Weights & Biases: Automated Tracking, Logging, and Collaboration for Research

We cordially invite you to our Weight and Biases practical workshop organized by Saarland University. The workshop will be held online on November 27th from 14.00 to 16.00.

Register here: https://forms.office.com/e/tXneWeKqTy

Automated tracking and logging are essential practices for well-executed research, and becoming proficient in these areas can significantly enhance your research productivity. In this introductory practical workshop, we will introduce you to Weights & Biases, the machine learning platform that can support your research. In addition to tracking and logging your experiments, Weights & Biases will help you identify issues with your models during both training and testing, enabling you to gain deeper insights into your parameters and monitor changes in your models and datasets.

During this workshop, we will focus on the beginner-level CIFAR-10 classification task and guide you on how to leverage Weights & Biases in your model training and testing. We will utilize Weights & Biases for hyperparameter searches, and as a conclusion, we will explore how to integrate Weights & Biases into your interactions with large language model APIs. We invite you to join us for an introduction to Weights & Biases. Enhance your research capabilities and productivity with this valuable workshop.

Looking forward to seeing you at the workshop,

by Israel A. Azime and Paloma García de Herreros García

Contents

  1. Introduction to CIFAR-10 Classification Task
  2. Setting up W&B
    • Installing W&B and Creating an Account
    • Experiment Logging and Alerts
    • wandb.init
    • wandb.log
    • wandb.watch
    • Visualizing Predictions
  3. Exercise 1
  4. Parameter Sweep, Hyperparameter search, and Artifacts
    • Exercise 2
  5. LLM Monitoring