# KDL HPC training

## Questions

- from Peter: will they be using Jupyter notebooks?

## Slides / background

- we have 90 min for this
- needs to introduce things like "what is a CPU?", "when is using HPC useful?", "what's a virtual environment?", "what is parallelisation?", etc
- we can base this on existing material from the Bioinfo MSc and Intro to HPC slides
- should have more focus on practical applications than theoretical computer science
- will need to check with Mary/Neil that we're pitching it at the right level and not missing out things that might seem obvious to us!

### Plans:
- use the "what is a computer"(?) and "components of a computer" sections of James's slides for the MSc
- no binary content
- keep at least some of the history of computing
- probably need to expand on the components?
- need to add content on going beyond using just one computer - how an HPC cluster is structured
- some content in the HAB HPC workshop that we can use
- need to add content on connecting - ssh keys?
- NB these participants will have username and password access, so will not need to generate an SSH key!
- so cover this only in a theoretical way
- need to add content on resource allocation / job scheduling / partitions - can base on the HPC training slides

## Hands-on session

- we have 3 hours for this
- focus on common principles that will apply to other HPC systems too (e.g. probably skip OpenOnDemand, but show how to connect to RStudio/Jupyter using an ssh tunnel)
- probably Jupyter since other parts of the week will use Python
- can skip/minimise content on using the command line as there's a dedicated session for this!
- would be nice to have a relatable research example - perhaps using Singularity to run https://github.com/kingsdigitallab/kdl-vqa

### Content:

- accessing software with modules
- virtual environments?
- submitting jobs
- job monitoring, cancelling jobs
- requesting resources including GPUs
- parallel jobs
- cover multicore jobs
- don't cover MPI jobs, probably not relevant for this audience
- leave additional material for Thursday?
- Singularity
- Jupyter notebook via ssh tunnel
- don't do this right at the end, give them time to try it out

### Plans?

- need to set up a docker container for the KDL-VQA software and adapt the material to use this

- develop example scripts that are more relevant:
- text processing - look for online training material to adapt
- Bristol course?
- https://blog.jetbrains.com/pycharm/2024/12/introduction-to-sentiment-analysis-in-python/
- matrix multiplication using tensorflow/pytorch - Max has scripts for this already
- https://docs.er.kcl.ac.uk/CREATE/TRE/guides_and_tutorials/tre_hpc/