How to Deploy ML Solutions with FastAPI, Docker, & AWS
ฝัง
- เผยแพร่เมื่อ 5 ก.ค. 2024
- 👉 More on Full Stack Data Science: • Full Stack Data Science
This is the 5th video in a series on Full Stack Data Science. Here, I walk through a simple 3-step approach for deploying machine learning solutions.
More Resources:
💻 Example Code: github.com/ShawhinT/TH-cam-B...
📰 Read more: towardsdatascience.com/how-to...
🛠️ Previous Video: • Text Embeddings, Class...
➡️ Data Pipeline Video: • How to Improve LLMs wi...
References:
[1] FastAPI Tutorial: fastapi.tiangolo.com/tutorial...
[2] FastAPI + Docker: fastapi.tiangolo.com/deployme...
[3] Deploying on AWS ECS: • AWS ECS Tutorial // Am...
--
Book a call: calendly.com/shawhintalebi
Homepage: shawhintalebi.com/
Socials
/ shawhin
/ shawhintalebi
/ shawhint
/ shawhintalebi
The Data Entrepreneurs
🎥 TH-cam: / @thedataentrepreneurs
👉 Discord: / discord
📰 Medium: / the-data
📅 Events: lu.ma/tde
🗞️ Newsletter: the-data-entrepreneurs.ck.pag...
Support ❤️
www.buymeacoffee.com/shawhint
Intro - 0:00
ML Deployment - 0:33
3-Step Deployment Approach - 1:52
Example Code: Deploying Semantic Search for YT Videos - 3:21
Creating API with FastAPI - 4:31
Create Docker Image - 11:13
Push Image to Docker Hub - 17:15
Deploy Container on AWS ECS - 19:46
Testing Gradio UI - 25:54
What's Next? - 27:07
More on Full Stack Data Science👇
👉Series Playlist: th-cam.com/play/PLz-ep5RbHosWmAt-AMK0MBgh3GeSvbCmL.html
💻Example Code: github.com/ShawhinT/TH-cam-Blog/tree/main/full-stack-data-science/ml-engineering
This is such a great video, no nonsense straight to the point!
Thank you so much, such videos are really very helpful
Glad to hear :)
Thank you!
Hey Shawn, videos on FastAPI and Docker from you would be great.
Thanks for the suggestion! I'll add it to my queue :)
This is fantastic stuff as I’m pulling out my hair on this same step.
You have the right idea for the next video, but I think the next one after that is making the chat interface publicly accessible
Great suggestion Brian! There are several ways one can do this. The simplest and cheapest would be hosting it via HuggingFace Spaces: huggingface.co/spaces/launch
However, for this specific use case the most practical option would be to embed it into my Squarespace website. I'll need to do some more digging to see the best way to do that.
@@ShawhinTalebi The streamlit cloud has been my go-to so far. I am toying with creating a react front end template but would like to see what others are doing
Bro create a video for handling post and get request and multiple endpoints using fast api dockerize and ECR and aws lambda functions
Great suggestion. I'll add that to my list!
I have a DL model which takes about 5 mins and 3gb GPU to process the query and to return result. I need to handle 5 queries per minute and I have a GPU with 8gb in GCP. How can I deploy such a model without memory leakage and I should be able to use the GPU at its full potential?
How big is that model? Do you have GPU parallelization enabled?
If it takes 5 min and 3GB to do one query with parallelization, the model may be too big to meet those technical constraints.
that was an awesome video, I have a task for one click ml model deployment on aws, azure and GCP, like one click on aws and other click on azure. CAn u please guide me shortly the roadmap...!
Thanks for your comment. Sorry I'm not sure what you mean by "one click ml model deployment". Could you share more details?