Indeed / W&B Commonly Asked Questions
Answers to questions from Indeed's butterfly team.
Created on August 20|Last edited on August 20
Comment
Hello Indeed team! We wanted to ensure we shared answers to some of the questions Mario had relayed to us as a pre-cursor to the enablement session so you can focus on any follow up questions and the material at hand! Please comment / send any follow up or feedback to us in here or in Slack through our joint channel #wandb-indeed
Questions
How can I create an account and get started?
Are there any restrictions or usage limits?
How do I find a specific model in Weights & Biases?
How to find the metadata, configuration & hyper-parameters of a model?
How do I visualize/compare the results of two separate models?
How do I visualize/compare the results of an offline evaluation?
How do I visualize/compare the results of a single model's multiple hyper-parameter tuning runs?
Can we check the test result for multiple ongoing hyper-parameter tuning runs in a way humans can tell what parameters seem promising?
What can we do with W&B and what can't we do? Can we check only those things related to NN model training? How about feature distribution or preprocessing time?
Can we obtain the "percentage change" for a model's metrics against a chosen baseline? e.g. prod model baseline vs incumbent model, 5% better accuracy metric
Feedback
I want a functionality to easily share offline test results with others in an easy to understand way.
I want to see AUCs per target and give individual up-and-down percentages.
Add a comment