LMPT is a library with functions to construct powerful, multi-step language model prompts. For example, using the write_book() function, you can ask in a single function call ask an arbitrary language model to write an entire book for you. This, under the hood, uses a structured series of prompts designed for this purpose that you can find in the write_book.py file.
- Install the LMTK library using
pip install -U lmtk
Please see the LMTK documentation if you have any issues with that install
- Run the relevant script from this repo using ones of the examples below.
If you have any issues with installing or using this code, please reach out to me (rafaelcosman@alumni.stanford.edu).
Just run:
python textbook_writing.py
You should then see on your prompt:
What would you like to write a textbook on? You can just give me a single topic, like 'Bioethics'.
Let's say you input Bioethics.
You should then see on your prompt:
- A list of the chapters for the Bioethics textbook
- An outline for each chapter
- Each chapter written out, section by section
This third step can take 30-60 minutes, please be patient. It's important that your internet connection is not interrupted during this process! Note also that textbook generation requires a LOT of queries to the OpenAI API, so if you have a limited budget this may eat into it very quickly.
Similar to writing textbooks, just run:
python journal_writing.py
You should then see in your prompt:
What would you like me to write a journal about? You can give me a signle topic like 'Geopolitics'.
Let's say you input Geopolitics.
You should then see on your prompt:
- A list of the journal article topics
- Each journal article written out in turn
edit.py will edit text to be clear and concise. It is used by other functions such as journal/article generation. To use it directly, just run python edit.py
