September 2022 Summaries

The field of machine learning has seen significant advancements, with models like Stable Diffusion and Whisper setting new benchmarks for capabilities. This surge in adoption highlights the importance of MLOps (Machine Learning Operations) to ensure successful product development and deployment. To facilitate this, Baseten provides flexible interfaces for ML models, allowing users to set their own model interface for specific use cases. Additionally, the platform offers robust application development workflows, including the ability to back up applications to GitHub, which enables best practices for development and confidence in making changes to production apps. The team at Baseten continues to innovate with new features and hires, aiming to support users in building innovative products and services around AI.

Stable Diffusion is an open-source image generation model developed by Stability AI, which can be deployed on Baseten using the Truss package. To deploy Stable Diffusion on Baseten, a Baseten API key must be created and installed PyPi's Truss, then the model weights are packaged for production using Truss. The Truss CLI is used to push the model to Baseten and serve it from the command line with the `truss predict` command. After deployment, a helper script can be written to save and display the output as a nice image file.