This repo contains reference examples for using the MosaicML platform to train and deploy machine learning models at scale. It's designed to be easily forked/copied and modified.
It is structured with four different types of examples:
- benchmarks: Instructions for how to reproduce the cost estimates that we publish in our blogs. Start here if you are looking to verify or learn more about our cost estimates.
- end-to-end-examples: Complete examples of using the MosaicML platform, starting from data processing and ending with model deployment. Start here if you are looking full MosaicML platform usage examples.
- inference-deployments: Example model handlers and deployment yamls for deploying a model with MosaicML inference. Start here if you are looking to deploy a model.
- third-party: Example usages of the MosaicML platform with third-party distributed training libraries. Start here if you are looking to try out the MosaicML platform with non-MosaicML training software.
Please see the README in each folder for more information about each type of example.
To run the lint and test suites for a specific folder, you can use the lint_subdirectory.sh and test_subdirectory.sh scripts:
bash ./scripts/lint_subdirectory.sh benchmarks/bert
bash ./scripts/test_subdirectory.sh benchmarks/bert