Weights & Biases

A commercial platform for experiment trackingexperiment tracking, n.: Keeping an automatic logbook of every model-training attempt — settings, data, code and results — so a team can compare attempts, repeat the good ones and see what changed. and model management, widely used in deep-learning research and industry, including at frontier AI laboratories. Its client library, wandb, logs hyperparameters, metrics, hardware statistics, media and outputs from training scripts to hosted or self-managed servers, where runs can be charted, compared and shared in collaborative reports. Further features include sweeps for hyperparameter searchhyperparameter tuning, n.: Trying out different settings for a model's training, such as learning speed or model size, and keeping the combination that performs best on data held back for checking., Artifacts for versioning datasets and models with their lineage, a model registrymodel registry, n.: A central catalog where a team stores every version of its trained models, with records of how each was made and which version is approved for use., and Weave, a toolkit for tracing and evaluating applications built on language modelslarge language model, n.: A computer program trained on enormous amounts of text to predict the words that come next. At vast scale, that simple skill lets it chat, answer questions, translate, summarize and write code.. The name alludes to the two kinds of learned parametersparameter, n.: One of the adjustable numbers inside an AI model that are tuned during training. Large language models have billions of them, and together they store what the model has learned. in a neural network, weightsweights, n.: The learned numbers inside a neural network that set how strongly each part influences the next. The finished set of weights is, in effect, the trained model itself. and biases.

Everyone on the team posts loss curves to Weights & Biases, so we can see the moment a run diverges.