Getting Started with W&B
Created on January 12|Last edited on January 19
Comment
Creating a Team
To collaborate using W&B, you can create a team to invite coworkers and share projects. Learn more: https://docs.wandb.ai/ref/app/features/teams
Logging Runs
Use Runs to automatically capture configurations and results from your code. Learn how to log your first Run using our Python library here: http://wandb.ai/quickstart/pytorch. For a more detailed guide about Experiment Tracking, learn more here: https://docs.wandb.ai/guides/track
Viewing your Runs
The configurations and metrics you have logged are automatically organized by W&B. You can view your logged Runs by navigating to the workspace link provided by W&B while your code is running. Learn more about Workspaces: https://docs.wandb.ai/guides/track/app
Customizing your Workspace
Each logged metric is automatically visualized in a panel on W&B. You can edit and add new panels dynamically to compare and visualize your Runs. Learn more about Workspace best practices: https://www.youtube.com/watch?v=dHyQdCon56g
Saving Insights
Use Reports to capture snapshots of your work alongside descriptions of what you’ve tried. All Reports can easily be shared using view-only links. Learn more about Reports: http://wandb.me/report-about-reports
Managing Models
You can organize model checkpoints within W&B using the Model Registry. Learn more about model management with the Model Registry and Artifacts: https://wandb.ai/registry/model
Optimizing Hyperparameters
Once your project is stable and you want to find the best configuration, use Sweeps to easily and intelligently explore the space of possible hyperparameters. Learn more about Sweeps: https://docs.wandb.ai/guides/sweeps
Evaluating Model Predictions
Find your model's edge-cases flexibly in W&B using Tables. You can log rich media (images, audio, etc.) alongside raw model predictions for each class and dynamically sort through them in W&B. Learn more about Tables: https://docs.wandb.ai/guides/data-vis
Add a comment