Accessing LUMI
Before accessing LUMI you need to be added to a project in My CSC with your CSC account. Access to LUMI is currently only through the command line interface, with access provided over SSH. Users using LUMI should have some experience with the Linux command line and familiarity with HPC systems. To log on to LUMI, you will first need to add your ssh key on your My CSC profile section.
The LUMI documentation offers a guide for creating SSH keypairs from which you should follow the instructions labeled "With a Finnish allocation".
Note
It might take a couple of hours until your public key is distributed to all LUMI nodes.
Warning
The private key should never be shared with anyone. It should only be stored on your local computer. Otherwise, anyone can steal the SSH key and impersonate you.
Once your public SSH key is available on all login nodes, you can access LUMI using
The LUMI documentation provides further information on how to connect to LUMI. For an introduction to LUMI, its architecture, modules and software stack, take a look at the LUMI training materials.
Storage
LUMI uses a Linux based parallel filesystem, utilising lustre. Access to different parts of the file system are based on the projects you are associated with. LUMI's storage areas and their intended use cases are described here.
Configuring the environment
To run quantum circuits on LUMI we need to load certain Python modules. Dependencies are managed via modules, which can be loaded.
To load the required modules, after connecting to LUMI, run the following lines in your terminal
This commands can also be added to your .bashrc to save repetition. In case you are curious, you can explore other available modules via module avail.
Using a custom Python environment
Loading helmi_qiskit or helmi_cirq on LUMI comes with a preconfigured Python environment for submitting jobs to Helmi. This environment can be replicated by users, by following the LUMI documentation on installing python packages. To access Helmi you will still need to load the helmi_standard module.
| Package | Version |
|---|---|
| iqm-client | >=20.0 < 21.0 |
| qiskit-iqm | >=15.1 < 16.0 |
| cirq-iqm | >=15.1 < 16.0 |
Newer versions of the above Python packages can be installed and may work with Helmi, however these are currently unsupported and may lead to errors.
As an alternative to the above, Python packages can be installed on top of helmi_qiskit or helmi_cirq into the Python user install directory by specifying python -m pip install --user whatsapp. This, however, may lead to increased dependency conflicts.
Submitting jobs
Once you have access to LUMI and have configured your environment you can learn how to run jobs on Helmi with the Running on Helmi.