Model search (MS) is a machine learning framework that implements AutoML algorithms for the purpose of scalable model architecture search. Its goal is to help researchers and people alike accelerate their process of exploration. The exploration is aimed at finding the best model architecture for their classification issues.

The library lets you run AutoML algorithms on your data. These include automatically finding the right model architecture, finding the right ensemble for models, and searching for the best-distilled models.

You can compare different models found during the search process. You can also create your own search space to customize different types of layers in neural networks.

Project Background

• • Library: Model Search AutoML framework 
  • Author: Google AI
  • Initial Release: 2021
  • Type: Helps with classification problems
  • License: Apache-2.0 License
  • Contains: Multiple trainers, a search algorithm, a transfer learning algorithm and a database
  • Language: Python, Starlark
  • GitHub: /google/model_search
  • Runs On: Linux, Mac, Windows

Applications

• Searching for the right model architecture
• Customizing neural networks layers
• Comparing different models