https://github.com/tauses/engagemachinelearning

Submitted as part of my Machine Learning exam.
https://github.com/tauses/engagemachinelearning

finetuned-model llama2 llm machinelearning

Last synced: about 1 month ago
JSON representation

Submitted as part of my Machine Learning exam.

• Host: GitHub
• URL: https://github.com/tauses/engagemachinelearning
• Owner: Tauses
• Created: 2025-05-11T20:53:02.000Z (about 1 year ago)
• Default Branch: master
• Last Pushed: 2025-06-28T14:18:37.000Z (12 months ago)
• Last Synced: 2025-07-14T03:30:13.865Z (11 months ago)
• Topics: finetuned-model, llama2, llm, machinelearning
• Language: Python
• Homepage:
• Size: 32.2 MB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 1
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
Gaming Chatbot & Player Profiler


A compact Python toolkit for analysing gamers behavioural data and chatting with them through locally‑run Llama 2 models. The project demonstrates end‑to‑end ML: clustering, classification, sentiment analysis and an interactive recommendation bot. Using the K-means and RandomForest pipeline for optimal prediction capabilities.

Key Features



  
Data profiling — K-Means clustering (TrainProfiler.py) categorizes players into six intuitive segments.


  
Engagement prediction — Random-Forest classifier (RFTrainBot.py) forecasts each player's engagement level.


  


Three chatbots

    


      


Heavy — fully automated ML + Llama; uses regex to extract user features (Chatbot_Heavy_Model.py).

      


Light — semi-automated, keyword-triggered ML for user-feature extraction (Chatbot_Lighter_Model.py).

      


Stupid — rule-based fallback with canned responses (Chatbot_Stupid.py).

    

  


  
Sentiment and keyword tracking with NLTK.


  
Local Llama 2 inference via llama-cpp-python (no external API keys required).



REQUIREMENTS

In order to run the light and heavy model, you must first install Visual Studio BuildTools for desktop. 

https://visualstudio.microsoft.com/visual-cpp-build-tools/

Under "Workloads" after installing the BuildTools Choose "Desktop Development with C++" and wait for it to install the packages.


Now you should be able to run it.

Customising the Models

Wish to try my models?


You can try editing the n_clusters in TrainProfiler.py.

There's quite a few other settings to mess around with aswell. Watch out for the context window, the max is 4096, setting it higher results in crashing the program.