Create & Deploy A Deep Learning App - PyTorch Model Deployment With Flask & Heroku

Create and Deploy your first Deep Learning app! In this PyTorch tutorial we learn how to deploy our PyTorch model with Flask and Heroku.
We create a simple Flask app with a REST API that returns the result as json data, and then we deploy it to Heroku. As an example PytTorch app we do digit classification, and at the end I show you how I draw my own digits and then predict it with our live running app.
~~~~~~~~~~~~~~ GREAT PLUGINS FOR YOUR CODE EDITOR ~~~~~~~~~~~~~~
✅ Write cleaner code with Sourcery: sourcery.ai/?... *
📚 Get my FREE NumPy Handbook:
www.python-engineer.com/numpy...
📓 Notebooks available on Patreon:
/ patrickloeber
⭐ Join Our Discord : / discord
If you enjoyed this video, please subscribe to the channel!
Code:
github.com/patrickloeber/pyto...
Links:
Flask installation:
flask.palletsprojects.com/en/...
PyTorch course & code:
• PyTorch Tutorials - Co...
github.com/patrickloeber/pyto...
PyTorch:
pytorch.org/
Heroku:
www.heroku.com/
You can find me here:
Website: www.python-engineer.com
Twitter: / patloeber
GitHub: github.com/patrickloeber
#PyTorch
----------------------------------------------------------------------------------------------------------
* This is a sponsored link. By clicking on it you will not have any additional costs, instead you will support me and my project. Thank you so much for the support! 🙏

[@patloeber]

If you build your own ML/DL web apps after this tutorial, feel free to share your GitHub repo here :)

[@IAmTheMainCharacter]

will do,

[@sarahjamal86]

Can’t wait to finish work and do this video 😍

[@patloeber]

Nice! I hope you like it!

[@tafikelmehdi2913]

A covid-19 prediction using chest x-ray images webapp : github.com/Avant-Gardiste/Machine-Learning-Portfolio/tree/master/Deep%20Learning/Detecting%20COVID-19%20with%20Chest%20X-Ray

[@krinodagamer6313]

ok

[@JosephRivera517]

This is really a great help for beginners.

[@patloeber]

thanks!

[@HassanHayatch]

Thanks, your tutorials are really amazing.

[@patloeber]

Thanks :)

[@muralikrishnapolipallivenk2572]

Thank you very much, Right On Time when I need it

[@patloeber]

You're welcome :)

[@emadgirgis6689]

Amazing work. Thank you.

[@patloeber]

Glad you like it :)

[@arshadAlieusafzay]

excellent tutorial. many thanks

[@patloeber]

glad you like it :)

[@yevg3907]

Great stuff surprised you didn’t finish your stuff with a simple front end maybe next part

[@patloeber]

Yes it was mainly about deploying a REST API, but maybe next time with Frontend

[@HighInFlavour]

Hi, do you have any tutorial on integrating a PyTorch chatbot into HTML? I followed your Chatbot with PyTorch tutorial (really great tutorial) and I have an HTML template which receives user input. How do I converse with the chatbot inside the html file instead of the terminal?

[@2LazySnake]

Thank you very much!

[@patloeber]

glad you like it!

[@2LazySnake]

@@patloeber senpai noticed me!

[@raminessalat9803]

I felt that I dont have the background to learn this tutorial, can you suggest some links to learn flask and horuku? cuz this was the first time i was hearing them

[@firuzjuraev5383]

Mind-blowing. Just amazing tutorial

[@manishsharma2211]

Thank you. I love how you say "Hey everybody"😌😌😌

[@patloeber]

Haha thanks :)

[@gurjeet333]

Hi, thanks for the wonderful video, it would be better you u can develop a complete UI experience rather than coming to terminal to view the predection

[@taneryilmaz6171]

I wonder can we extract data from pdf math graph useing pytorch CNN model automatical (without human help) ? thanks in advance.

[@user-hr5dc4ly3q]

you have very good knowledge in python and json so which books and websites was your reference to be master in python

[@mr.nobody8079]

flask and django both needs different model formats?........ plz explain........

[@bingsenlim3121]

just in case anyone here is using Windows, gunicorn does not work on windows. You'll get an error ModuleNotFound:fcntl because gunicorn uses fcntl. You can't run pip install fcntl on Windows because fcntl is only for Unix-based OS

[@gbtech8602]

@so beautiful memory Thank you, this help me

[@owenlie]

@so beautiful memory Thank you, this helps me.

[@souvikbiswas284]

@so beautiful memory many thanks..

[@alisonjmarch]

@@sobeautifulmemory6170 Thank you, saved my life, was frustrated about it!!

[@masoudrahimi3696]

this toturial is great and I have watched until here. But, this one can be improved.

[@aryankhatana]

Can you do a video on deploying an RNN model using Flask, please?

[@patloeber]

My next videos will be about RNN

[@alexnim4873]

