Skip to content

Latest commit

 

History

History
33 lines (20 loc) · 1.89 KB

README.md

File metadata and controls

33 lines (20 loc) · 1.89 KB

Demo

Text Generation using 30M-param LSTM Model

The model is trained on a corpus of 9000+ articles from StackExchange Academia. The dataset was found from this archive The compute resources used in training are a single GPU (1070 Ti) for approximately 12 hours.

Model Diagram

Spoiler Image

How to Use

The model is available to try online here. However, if you want to run it on your own, the code is provided. First, run main-text-gen-rnn.ipynb. A folder should be created under models. This folder contains the model weights and the vocab. Next, if you want to convert it to Tensorflow.js format, run tfjs-convert.ipynb.

Sample Output

  1. No starting text
I recently published my first research paper and it was accepted in a peer-reviewed journal in the field of Ai. I worked in private engineering in Europe (mostly in area, linear algebra, coding, and field topics that I have never seen on a topic that is required in the field on research and also engineering). I am worried about the quality of the world at the moment, but I know that this source of my ability of doing research is for all of those fields, being a big"
  1. Starting text: "I am a"
I am a Masters student in math and thus new to academia and publishing in general. I am a coauthor on a number of recently submitted papers and had the (somewhat unreasonable) thought about it, now after reading on the blog, more proofreading. Now, to be clear, I failed a lot of emotional and open process as I can look at the methods eventually novel stuff.

My thought is that this may mean that I would also consider the" how you will somehow"