qsimcirq can handle large circuit?

[CoolMLAI]

Hi
when i tried to simulate statevector, i encontered the following error.

RuntimeError: qsimcirq cannot handle 33-qubit circuits due to numpy limitations

i wonder qsimcirq can't handle large circuit.

Thanks

[CoolMLAI]

qsim development is stopped?
why no answers?

[mhucka]

qsim can handle large circuits – people have done 40+ qubits. Can you please provide more information?

1. What version of qsim do you have installed?
2. What version of NumPy?
3. What platform are you running on? (Linux, Mac, Windows, Colab, ...?)

[CoolMLAI]

Thanks for your reply.

I have send the detail of my current situation to your email.
will you please check it out?

Thanks.

[mhucka]

Thanks. By the way, in the future, please just type your answer here instead of emailing it. That will make the answer available for other users (who may also have similar questions) and serve as a record of questions and solutions.

Here I copied the email:

What version of qsim do you have installed?

Since I am using a multi GPU, I am using cuQuantum Appliance docker container. (nvcr.io/nvidia/cuquantum-appliance:25.06-x86_64)
Inside the docker, qsimcirq python package version is 0.15.0. (Not sure how to check the qsim version.)

What version of NumPy?

The Numpy version is 1.26.4 in that docker container.
(docker run --rm nvcr.io/nvidia/cuquantum-appliance:25.06-x86_64 pip show numpy returns that)

What platform are you running on? (Linux, Mac, Windows, Colab, ...?)

I am running on Ubuntu 22.04. I am working with .py file, not colab.

Other people have reported problems with using > 32 qubits with NumPy 1.x, so that is likely the reason it is failing in your environment. Is there a more recent version of that Docker container available, by any chance?

[Rabbit692]

Same issue.

And no, that seems to be the latest docker. https://docs.nvidia.com/cuda/cuquantum/latest/appliance/index.html

25.06 is the latest. @mhucka

[CoolMLAI]

Hi @mhucka
Thanks for your reply.
Docker container what i used is the latest and i tried the others but they all use numpy 1.26.4.
So could you plz let me know how i can complie the qsim in order to use multi gpu and run the code for solving the more than 33 qubits?

[mhucka]

The current instructions are out of date. I have to update them. I hope to do that this week.

[CoolMLAI]

Hi @mhucka
Is there any update?