Running CUDA Quantum on Colab/Kaggle

[essaathar]

Was planning on implementing a Quantum Algorithm on CUDA. Since I do not have a GPU on my local machine, I've been using Google Colab and/or Kaggle for doing CUDA C Programming so far. Can I also similarly use CUDA Quantum in Colab/Kaggle, since I cannot find anything on this. Please let me know.

[bmhowe23]

#1096 describes how to get the Python wheels working in Google Colab. Does that work for you? Or are you strictly asking about how to use the nvq++ compiler in that environment?

[essaathar]

Thank you! That works for now. Now just two questions :

• How to use the nvq++ compiler in Colab, if I choose to do CUDA QC in C++?
• Is there a tradeoff or an edge between doing CUDA QC in Python vs C++?

[bmhowe23]

I am not a Colab expert, but I am not aware of any way to run typical command-line activities in Colab, and that may limit your ability to use nvq++ on Colab. That being said, you might want to consider https://nvidia.github.io/cuda-quantum/latest/install.html#dgx-cloud if you access to the command-line interface with cloud-provided GPUs.

Regarding Python vs C++, there is some documentation here that may be useful for you. The examples show how to do similar items in both C++ and Python. The syntax is obviously different between the two languages, but both C++ and Python implementations have access to the same CUDA Quantum simulators and real QPU backends.

[essaathar]

Is it then not possible to use the C++ API on colab?

[bmhowe23]

That's correct ... I don't think it is possible to use the C++ API on Colab. However, you can use both the C++ API and Python API on Nvidia Quantum Cloud. NVQC is currently in Early Access phase; instructions on how to apply are provided on the linked documentation.