Kai Liang Teoh@kailiang

Users have the ability to upload a range of file formats, including PDF, DOCX, MD, TXT, JPG, and PNG, to a specific application. The program analyzes these files, extracting text through techniques such as parsing and OCR for images. The AI71 model examines the information, allowing for functions like natural language processing, and answering user questions. The app needs a user-friendly interface for uploading files, backed by the ability to accurately handle various formats. In general, incorporating the AI71 model enables effective communication and use of data, giving valuable feedback and analysis depending on the content that is uploaded.

This project implements a DeepSeek LLM-based sentiment analysis application using Streamlit for an interactive web interface. The model, deepseek-ai/deepseek-llm-7b-base, is loaded with transformers and deployed via Google Colab. The app accepts text input from users and analyzes sentiment by generating a response from the model, returning only one of three possible outputs: POSITIVE, NEGATIVE, or NEUTRAL. The pipeline is optimized for automatic device selection (device_map="auto") and utilizes offloading to manage memory efficiently. Streamlit serves the app on port 8501, and a Colab proxy exposes it via a public URL.