atoshev

@atoshev

Joined on Dec 8, 2021

  • Open TODOs as soon as the solver is stable enough, publish to MLOSS Clones to remove (1):[x] some scripts on Mac code/jax-sph [x] lundquist git/jax-sph has SitL script - archived [x] ??? mikoto code/iclr_workshop_2024/jax-sph [x] working dir: mikoto code/tmp/jax-sph - can be safely deleted [x] LagrangeBench dataset code on mikoto code/sph-dataset-jax - on github References phiFlowworkshop paper
     Like  Bookmark
  • Zoom Link: https://tum-conf.zoom-x.de/j/68020718644?pwd=Q0J5bHFOZGttM3V2SzR0cTNNNGRIQT09 References GNNs Videos [ ] Stanford CS224W: Machine Learning with Graphs | 2021 - The all-time best Stanford course to GNN. I started learning GNNs with these videos, but later I realized that 90% of the theory/vocabulary is not really needed for our physical applications. [x] 1. Theoretical Foundations of Graph Neural Networks - all you need to know on GNN theory. [x] 2.1 Three part series on GNN Theory (Part 1, Part 2, Part 3) - more in-depth version of the above.
     Like 1 Bookmark
  • There are many ways to get a job done using Python, i.e. different IDEs (Colab, VS Code, PyCharm, Spyder, etc.), different libraries (numpy, PyTorch, JAX, etc.), different hardware (CPU, GPU, TPU, parallel CPU/GPU/TPU, etc.). This brief overview should help you find the best setup for given your hardware and operating system. GPU vs CPU If your hardware does include a CUDA GPU (How to know if my GPU supports CUDA), then you should probably use it. Note, MacBooks do not have CUDA GPUs and Windows machines require a somewhat different setup than we will present here (e.g. Installing Pytorch with CUDA support on Windows 10; our setup during the exercises is based on Ubuntu 20.04., e.g. this tutorial. If you do not have a CUDA GPU, you should consider using Google Colab with GPU runtime, see How to use Colab. Why GPU? By moving the core computations to the GPU, the training of deep learning models can be accelerated with up to one order of magnitude. With PyTorch it is relatively straightforward to implement that. You will often see lines line device = torch.device('cuda:0' if torch.cuda.is_available() else 'cpu') and inputs, labels = data[0].to(device), data[1].to(device).
     Like  Bookmark