amitgm/llm

Code for various Large Language Models

LLm for various Large Language Models

Large Language Models (LLMs) leverage billions of parameters to train on vast amounts of text data. Industry leaders like Google, Meta (formerly Facebook), and OpenAI have developed some of the most prominent LLMs. Here are several applications where Large Language Models are utilized, often in conjunction with libraries from Hugging Face:

1. Text Generation: LLMs excel at generating coherent and contextually relevant text. Applications range from chatbots and virtual assistants to content creation tools. Hugging Face provides libraries like Transformers for fine-tuning and deploying LLMs for text generation tasks.

2. Language Translation: LLMs are instrumental in machine translation systems, enabling accurate and fluent translations across multiple languages. Hugging Face's translation models, built on top of Transformers, facilitate language translation tasks with ease.

3. Text Summarization: LLMs can reduce lengthy documents or articles into concise summaries, aiding in information retrieval and content summary. Hugging Face's summarization models offer efficient solutions for generating abstractive summaries.

BELOW ARE SOME OF THE APPLICATIONS CREATED USING LARGE LANGUAGE MODELS

CHAT WITH PDF

Engage with PDFs effortlessly through Chat with PDF. This tool allows you to upload single or multiple PDF files and interact with them seamlessly. Upon upload, you can pose a variety of questions related to the content, and the system will generate responses based on the set temperature, which influences creativity. Additional features include a concise introduction to the document and a memory function to review previous questions and answers. Powered by Google PaLM LLM, Chat with PDFs offers a user-friendly experience for exploring PDF content.

A video demo of the project has been shown on Chat with PDF.mp4

IMAGE TO STORY APP

Transform any image into a humorous narrative with the IMAGE TO STORY application. This innovative tool utilizes cutting-edge technologies like Google PaLM LLM and various Hugging Face models. By employing Salesforce/blip-image-captioning-base, it seamlessly converts the image into text, while the kan-bayashi_ljspeech_vits model brings the story to life through voice narration.

FINE TUNING LLMS

Fine-tuning LLMs on various datasets using the state-of-the-art LLMs such as Falcon, Zephyr, Mistral and many others. The Google Colab notebooks give us ideas on different use cases of fine-tuned LLms