Course Wiki

Preparing a Docker image — Machine Learning Zoomcamp Module 9

Refer to updates.md for info on running TF lite in 2024.

Machine Learning Zoomcamp › Module 9: Serverless Deep Learning › Preparing a Docker image

Notes

Refer to updates.md for info on running TF lite in 2024.

Using pip install for TF-Lite binaries

When using pip to install the compiled binary, make sure you use the raw file, not a link to the github page.

Correct:

pip install https://github.com/alexeygrigorev/tflite-aws-lambda/raw/main/tflite/tflite_runtime-2.14.0-cp310-cp310-linux_x86_64.whl

(Note /raw/ in the path)

Also correct:

pip install https://github.com/alexeygrigorev/tflite-aws-lambda/blob/main/tflite/tflite_runtime-2.14.0-cp310-cp310-linux_x86_64.whl?raw=true

The wheel file above is for Python 3.10. Check other available compiled TF lite versions here.

Not correct - won’t work:

pip install https://github.com/alexeygrigorev/tflite-aws-lambda/blob/main/tflite/tflite_runtime-2.14.0-cp310-cp310-linux_x86_64.whl

If the file is incorrect, you’ll get an error message like that:

zipfile.BadZipFile: File is not a zip file

ENTRYPOINT vs CMD

This link explains the difference between them: https://stackoverflow.com/a/34245657

ENTRYPOINT specifies a command that will always be executed when the container starts. CMD specifies arguments that will be fed to the ENTRYPOINT.

In case of the lambda base pacakge, the authors already specified the entrypoint and we only need to overwrite the arguments passed to the entrypoint,

• Notes from Peter Ernicke

Key concepts

• Docker

Related notes

• mlz-m01-setting-up-the-environment
• mlz-m05-deployment-to-the-cloud-aws-elastic-beanstalk-op
• mlz-m05-environment-management-docker
• mlz-m05-intro-session-overview
• mlz-m05-python-virtual-environment-pipenv
• mlz-m09-api-gateway-exposing-the-lambda-function
• mlz-m09-aws-lambda
• mlz-m09-creating-the-lambda-function
• mlz-m09-explore-more
• mlz-m09-introduction-to-serverless
• mlz-m09-preparing-the-code-for-lambda
• mlz-m09-python-3-12-vs-tf-lite-2-17
• mlz-m09-summary
• mlz-m09-tensorflow-lite
• mlz-m10-creating-a-pre-processing-service
• mlz-m10-introduction-to-kubernetes
• mlz-m10-overview
• mlz-m10-running-everything-locally-with-docker-compose

Sources

• Video
• Lesson file