Repositories
1. Battle Game
Repo Description: A simple 2D battle game developed in Python using the Pygame library. Players engage in real-time combat against each other, utilizing different weapons and abilities. The game supports 1vs1 PVP mode with player movement, attack controls, and health/energy management.
Key Features:
- Real-time 1vs1 battles
- Health and energy management
- Interactive environment and effects
- Easy-to-use controls via keyboard
Setup & Run:
- Clone and install dependencies via pip.
- Run the game script or compile it into an executable with PyInstaller.
Link: Battle Game Repo
2. Music Genre Classification
Repo Description: This project focuses on music genre classification using machine learning techniques. It involves the training and deployment of a pre-trained audio model fine-tuned on the GTZAN dataset, utilizing the HuggingFace platform for model deployment.
Key Features:
- Music genre classification with pre-trained models
- Utilizes HuggingFace for model training and inference
- Data preprocessing, feature extraction, and model evaluation pipelines
Setup & Run:
- Set up both Conda and virtualenv environments.
- Train models with the provided Python scripts and run them using custom pipelines.
Link: Music Genre Classification Repo
3. MLOps Challenge
Repo Description: This project implements an end-to-end ML solution to classify keystrokes, serving the trained models via a REST API. It focuses on building a reliable data pipeline, model training, and deployment in a production environment.
Key Features:
- Data processing and feature extraction pipeline
- Model training and evaluation for keystroke classification
- REST API for inference serving
- Deployed API on AWS EC2 instance for public access
Setup & Run:
- Set up both development and production environments.
- Train models and serve them via a REST API using FastAPI and Uvicorn.
Link: MLOps Challenge Repo
4. LLM QA Retrieval
Repo Description: This project involves building a QA retrieval system using Large Language Models (LLMs). It includes zero-shot and few-shot learning techniques to answer questions from fetched news data, employing ChromaDB for semantic search and text generation using LLMs.
Key Features:
- Fetches news articles and processes them for semantic search
- Implements a QA pipeline with text generation using LLMs
- Configurable pipeline for different query types and news sources
- Integration with Weights & Biases for model tracking and reporting
Setup & Run:
- Fetch and chunk news articles, then run the QA pipeline for answering queries.
- Use pre-configured models for zero-shot or fine-tuned QA.
Link: LLM QA Retrieval Repo
5. LLM Text Generation
Repo Description: A project dedicated to exploring the text generation capabilities of Large Language Models (LLMs). It focuses on model fine-tuning, text generation tasks, and deploying these models for real-world applications.
Key Features:
- Fine-tuning of large language models for text generation
- Data preprocessing and pipeline management
- Visualizations and evaluation of generated texts
Link: LLM Text Generation Repo
6. Activity Recognition
Repo Description: Machine Learning Challenge for Human Activity Recognition. This project is organized to facilitate data processing, model training, and evaluation for recognizing human activities based on sensor data.
Key Features:
- Human Activity Recognition Data Analysis
- Data preprocessing and Pipeline Management
- Exploratory Data Analysis and Machine Learning Development
- Visualizations and Final Presentation
Setup & Run:
- Set up both Conda and virtualenv environments.
- Train models and perform EDA with the provided Python scripts and run them using custom pipelines.
Link: HAR Prediction Repo