Skip to content

Latest commit

 

History

History
16 lines (11 loc) · 599 Bytes

README.md

File metadata and controls

16 lines (11 loc) · 599 Bytes

TF-GPT

A TensorFlow implementation of GPT. It implements a stack of decoder blocks for autoregressive text generation, allowing you to train your own foundation models and (smaller) LLMs.

Usage

To run, simply use the command line:

python main.py

If you want to train on a custom text file (that fits in RAM) then run the following command, substituting myDataset.txt for your text file. If you don't specify your file, it'll train on the HuggingFace Wikipedia Dataset.

python main.py --data="myDataset.txt"