Reflections from Previous Training Events

Reflections From Previous Training Events

Contents of this documents and quicklinks：

1. Julia

deeper HPC related issues with Julia
developing programs locally that targets HPC i.e. use of Docker and/or AppTainer
Writing Julia for Python programmers
ML using Julia (SciML)
- how about split ML part in Julia workshops into a separate ML-using-Julia lesson
- Francesco?
Julia for data scientist and review of some packages for CAE and other demanding projects (ray tracing, CFD machine learning …)

1. Python HPDA
- If we split them, we could also cover more things about big data storage and retrieval (S3, databases, tips on parallel file systems…) (Francesco)
- data management
- data storage
- database
1. Python HPC
- it would be beneficial to have comparison of all parallelization methods and best use cases as a summary
- like an overview of common data formats
- Someone in the feedback mentioned having a KB of tips and tricks, sounds interesting! (Francesco)
- ashwin!
1. Python ML/DL
- Practical machine learning
- Practical deep learning
1. a blogpost about data format/management in all eurohpc systems

lesson material practical deep learning from csc

theory on ML and DL at a deeper level
https://github.com/csc-training/intro-to-dl
more ML algorithms/applications
- workshops at intermediate level
model optimization on AMD GPUs
how to make use of GPU resources
measure the performance of network
- performance issues
it will be interesting to have a workshop to learn to use alpha fold
- https://liu.se/en/news-item/liu-forskare-far-alphafold-att-forutse-valdigt-stora-proteiner
- another framework jax
advance DL on specific models and application on prace machines for instance
- rotations about CASTIEL2 course
wrap models in web server (Flask, FastAPI, etc) ONNX Runtime Sagemaker/Google Cloud ML/Azure ML
- Example for using ONNX with KB-whisper: https://whisper.mesu.re/ | https://github.com/PierreMesure/whisper-web
explainability on different models, further hands-on with new model architectures, best practices about project structuring
ML/AI (tensorflow/PyTorch)
DL transformers
AI for chemistry
- maybe we can arrange more events from AI Factories perspective like AI4Science
Comparing different Deep-Learning techniques on the same dataset, training in parrallel and on a single CPU
- start from webinars focusing on regression problems, classification problems, etc.

OpenACC-CUDA
- arrange one workshop at H2
MPI
- piggy-back with our collaborators
OpenMP offloading
OpenMP with HIP
Hybrid programming
CUDA-Aware MPI
Modern C++ for scientific computing
Profiling and code optimization
Solving specific physical problems
- we may consider to write blogposts to solve specific problems using varied programming models
- like DFT, MD, CFD computations
Profiling for bottlenecks
Programming FPGA
Programming using other accelerators
SYCL concurrent programming all system resources (CPU+GPUs from multiple vendors+accelerators…
RUST programming
GPU toolchains and libraries