPU (A100) in Google Colab?
2. Are there any CUDA or system-level dependencies required beyond the pip package?
3. Is there a reliable workaround for this GPU support issue in Colab?

---

Environment:

• Google Colab Pro
• GPU: NVIDIA A100
• Python: 3.11
• Qiskit: latest (August 2025)

Any insight or working configuration would be greatly appreciated!

Answer 1

I found the solution! ? Since Google Colab often runs in a CUDA-enabled GPU environment, the fix is to install the GPU-compatible version of Qiskit Aer:

!pip install qiskit-aer-gpu-cu11

Additional note: I also noticed that sometimes Colab connects you to a server that doesn't allow executing these instructions properly—even with the correct packages installed. I didn’t change anything in the code; I just restarted the Colab runtime and reinstalled the GPU version of Aer, and it started working again.