Problems with enabling MPI on lightning.gpu

PennyLane Help
1

Hello!

I have a problem with running the testing the library of lightning.gpu. I need to compile it and run the test to further use it in a QML model. I followed the instructions on Lightning-GPU installation — Lightning 0.42.0-dev29 documentation and was reading another post that was made in January 9, that suggested to install manually also the library of pennylane.

The main issue i am facing is that after installing everything it wont run the tests, I am getting the following error

$ mpirun -np 2 python -m pytest mpitests --tb=shorthort
================== test session starts =============
platform linux -- Python 3.10.18, pytest-8.4.0, pluggy-1.6.0
================== test session starts =============
platform linux -- Python 3.10.18, pytest-8.4.0, pluggy-1.6.0
rootdir: /home/jlopez126/qmlp-malware/pennylane-lightning
configfile: pyproject.toml
plugins: flaky-3.8.1, xdist-3.7.0, mock-3.14.1, cov-6.1.1
collecting ... rootdir: /home/jlopez126/qmlp-malware/pennylane-lightning
configfile: pyproject.toml
plugins: flaky-3.8.1, xdist-3.7.0, mock-3.14.1, cov-6.1.1

collected 3551 items
collected 3551 items

--------------------------------------------------------------------------
prterun detected that one or more processes exited with non-zero status, thus 
causing the job to be terminated. The first process to do so was:
   Process name: [prterun-IQSeC-3258155@1,1]
   Exit code:    14
--------------------------------------------------------------------------
2

This is a update, I went back and try it in a clean enviroment in python 3.10. Currently i am using Miniconda to manage my enviroments in the server. I followed the following guide Trouble installing lightning-gpu with mpi support under Python3.10 - #11 by CatalinaAlbornoz

So i ran the following for the pennylane gpu:

RUN git clone https://github.com/PennyLaneAI/pennylane-lightning.git && \
cd pennylane-lightning && \
pip install -r requirements.txt --upgrade && \
python -m pip install . && \
PL_BACKEND="lightning_gpu" python scripts/configure_pyproject_toml.py && \
CMAKE_ARGS="-DENABLE_MPI=ON" python -m pip install . -vv

I also installed by hand pennylane and just did

# Copy first pennylane from git
cd pennylane/
pip install .

The thing is that when running the tests:
image
only 9 tests pass

3

Hi @NotSamus
I talked to someone from the team and they had the following requests and suggestions:

  • Could you run a simple circuit with "default.qubit" and "lightning.qubit" to validate the installation of PL and Lightning-Qubit on your machine?
  • Could you install and test LGPU without MPI on your machine? Note that for both LGPU and LGPU-MPI, you need to set the CUQUANTUM_SDK envvar following the installation guideline.
  • Could you run make test-cpp-mpi from ./pennylane-lightning to check the installation of the core routines and libraries (e.g., custatevec, mpi, …)?
  • There aren’t any Makefile commands to check LGPU-MPI python tests. You need to run the following to check these tests: python -m pip install mpi4py and PL_DEVICE=lightning.gpu mpirun -np 2 python -m pytest mpitests --tb=short -x

I hope this is helpful.