Deploying ML Models in Production: An Overview

Подписаться 47 тыс.

Просмотров 42 тыс.

50% 1

The deployment of ML models in production is a delicate process filled with challenges. You can deploy a model via a REST API, on an edge device, or as as an off-line unit used for batch processing. You can build the deployment pipeline from scratch, or use ML deployment frameworks.
In this video, you'll learn about the different strategies to deploy ML in production. I provide a short review of the main ML deployment tools on the market (TensorFlow Serving, MLFlow Model, Seldon Deploy, KServe from Kubeflow). I also present BentoM - the focus of this mini-series - describing its features in detail.
=================
1st The Sound of AI Hackathon (register here!):
musikalkemist.github.io/theso...
Join The Sound Of AI Slack community:
valeriovelardo.com/the-sound-...
Interested in hiring me as a consultant/freelancer?
valeriovelardo.com/
Connect with Valerio on Linkedin:
/ valeriovelardo
Follow Valerio on Facebook:
/ thesoundofai
Follow Valerio on Twitter:
/ musikalkemist
=================
Content:
0:00 Intro
0:36 ML deployment strategies
1:32 Basic ML deployment
3:27 Disadvantages of basic ML deployment
4:57 Overview of ML deployment tools
9:54 BentoML
14:00 What's next?

Наука

Опубликовано:

2 авг 2024

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 22

@Erosis 2 года назад

ML Ops is such a rarely taught topic in academia. Thank you so much for this explanation. I hope there is more to come!

@ValerioVelardoTheSoundofAI 2 года назад

Indeed. Most of the time, academia is focused only on the research side. Productisation is mainly done in the industry.

@fcojperez 2 месяца назад

Good summary of all different ML "framework" (if we may agree on that) to deploy a ML model. Thank for that 🙏

@stefanvasilev8948 Год назад

Thank you very much! This was the most helpful resource I've found on the topic!

@travelwithak_ Год назад

Thanks very much for the detailed explanation. Really helpful👏

@Diego0wnz 2 года назад

Almost starting as a data science consultant and I really need this video series!!! Comes at perfect time

@ValerioVelardoTheSoundofAI 2 года назад

Nice! Good luck with the new enterprise :)

@maxwohrle1049 2 года назад

Great video as always. You really helped me a lot with my master thesis. This video series is yet another important building piece in dealing with audio machine learning projects. If I may add a comment. The Mnist dataset is a good example to demonstrate the deployment, but since the channel is called "the sound of ai" it would also be very interesting to see how to deploy a model that classifies an audio stream in real time.

@ValerioVelardoTheSoundofAI 2 года назад

Thank you Max. I will cover audio classification in realtime in a future series. I'd like to keep this mini-series is domain agnostic.

@AbdurrahimKururu 3 месяца назад

this is very helpful thanks

@arumugam2615 2 года назад

Great ...

@ahmedgoniboulama8423 Год назад

Hello, great video, well explained, i would like to know if BentoML can be used to Deploy MeshSegNet models?

@Coolblockj Год назад

Hey! I've been trying to work on deploying some complex models and wondered if you had any experience in industry tools that do this well The scenario is essentially I have two models that need to be run, but one of the features they use is a completely separate model. Is there any industry standard software you know of to make that easier to do in a more scalable way? And if I eventually add more models that then depend potentially on those?