Programmatically access your machine learning system of record
Accessing your model weights, metrics, hyperparameters, images, and other workflow artifacts should be easy for 10 or 10,000 experiments. Our new Python API client allows you to programmatically access your Comet.ml workspace, project, and experiment data. By using the API Client, you can reduce the amount of code you write to interact with our REST API 👍🏼
With programmatic access to your experiment data, you can:
- Connect Comet.ml to your CI/CD system and automatically deploy to production the best model
- Fetch a model to continue training or for transfer learning purposes.
- Run meta analysis tests such as hyperparameter importance
- create variations of plots (like our average performance plot above)
- build custom reports around your model results
See this interactive tutorial of the Comet.ml Python API Client:
Not using Python? No problem!
The API is fully available as standard, language agnostic REST endpoints. Please see full specification here.
Liked this article? You might find these other articles useful:
- Comet.ml Release Notes — updates of new features and fixes!
- Real-Time Model Performance Visualizations with Comet.ml
- Comet.ml cheat sheet: supercharge your machine learning workflow