I just wanted to confirm if lightning.kokkos is still an active plug-in for pennylane.
I see the blog-post with installation but it is not listed on the plugin page.
Blog-post: PennyLane goes Kokkos: A novel hardware-agnostic parallel backend for quantum simulations | PennyLane Blog
Official Plug-In List: Plugins and ecosystem | PennyLane
Looking forward to getting some clarity over its current status. Thank you!
lightning.kokkos is still an active plugin. It’s not yet on the PennyLane Install page but it is indeed an active plugin.
You can install it by running
pip install pennylane-lightning-kokkos
Let me know if you have any additional questions!
Edit: If you just run
pip install pennylane-lightning-kokkos it will work, but only with CPU (OpenMP). To have it working with GPU devices, for example, you would need to build from scratch so the compilation can configure everything regarding your local hardware. The full installation instructions are here in the documentation.
Thank you for confirming, I will install it and reach out again if I face issues with it.
Hi @CatalinaAlbornoz ,
Doing a pip install using Anaconda is giving me the following error.
Can you kindly identify what might be causing it and a solution for successful installation.
Hey @kamzam , could you please include the things above where the screenshot cuts off? We’re missing parts of the output that would be helpful here.
PFA complete error output on pip install command for kokkos.
Environment Information: Basic python 3.8 with pennylane-lightning before running pip command for kokkos. Just in case that information helps as well.
kokkos_pip_install_error.txt (9.7 KB)
Unfortunately we do not currently release prebuilt wheels for Lightning Kokkos on Windows. If you have access to WSL on Windows, you can install the pre-built Linux versions with OpenMP support directly using pip.
If you wish to manually build the project under Windows, I suggest reading the installation guide here. Since Windows natively provides a much older version of OpenMP than Linux, you may have trouble compiling with the default backend, in which case you can attempt to use the Kokkos threads backend with
-DKokkos_ENABLE_THREADS=1. Though, once again, this may not be fully supported under Windows or by Kokkos as well as the OpenMP backend.
Otherwise, if the above fails, you may be best to use the
lightning.qubit device, as this will be guaranteed to work under Windows.
Hope this helps.
Hello, what would be the gpu installation for Colab?
Hi @kevinkawchak ,
Here are the steps you can follow to use
lightning.gpu on Google Colab:
- Open a new notebook
- In the navigation bar at the top go to Runtime → Change runtime type
- Choose a GPU
- In the first cell of the notebook run
!pip install pennylane custatevec-cu11 pennylane-lightning-gpu
- Run a simple code (such as the code below) to test that everything works
import pennylane as qml
dev = qml.device("lightning.gpu", wires=2)
Note that not all GPUs are compatible with lightning.qubit so if you have questions about compatibility please let us know.
Also please note that the code above will probably run faster using
lightning.qubit on a CPU that using
lightning.gpu on a GPU. For computations under 20 qubits I recommend using
Finally, remember that Colab only gives you access to limited GPU resources so be careful so that they don’t limit your GPU capacity.
I hope this helps!
Hello, is this the only code needed in Colab to access lightning.kokkos CPU/TPU/GPUs?
!pip install pennylane pennylane-lightning-kokkos
!pip install pennylane custatevec-cu11 pennylane-lightning-kokkos
Hi @kevinkawchak, I don’t know whether pennylane-lightning-kokkos works on Colab. But please let us know if you run into any issues.