hi, you mentioned at the end that the normal torch and torchvision is too large for heroku, hence you had to select a smaller version. what can i do if i need the full version of torch and/or torchvision for my app on heroku? will paying for a heroku account fix this problem? I'm building a web app that uses BERT, an NLP library that is built off pytorch - but heroku is telling me that the in install exceeds the 500MB slug size. Or perhaps there is a lightweight version of BERT out there... i need to install "sentence-transformers==0.3.9" from my requirements.txt file Either way, thanks for the great video :) appreciate it

[@RC-qi6hs]

What did you do then, im also facing a similar situation. Thanks

[@alexnim4873]

@@RC-qi6hs hi, it's been a while since i worked on this. but i believe i installed the following version of torch and torch vision in my requirements.txt torch==1.6.0+cpu torchvision==0.7.0+cpu this means that these versions will only process your data using cpu resources this should be fine as inference does not really need a gpu to process the data unless you're processing exceptionally large amounts of data - in this case i'd move over to docker hope this helps!

[@kyomdonalddogo5775]

Thank you sir for your video. Sir please can I ask you tips on how to deploy a seq2seq model using flask only? Thank you

[@patloeber]

thanks! I'll have a look into that

[@kyomdonalddogo5775]

@@patloeber sir thank you for your response. I need your help sir. I have made pytorch models, but loading them to be used for production, I get errors. Sir please help me with this

[@afdp3849]

how about if the model is tranfer learning?

[@ibobak]

Thank you for a nice tutorial. But I can't understand one thing: why VSCode is so stupid that it doesn't see your typos? I am using PyCharm and it is highlighting them for me just immediately....

[@IAmTheMainCharacter]

can you suggest some project? I have 8 months of time for one project.

[@patloeber]

Whatever interests you the most. I like object detection tasks where you learn more about CNNs. If you have enough time you can also build a nice frontend for your app.

[@abhaytiwari6401]

Great Video Kind Sir, I want to make a movie recommendation app I get the data using tmdb API i want to take input from users like two or three movies they like and recommend similar movies using a ML Model Can you suggest few things 1. How should i proceed with this , Get a csv file from Tmdb api and get similar movies using collaborative filtering 2. which ML library i should use for this or which ML algorithm ?

[@patloeber]

great project! I suggest using either scikit-learn and testing different classifier, or PyTorch if you like to try Deep Learning

[@souvikbiswas284]

i cannot use gunicorn on my windows PC. what is the alternative for that?? edited : you can use "waitress-serve" inplace of "gunicorn". you have to install waitress first

[@alisonjmarch]

What is the line you put inside Procfile? I am experiencing the same issue, and have waitress installed, however I cannot figure out the correct code format inside Procfile.

[@giftchidi8774]

@@alisonjmarch were you able to get past it

[@1UniverseGames]

"Using backend: pytorch". How can I get rid of such thing when run python code, any help

[@saurrav3801]

Bro ....I search everywhere but I didn't get any satisfied result about ..how to implement KNN in pytorch.....

[@patloeber]

PyTorch main focus is on deep learning and neural nets, so for KNN other frameworks might be better suited. But you can implement it with tensors following my machine learning from scratch tutorial about KNN

[@alisonjmarch]

How do you run gunicorn on Windows System? I am lost after trying it on my Windows laptop and vscode and running heroku local won't work, it says 'gunicorn' is not recognized as an internal or external command, operable program or batch file.

[@alisonjmarch]

So what will be the correct line to add if using waitress instead of gunicorn in the Procfile?

[@giftchidi8774]

@@alisonjmarch having the same problem now. I'm completely stucked

[@nobody2937]

```heroku login -i``` brought me : Error: Your account has MFA enabled; API requests using basic authentication with email and password are not supported. Please › generate an authorization token for API access. Any further suggestions?

[@ikbalnayem8914]

"Process running mem=645M(126.0%) Error R14 (Memory quota exceeded)" Why my application does not release the resource on heroku? But in my local machine it does not consume much memory

[@souvikbiswas284]

did you installed just the cpu only part for torch and torchvision??

[@kimm8941]

34:00

[@anounTT]

oof, I enjoyed watching you troubleshoot. Too bad it was only typos. You don't see that in many tutorials.

[@puneetwalia8641]

it isnt very much accurate with the paintbrush

[@NobodyIikesyouduh]

This lecture is most confusing in the pytorch series due to many screen switches. I suggest to rerecord this by explaining the already written code.

[@patloeber]

thank you for the feedback! will try to improve this

[@naknakn7807]

you're so cute ;)

[@patloeber]

Thanks :)