Skip to content

Benchmarking next word predictors from the Pre-Transformer Age 👴🏻

Notifications You must be signed in to change notification settings

dominik-pichler/LST-ME_KNOW

Repository files navigation

                                             ______
                                            /      \
                                           |        |
                                           |:/-\\--\.
                                            ( )-( )/,
                                             | ,  .
                                            / \- /. \
                                           | | |L / \ \
                                          / /  \/    | *
                                         / /          \  \
                                         | |      []   |\ |
                                        /| |           ||  |
                                        || |           ||  |
                                        |  |           ||  |
                                        /_ |___________||  |
                                       /_ \| ----------||  |
                                       /_ /|           ||  |
                                      /  | ||          ||  |   
                                      \//  ||          ||  |
                                      /  | ||    T     ||  |
                                     /   | ||    |     |
     "LSTM, Show me the next word"  / 

L(ets) S(how) T(he) .... - Next Word Predictor

This is a tool to train and evaluate an LSTM Model for next word prediction. For help, just run

make help

How to run

Raw

Requirements:

  • MacOS
  • Python 3.12

If you want to train the model, just run:

sh run.sh

Docker

For reproducibility, this project is been offered in a containerized version as well. In order to run it in a container, all you need to do is run:

docker-compose up 

About

Benchmarking next word predictors from the Pre-Transformer Age 👴🏻

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published