Skip to content

akseljoonas/Philosophy-GPT

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

89 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Philosophy GPT

This project, inspired by the works of Friedrich Nietzsche and GPT-2, implements a nano-scale generative decoder-only transformer in JAX which was trained on 2 a100 on a high-performance computing cluster @ RUG.

✨ Highlights

Dataset

Our dataset comprises the complete works of Nietzsche, consisting of 3,411,407 characters. After preprocessing, we refined the dataset to 3,396,780 characters, ensuring a rich textual corpus for our model and no unnecessary characters.

Training and Optimization

We leveraged the computational efficiency of JAX and Flax, with Just-In-Time (JIT) compilation and parallel computing with pmap on the RUG's supercomputer, Habrok.

📈 Results

Quantitative Analysis

Our transformer model showed a stable decrease in both training and evaluation losses, outperforming the benchmark model significantly.

Qualitative Analysis

The benchmark model produced text like:

"Misterel of is r Thin lfe n aneacoucereagencous t Mer ete.. aler lllorivede out effore id ivity the"

Whereas our transformer model generated:

"Education is a fundamental necessity, always. It is similar with good music and art—they are often misunderstood or misleading, much like fish in a vast sea. Among creatures, a lion stands apart, embodying a solitude that defines its essence."

The improvement is evident in the coherent and essay-like structure of the output, demonstrating the model's ability to capture Nietzsche's stylistic essence.

🧠 The Ultimate Question

What is the meaning of life? Our model’s whimsical attempt to answer this eternal question:

"The meaning of life is with rising delays, like the sting of their fellow artists linger."

📄 Report

For a detailed account of our methodology, challenges, and insights, refer to our comprehensive project report.

🧑‍🤝‍🧑 Team

  • Aksel Joonas Reedi
  • Mihkel Mariusz Jezierski
  • Elisa Klunder
  • Mika Umaña

This project was developed as part of our bachelor's studies

About

124M Parameter Decoder-only GPT

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors