### Interesting things - [Twin Graph](https://github.com/aws-samples/twingraph) - provides a Python-based high-throughput container orchestration framework for simulation, predictive modeling and optimization workflows. - It supports asynchronous multi-host computing, dynamic directed-acyclic-graph (DAG) pipelines, and recording of custom workflow artifacts and attributes within a graph database, for repeatability and auditability. - Applications include high-throughput simulations, simulation-driven optimization, Digital Twins and machine learning. - [[Physics Informed Neural Operators]] - [[Fourier Neural Operators]] - Physics Informed Neural Networks PINNs for [[Forward modelling]] and [[Inverse Modelling]](https://arxiv.org/abs/2010.09147) - [NVIDIA Modulus](https://developer.nvidia.com/modulus) - an open-source framework for building, training, and fine-tuning Physics-ML models with a simple Python interface. - [[Wind Farm Simulations & Optimisation]] is an application that takes up a very high percentage of HPC usage for data centres. Wind Farm Layout Optimisation and Active Wake Control in Wind Farms. - [[Benchmarking Specifics ]] - [[Full Waveform Inversions]] - [[ Reverse Time Migration]] - [[Inverse Modelling]] - [[Forward modelling]] - [[Seismic data regularization and de-noising]] - [SEISCOPE Consortium](https://seiscope2.osug.fr) and [Optimisation Toolbox](https://seiscope2.osug.fr/SEISCOPE-OPTIMIZATION-TOOLBOX-274) Seismic data Higher dimensions 5 dimensions: X,Y,Z, + Source Receiver Distance + Azimuth 1 TB to 100 TB in size HDF5 DATA: https://www.hdfgroup.org/solutions/hdf5/ 5 million Parameters TransUnet https://github.com/Beckschen/TransUNet https://sh-tsang.medium.com/review-transunet-transformers-make-strong-encoders-for-medical-image-segmentation-ae2eb9d8ee76 Seismic data compression in time processing High density, single sensor Data Size per 1000 km squared is ca. 500 TB [[MPI Jobs]] Jobs segmentation into preconditioning, shot computation and reduction [[Lossless Compression]] - [NVcomp](https://developer.nvidia.com/nvcomp)https://cds.cern.ch/record/2825247/files/report.pdf [[Lossy Compression]] - ZFP, S23 and Bitcomp for snapshot storage, Tolerance is 10^-3 Compression Algos for FWI Drilling Services x Data Science Reservoir Navigation Multi-physics for single-well, cross-well and well to surface Pre-well planning Preambility variations Vertical and horizontal resistivity Water contacts in oil & gas bearing sand. ZFP compression: Opern-source software for compressed floating point arrays https://computing.llnl.gov/projects/zfp Cloaking and Holography Occa: Open Concurrent Compute Architecture https://libocca.org/#/ https://github.com/libocca/occa Centre for Exascale Discretisation https://www.exascaleproject.org/research-project/efficient-exascale-discretizations/#:~:text=The%20Center%20for%20Efficient%20Exascale,to%20run%20efficiently%20on%20future NEK https://nek5000.mcs.anl.gov https://arxiv.org/abs/2104.05829 https://github.com/paranumal/libparanumal mFEM https://github.com/mfem/mfem