calebtalley2024 / gpt_toy_tutor

Installation

• Install libraries by using pip install requirements.txt
• put the .env file in the command_line_toy folder

Workflow

• Edit in Jupiter Notebooks (.ipynb), Export the file as a Python (.py) file

Files: py/ipynb

• learning: code used for teaching student and/or GPT
• students: handles MongoDB database and local Student Objects
• memory: handles MongoDB database and local MemPropmt Objects
• database_connect.py: connects project to MongoDB database
• data_preprocessing.ipynb: preprocesses subtopics into format identical to subtopics.csv

Data

• GPT_tutor_topics(subtopics_included): math data
• subtopics.csv: math data preprocessed

User

• GPT gives students questions

Trainer

• Train GPT Tutors External Databases

What's Next?

• Change name, GPT Tutor is already taken

Reinforcement Learning

• Neural UCB

Robustness Testing

• Different topics
• Program synthesis section
• MemPrompt feedback effectiveness

Sources

• sauxpa's implementation of Neural UCB
• Rest of sources are in Paper in APA 7 format