FastAPI for Machine Learning: Live coding an ML web application.

Welcome! Join us for a live workshop where you can follow along with the creator of FastAPI Sebastián Ramírez to build your very own AI image generation web application! He will outline the core components of the FastAPI web framework, and his application will leverage the newly-released Stable Diffusion text-to-image deep learning model.

We will be taking questions during the event. Please submit your question or upvote others' here:

Speakers
Sebastián Ramírez, Creator of FastAPI, Senior Staff Software Engineer at Forethought

Dr. Greg Loughnane, Head of Product & Curriculum at FourthBrain

Let us know how we're doing? We will be giving out discount codes for a selected number of people who fill out the survey:

Looking to connect with your peer learners, share projects, and swap advice? Join our AI community:

To learn more about us and signup for future events:

[deeplearningaicommunity]

0:22 Introduction
2:06 Sebastian Ramirez
2:48 What is FastAPI?
5:00 Live code an ML API with FastAPI
39:42 Audience questions
40:57 What development options for FastAPI web app is available, any difference in the way the app needs to be packaged for server less vs e.g AWS lambda vs ec2?
42:40 Best practices for scaling up/handling thousands of request in parallel?
47:00 What is the best approach to retrain the model we have trained outside our application?
51:00 FastAPI vs Flask for ML Web app?
59:00 Closing

[nunofarinha]

Please, don't forget to share the link to the repository with the code of this workshop. Thanks for organizing these events and making them available to everyone

[thisaintmyrealname]

FastAPI made me love backend development again.

[Lord-V]

Tiangolo. The legend.
I remember talking to him about FastAPI standards a few years ago when FastAPI was a small and upcoming project.
So great to see that the community is rapidly adopting it !

[shanyachaubey]

The way Sebastian explains such technical concepts is so awesome.

[anaareias]

being able to run code from a .py file interactively in a jupyter notebook is what I've been searching for all this time!

[FrancisRodrigues]

Thanks for explaining the ML model on Docker -- during the final minutes of this video.

[BhanuChander-SYL]

Thank you for this wonderful Live session. It was very useful and detailed :)

[WordsThroughTheSky]

FYI, at 15:08, if you're having trouble with `image = pipe(prompt)['sample'][0]` as of 2024-01-03 it's `image = pipe(prompt)['images'][0]`

[zaheeruddinfaiz]

7:29 Hi @taingolo you should get your laptop battery checked. It jumped from 6% to 37%. 😁

[mytubejaved]

Is there any sanity or logic in using this fast API? If you are using stable diffusion? Why one will not use stable diffusion directly?

[shjmpzs]

I know this maybe a silly question but at 21:47 how do I know that the "item_id" is going to be "7"?

[santhoshkumarmatlapudi]

Please reply if you see this, I am not able to connect clickhouse database from fastapi

[surajpratapsinghsayar]

Kudos to DeepLearnigAI, great guy Sebastián

[so_you_can_go]

Guys, you promised the link to the repository. Please! :)

[theflag]

ipykernel doesn't seem to be enough anymore, tells me that I need the notebook package to run interactive cells

[Ting-ChiangYi]

I get an error message on the line
pipe.to("cuda")

The message is
AssertionError: Torch not compiled with CUDA enabled

Is it possible to run this demo without a CUDA enabled GPU?

I have been googling and trying various things for several hours but couldn't get it to work.
thank you

[mohammadirfandsa]

Hi All, I am getting below error

TypeError Traceback (most recent call last)
Cell In[11], line 1
----> 1 pipe =
2 "CompVis/stable-diffusion-v1-4",
3 revision="fp16",
4 torch_dtype=torch.float16,
5 use_auth_token=token,
6 )


452 if issubclass(class_obj, class_candidate):
453 load_method_name =
--> 455 load_method = getattr(class_obj, load_method_name)
457 loading_kwargs = {}
458 if issubclass(class_obj, torch.nn.Module):

TypeError: attribute name must be string, not 'NoneType'