Offline Machine Learning Model Development

Our project focuses on developing a machine learning model utilizing a confidential dataset. We aim to interact with this model to retrieve specific statuses from our database. Given our offline environment, we will provide the hired individual with access to an isolated virtual machine for development. Planned Components: - Data Storage: We intend to use ChromaDB for storing datasets and documents, as it supports internal local hosting. - Embedding Model: We plan to implement the BAAI/bge-large-en-v1.5 model, which transforms text into 1024-dimensional vectors, facilitating tasks like retrieval and semantic search. ([login to view URL])) - Interactive Model: We are considering the Llama 3.2 1B model for future interactive capabilities. This lightweight, multilingual model is optimized for tasks such as personal information management and knowledge retrieval, making it suitable for deployment on edge devices. ([login to view URL])) - Data Preprocessing: Include tools and methods for cleaning, normalizing, and preparing datasets for model training. - Model Training Pipeline: Develop a robust training pipeline using frameworks like TensorFlow or PyTorch. - Evaluation Metrics: Define key performance indicators such as accuracy, precision, recall, and F1-score for model evaluation. - Data Augmentation: Implement data augmentation techniques to enhance the training data and improve model generalization. - Error Handling: Develop robust mechanisms for logging and handling errors or exceptions that may arise during model deployment. - Version Control: Use tools like Git for version control to manage changes in the model and datasets effectively. - Security Protocols: Define security measures to ensure the confidentiality and integrity of the dataset and model. - Documentation: Provide detailed documentation for all components and processes involved, ensuring easy maintenance and scalability. - Deployment Strategy: Outline strategies for deploying the model in an offline environment, ensuring smooth integration with existing systems. - Model Optimization: Implement techniques such as quantization and pruning to optimize the model for faster inference and lower resource consumption. The model should be developed using PyTorch to leverage its dynamic computation graph and efficient optimization tools. The primary goal for the model's output is to retrieve specific statuses from our database. The primary use case for the model's output is status monitoring from our database. The expected timeline for project completion is 1-3 months.