AdityaNG/kan-gpt
A PyTorch implementation of GPT models using Kolmogorov-Arnold Networks (KANs) as an alternative to standard MLPs for language modeling.
Not currently ranked — collecting fresh signals.
star history
This repository implements Generative Pre-trained Transformers using Kolmogorov-Arnold Networks (KANs) instead of traditional multi-layer perceptrons. KANs replace linear weight parameters with spline-based functions, offering an alternative approach to transformer architectures. The project provides a working GPT implementation with text generation capabilities using the GPT2 tokenizer.
Frequently asked
- What is AdityaNG/kan-gpt?
- A PyTorch implementation of GPT models using Kolmogorov-Arnold Networks (KANs) as an alternative to standard MLPs for language modeling.
- Is kan-gpt open source?
- Yes — AdityaNG/kan-gpt is open source, released under the MIT license.
- What language is kan-gpt written in?
- AdityaNG/kan-gpt is primarily written in Python.
- How popular is kan-gpt?
- AdityaNG/kan-gpt has 724 stars on GitHub.
- Where can I find kan-gpt?
- AdityaNG/kan-gpt is on GitHub at https://github.com/AdityaNG/kan-gpt.