Deploying Amazon Chronos Using Lambda Web Adapter

This repository provides a simple guide to deploying Amazon Chronos, a time-series database, using AWS Lambda Web Adapter. By using Lambda, you can run Chronos without worrying about servers. This setup is cost-effective and scales automatically, making it easier to manage your time-series data in the cloud.

Key Dependencies:

• Amazon Chronos: https://www.amazon.science/code-and-datasets/chronos-learning-the-language-of-time-series
• Lambda Web Adapter: https://github.com/awslabs/aws-lambda-web-adapter
  • Docker: https://www.docker.com/
  • FastAPI: https://fastapi.tiangolo.com/

Other references:

• Overview: Get the Future by Chronos of Amazon's Timeseries FM

Files

• main.py: Implementation Using Chronos and Example with FastAPI
• test_main.py: Testing FastAPI with Pytest
• live-demo-scripts.md: Cheat Sheet for Live Demo
• demo_seq.csv: Time-series data for demo

Getting Started / Local Development

Prepare:

• Python3 (3.12 or above)
• Docker (CE 27.1 or above)

git clone https://github.com/ma2shita/amazon-chronos-with-lambda-web-adapter
cd amazon-chronos-with-lambda-web-adapter
python3 -m venv chronos-libs
source chronos-libs/bin/activate
pip install -U pip
pip install "huggingface_hub[cli]"
pip install git+https://github.com/amazon-science/chronos-forecasting.git
pip install uvicorn fastapi pandas matplotlib
pip install pytest httpx
huggingface-cli download amazon/chronos-t5-tiny
pytest test_main.py
#=> Congrats! You can edit `main.py` and `test_main.py`.
uvicorn --reload --port=8080 main:app
#=> curl localhost:8080
#=> curl -X POST -H "Content-Type:application/json" -d '{"historical":[1,2,3,4,5,6,7,8,9,10,11,12,13,14,15]}' localhost:8080/forecast.json

Continuation

cd amazon-chronos-with-lambda-web-adapter
source chronos-libs/bin/activate
pytest test_main.py

Build container then push to Amazon ECR

## Prepare
aws ecr create-repository --repository-name ${REPO_NAME} # Can be made in the management console.

## Build
docker buildx build --target prod_runner .
#=> You can enter container: `docker run -it --rm ${NEW_IMAGE_ID_AND_TAG} /bin/bash`
docker tag ${NEW_IMAGE_ID_AND_TAG} ${ECR_URL_AND_TAG}
aws ecr get-login-password | docker login --username AWS --password-stdin ${ACCOUNTID}.dkr.ecr.${REGION}.amazonaws.com
docker push ${ECR_URL_AND_TAG}
## NEXT: Create or Update a Lambda function based on this container

## Tips: Update a Lambda function via AWS CLI
aws lambda update-function-code --function-name ${MY-FUNCTION} --image-uri ${ECR_URI}
while true ; do aws lambda get-function --function-name amazon-chronos-t5-tiny_fastapi0 | jq -r .Configuration.LastUpdateStatus ; sleep 5 ; done
# waiting for `Successful`

References

Chronos

• Chronos: Learning the language of time series
  • amazon-science/chronos-forecasting
• Google Colabで時系列基盤モデルを試す

Tools and Developments

• huggingface
  • Manage huggingface_hub cache-system Learn the Cache system on Huggingface
  • Command Line Interface (CLI) huggingface-cli
• Amazon ECR and Lambda Web Adapter
  • docker/library/python
  • awsguru/aws-lambda-adapter
  • SinatraをAWS Lambdaで動かす (Lambda Web Adapter 0.8.1 使用) Step by Step for Lambda Web Adapter (It's own content)
• Docker
  • マルチステージビルドの利用 Multi-stage build for Docker
  • JSONArgsRecommended Learn the CMD, ENTRYPOINT
• FastAPI
  • FastAPI

Author

This project is maintained by Kohei "Max" MATSUSHITA.
If you have any questions or suggestions, feel free to reach out via GitHub.

Contributing

We welcome contributions to this project! Here’s how you can help:

1. Bug Reports: If you encounter any issues or bugs, please feel free to submit a bug report. We appreciate your help in making this project better.
2. Pull Requests: You can directly create a pull request to contribute code or features. There’s no need to create an issue first unless you want to discuss the changes before starting.
3. Coding Style: There is no specific coding style required. Just make sure your code is clear and well-structured.
4. Testing: Please ensure that your code is tested before submitting a pull request. This helps maintain the stability and reliability of the project.
5. Documentation: There are no specific documentation requirements. Feel free to add documentation if you think it’s necessary, but it’s not required.
6. Review and Merge Process: We follow a simple pull request and merge process. Your contributions will be reviewed and merged if everything looks good.

Thank you for your contributions!

License

This project is licensed under the MIT License - see the LICENSE file for details.

EoT