Author: Irene Mendez Guerra ([email protected])
Notebook | W&B interactive dashboard | W&B report
Resources:
Weights and biases (W&B) is a popular MLOps platform used to track and manage machine-learning experiments. It aims to bring order, reproducibility, and efficiency to the chaotic world of ML experimentation.
To do this W&B provides (among other specialised features):
- Experiment tracking
- Model and dataset versioning
- Visualisations
- Hyperparameter sweeps
- Collaborative environment
- Integrations with common AI-tools (PyTorch, TensorFlow, JAX, Keras, HuggingFace, etc.)
Together, these capabilities make ML research, faster, more reliable, reproducible, scalable, and collaborative.
This tutorial covers the basic functions going through:
- W&B registration and installation
- logging, downloading, and using artifacts
- logging models, metadata, and metrics
- performing hyperparameter sweeps
- visualising metrics (W&B website)
- creating a report (W&B website)