The Best Way to Deploy AI Models (Inference Endpoints)

Рет қаралды 11,834

Күн бұрын

Unlock your AI model's full potential with serverless deployment 🚀 Dive into our comprehensive guide on deploying open-source models with Hugging Face and shape the future of AI! 💡🤖
Notebook: colab.research.google.com/dri...
🤝For all sorts of projects, reach out to me via email on the "About" page of my channel.
📞Consulting: calendly.com/vrsen/ai-project...
🐦Twitter: / __vrsen__
Intro 00:00
Understanding the Tradeoffs: Different Deployment Options 00:44
Serverless Deployment: An Efficient Solution 02:32
A Practical Walkthrough: Deploying a Model from Hugging Face 03:33
Conclusion 04:57
About: Explore the ins and outs of AI model deployment in this comprehensive video tutorial. We'll cover popular options such as cloud-based, on-premise, edge, and serverless deployments, focusing on their trade-offs in cost, latency, and scalability. Learn how to optimally deploy open-source models from Hugging Face, harnessing serverless deployment's power to unlock your AI model's full potential. Understand the future trends in AI deployment and engage in a practical walkthrough for serverless model deployment using Hugging Face's inference endpoints. Ideal for AI enthusiasts seeking to enhance their knowledge in efficient model deployment.

Пікірлер: 36

@bjorginson Жыл бұрын

Great editing and information delivery, keep it up!

@pequod4557 11 ай бұрын

Very informative video & straight to the point. Thanks for making these!

@mykhailoshchuka Жыл бұрын

Great video! The visual presentation of information is very creative and helps me get more out of this video!

@HeyFaheem 5 ай бұрын

Highly informative yet underrated

@KelvinWKiger 8 ай бұрын

Excellent explanation, thank you.

@user-nr4jp7vi2y 9 ай бұрын

Great great video!!

@udaym4204 9 ай бұрын

hats off to you Thank you.

@father_mihai 9 ай бұрын

Solid video! I've been wondering whether hugging face charges per usage minutes, or whether it charges for non-stop uptime. I couldn't find anything online in layman terms, so I stayed away from it /9didnt want to spend 500$+ per month for a small app). This finally answered the question. Thanks man!

@evanscastonguay Жыл бұрын

really good video as usual. :)

@vrsen Жыл бұрын

Thank you, Evans🙏 Next one will be better!

@user-ew8ld1cy4d Жыл бұрын

Good one VRSEN!

@vrsen Жыл бұрын

Thanks you🙏

@asdasdaa7063 6 ай бұрын

My question is i see, "After 15 minutes" it becomes serverless? Does that mean for those 15 minutes, before it becomes serverless, you will be charged? If the answer is yes then its better to use something like azure functions or something maybe?

@stefanvasilev8948 Жыл бұрын

Simple and straight to the point

@karamjittech 11 ай бұрын

Nice video, but hugging face charges for the defined minutes. Any video on replicate or any other options we have?

@MohanKumar-gj9th Жыл бұрын

Cool video animation

@Ryan-yj4sd Жыл бұрын

nice

@vintagegenious Жыл бұрын

Awesome mate, I actually needed your videos for the project I am working on right now. Do you know if HuggingFace endpoints are "private", can you host models trained on personal GDPR data ?

@vrsen Жыл бұрын

Yeah, you can make it “protected,” than only you will be able to call it with your api key

@vintagegenious Жыл бұрын

@@vrsen Nice

@Ryan-yj4sd Жыл бұрын

How to do batch inference. I have 5mm prompts I want to run. Is it possible to for reasonable cost?

@vrsen Жыл бұрын

I think you will have to run them 1 by 1 in a loop, unfortunately.

@heski6847 8 ай бұрын

спасибо

@shaonsikder556 11 ай бұрын

What are the cost of serverless deployment?

@admiralhyperspace0015 11 ай бұрын

I don't use inference endpoints because its not by the second charge, It charges as you said 15 minutes above when you stopped inference. That is fifteen minutes everytime I did not use my gpu. I have been looking at other providers just for this issue.

@vrsen 11 ай бұрын

Check out replicate