pytorch friends

  • einsum/einops: https://einops.rocks/pytorch-examples.html

    Image

  • optuna: hyperparameter optimization framework

  • tensorboard / wandb (weights and bias) / comet ml: track experiments

    Comet ML - Build better models faster

  • albumentations / Augmentor: for image data augmentation

  • MMdnn: ML framework conversion

    img

  • accelerate: for distributed computing (DDP) & mixed precision training

  • netron: for visualize network structure

    img

  • kornia: differentiable OpenCV on GPU

  • torchprofile, thop: compute parameters, FLOPS, MACs,

  • hydra

python library

  • loguru
  • numba: for parallel computing (https://thedatafrog.com/en/articles/make-python-fast-numba/)
  • joblib: multiprocessing (equivalent to parfor in Matlab)
  • carbon: share the beautiful image for your code
  • nose2: for unit test

deployment

  • MLflow; Kubeflow; Clear ML
  • https://mymlops.com/

visualization / demo

  • plotly / seaborn
  • gradio / streamlit
  • dearpygui
  • lovelyplots
  • ManimML

data labeling

  • label studio

guidelines

  • use percommit: https://towardsdatascience.com/4-pre-commit-plugins-to-automate-code-reviewing-and-formatting-in-python-c80c6d2e9f5#1bec