In this project I used a 1 layer character based LSTM to try and generate inspirational quotes. Model was trained for 7 hours with a GPU using google colab. I got a validation loss of 1.387 using categorical crossentropy.
First you'll have to unzip quotes.zip to be in the same directory as generate.py
Then you can run this to generate a quote
python generate.py
Run this command for information on the arguments
python generate.py -h
If you want to run this on your local machine you will need numpy, pandas, and tensorflow 2.
You can generate quotes on Colab here
The confident with the world and a fact to decide that the more than the part of the fact is matter
glory of humanity, but for the person but when the hand they could have to be always like the hand in the rest of harmous and distant great events to the self-striggle of the people
to read books that I'm not enjoying for all that is the experience is a part of the same concern. The Proved and Woman is a thing
like the latest rage in tattoo because we believe that others forget that it was
Quotes were taken from https://github.com/ShivaliGoel/Quotes-500K
I based my model off of
Andrej Karpathy's LSTM.