Weights & Biases

The system of record for machine generated code


Save everything. Stay organized.

Save your weights in the cloud with your hyper-parameters and notes. Restore any experiment back to it's original state for reproducibility and iteration.

Get started now   ❯ 

Powerful CLI & Modules

W&B stays out of your way and let's you focus on iteration

  • Keep track of your configurations
    Override config variables with our YAML file or use your existing flags
  • Describe your experiments
    Use Markdown to describe your experiment we'll store it next to your meta files
  • Store your machine generated data
    Weights files, model definitions, you name it. We'll store it securely in the cloud

Compare experiments

We keep the training logs and parse loss and accuracy.

  • Store test datasets for automated evaluation
    CI is coming soon...
  • Search across hyper-parameters
    Discover how different parameters impact accuracy
  • Reference training code in GitHub
    Go back to the original training code and restore an experiment locally anytime.

framework Agnostic

Use whatever machine learning framework you prefer.

Get Started Now

cd my_training_git_repo
pip install wandb
wandb init