Înapoi la articole →

Cel mai recent articol

Ultima verificare

Urmăritori 0 urmăritori
Articole: 19
  1. Some interesting apps I've published on the Play Store now that I'm choosing to allow my account to be terminated (for my own safety)

  2. Overlapping waveform splitting in two places (C++ and Javascript) for multi-threaded Demucs in the browser

  3. I finally ended up with the monorepo I always dreamed about, through an organic evolution of my projects.

  4. Recapping major updates to freemusicdemixer.com this year!

  5. Trying (and mostly failing) to improve my 2021 neural network in 2023: automatic mixed precision, bfloat16, TF32, ONNX, TensorRT, nested tensors, weight pruning, and more

  6. How I adapted the Open-Unmix neural network PyTorch inference in C++ for WebAssembly with a custom streaming LSTM, mixed-integer quantization, and Demucs waveform segmentation

  7. Why the Canadian federal government is trying to ban an educational wireless hacking device

  8. Why do so many FFT and STFT libraries have different outputs?

  9. Ignorance is BLISs

    Exploring different BLAS library backends (OpenBLAS, AOCL BLIS, Intel MKL) for C++ Eigen project demucs.cpp on my AMD Zen 3 5950X.

  10. Tracking the evolution of my GitHub portfolio from 2014 to 2022, and how I chose new projects to work on

  11. Welcome

    Introduction to my new blog

  12. Speed is the product

    Exploring SIMD and auto-vectorization in WebAssembly to accelerate the C++ Eigen demucs.cpp code in freemusicdemixer.com.

  13. Introducing this pragmatic load balancing algorithm and describing a real world use case

  14. How I Python

    Describing my Python dev setup (virtualenvs, etc.) as someone who has had to work with virtualenv/pip and conda side-by-side

  15. I finally fixed an old bug in pitch-detection, my favorite among all of my open-source projects

  16. I finish projects by starting them

  17. Describing new ideas for managing my dotfiles for work-life balance and separation of concerns, and some tips on my personal and workplace learning philosophies

  18. How I adapted the Demucs neural network PyTorch inference in C++ for WebAssembly, with a more complicated architecture than Open-Unmix

  19. GEMM convolutions, debugging, and performance improvements to implement Demucs inference in client-side WebAssembly