Zenobot: Flask API, React Front End, LSTM Proverb Generator


(Garrett W) #1

Hi all,

https://zenobot.garrettwatson.io

I’ve been stuck on the silly idea of creating a simple proverb generator and serving it up for my portfolio since I first started the fast.ai part 1.

Thought I’d share it with you all since I just finished it last night! I hit a surprising number of snags trying to save the model for CPU inference, and then the front end was a challenge for various reasons. I’m excited to move on to my next project :smiley:

I’m thinking of writing a tutorial on Medium if that’s something anyone’s interested in. My code isn’t too pretty, but it’s been an amazing learning experience! Thanks to Rachel, Jeremy et Al for this incredible course, and thanks everyone for any feedback!


(Theodoros Galanos) #2

Seems pretty cool, thanks for sharing! Would you feel like sharing the code for us? If not, the medium article would be valuable. Deploying a model like this is one my biggest gaps at the moment, as I’m sure the same applies to many more of us.

Kind regards,
Theodore.


(Garrett W) #3

Sure! It’s on my github, here:
https://github.com/garrowat/zenobot

It uses the flask-classful library, whose documentation is a joy to read, personally.

I figure I should write a walk-through because, at least for me, it was tough getting the model exported and imported properly before I could serve it up.

Thanks for your interest :slight_smile:


(Jonathan Mist) #4

Congratulations! I’m definitely interested in a walkthrough of getting inference to run on a CPU