Skip to content

ATLAS-Illinois/AI_town_simulation_with_LAMA2

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

16 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Abstract

This project is an alternative to Generative Agents, substituting the GPT model with the Mixtral model from the University of Illinois at Urbana-Champaign. The aim is to conduct simulations with specific persona settings and evaluate the Mixtral model's performance in interactions resembling human-like NPC communication.

Files Modified

  • reverie/backend_server/persona/prompt_template/gpt_structure.py
  • reverie/backend_server/persona/prompt_template/run_gpt_prompt.py
  • reverie/backend_server/reverie.py
  • All files under reverie/backend_server/persona/prompt_template
  • reverie.py

For the details of the changes, see the comments within the files.

New File Added:

  • run_simulations easier method to run the simulation with large number of steps
  • environment/frontend_server/storage/base_the_vile_n5 5 person version of simulation with some evial personality
  • reverie/backend_server/API.py (Used for accessing the Llama model From university)

Note: If you are the student in the university, asked the acess account and password First and you need enter the account and password into this API file before any simulation beginning

Simulation Steps

Step 0

Download all required envirnoment which stored in 'environment/frontend_server/requirements.txt'

Step 1

Open a terminal and navigate to the frontend_server directory.

cd environment/frontend_server

Then run the server.

python manage.py runserver

Step 2

In another terminal, navigate to the backend_server directory.

cd reverie/backend_server

Then run the backend.

python reverie.py

Step 3

Follow the on-screen instructions for setup.

  1. You'll be prompted to Enter the name of the forked simulation:
  • Go to environment/frontend_server/storage to see package names:
    • base_the_ville_isabella_maria_klaus: a smaller simulation with 3 people
    • base_the_ville_n25: a larger simulation with 25 people
  • Chooce the enter the package name which you want to simulation
  • For testing, considering less computation cost, enter:
base_the_ville_isabella_maria_klaus
  1. Then it would shows Enter the name of the new simulation: and you can type any name like:
test-simulation
  1. Finally, in Enter option:, type run followed by a number to specify the simulation duration like:
 run 10

Final Steps:

You would able to see the Enter option: again after one simulation complete:

  • To save the results, type fin
  • To exit without saving, type exit
  • To run another simulation, type "run" followed by a number.

About

To solve the problem of highly billing of GPT4, this repository is trying to change the GPT request to LAMA2 request

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors