https://github.com/jcaip/quickthoughts
pytorch reimplementation of qt paper
https://github.com/jcaip/quickthoughts
Last synced: 12 months ago
JSON representation
pytorch reimplementation of qt paper
- Host: GitHub
- URL: https://github.com/jcaip/quickthoughts
- Owner: jcaip
- Created: 2019-04-22T18:14:51.000Z (about 7 years ago)
- Default Branch: master
- Last Pushed: 2022-12-08T05:06:36.000Z (over 3 years ago)
- Last Synced: 2025-05-08T21:15:12.611Z (about 1 year ago)
- Language: Python
- Homepage:
- Size: 313 KB
- Stars: 9
- Watchers: 1
- Forks: 1
- Open Issues: 32
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# QuickThought
pytorch reimplementation of quickthoughts paper: https://arxiv.org/pdf/1803.02893.pdf
This is a very rough implementation atm, so feel free to reach out if you're having trouble running the code.
There's also a lot of code for some experiments I ran [here](https://jcaip.github.io/Quickthoughts/) in this repo.
### Setup
First you'll need to pull down the bookcorpus submodule and follow the instructions there to get `all.txt`.
Next move `all.txt` to the root folder and then run `make data`, which will process the data into `cleaned.txt`.
Next grab the pretrained word vectors by running `make pull-vec`.
You'll be able to train with `make train`.
The model will pull from a config file, which can be edited to tune parameters.