Building a Llama 3.2-11B-Based RAG System with Vision Model Integration

October 21, 2024

Introduction

Llama 3.2-11B is a state-of-the-art multimodal large language model developed by Meta, featuring 11 billion parameters. This model is part of the Llama 3.2 collection, which includes both vision and text capabilities, marking a significant advancement in the integration of visual and linguistic understanding. Unlike its predecessors, Llama 3.2-11B is designed to handle complex tasks that involve both text and images, such as image captioning, visual question answering, and document understanding.

One of the standout features of Llama 3.2-11B is its ability to process high-resolution images alongside text, enabling it to perform tasks that require reasoning over visual data. The model employs a unique architecture that integrates an image encoder into the language model through a series of cross-attention layers. This allows it to bridge the gap between visual inputs and language outputs effectively. Also, it supports a context length of up to 128,000 tokens, significantly enhancing its capacity for handling extensive data inputs.

The applications of Llama 3.2-11B are diverse. For example, in education, it can be used for interactive learning tools that require image-based content generation or analysis. In business contexts, it can assist in extracting information from graphs or charts to provide insights based on visual data. Alternatively, it can enhance customer service through AI-driven solutions that require both visual comprehension and conversational abilities.

Our Aim in This Tutorial

In this guide, we'll walk you through the process of building a Retrieval-Augmented Generation (RAG) system using the Llama 3.2-11B model, enhanced with vision capabilities. The goal is to process images, extract text from them, and store this data in a high-performance vector database like Qdrant. This setup enables efficient retrieval of information when users ask questions.

The process starts by taking an image URL and using the Llama 3.2-11B model to extract text from the image. The extracted text is then stored in Qdrant, allowing for quick lookups and retrievals based on queries. Building on this, we'll create a Question and Answer (QnA) system that retrieves relevant information from the stored data and generates responses using the Llama model.

This project demonstrates how Llama 3.2-11B can seamlessly integrate vision tasks with text-based retrieval, creating a powerful system capable of handling both image processing and dynamic QnA functionality.

Let’s Code

The Installations

This command installs libraries for NLP, embeddings, vector search, and building machine learning interfaces.


!pip install -q sentence-transformers transformers qdrant-client groq gradio

Imports

Some required Imports:


from PIL import Image
import io
import torch
from transformers import LlamaTokenizer, LlamaForCausalLM
from sentence_transformers import SentenceTransformer
from qdrant_client import QdrantClient
from qdrant_client.http.models import VectorParams, Distance
import os
import requests
import torch
from PIL import Image
from transformers import MllamaForConditionalGeneration, AutoProcessor
from groq import Groq
import gradio as gr

Hugging Face Login


from huggingface_hub import notebook_login
notebook_login()

Setup

In this code block, we’re setting up a sophisticated natural language processing pipeline using several powerful tools:

LLaMA 3.2 11B Model: We load the "Llama-3.2-11B-Vision-Instruct" model, a large language model designed for both text and vision tasks. It’s configured with bfloat16 precision to balance memory efficiency and computational power. The device_map="auto" setting automatically assigns the model to the most appropriate hardware (like GPU or CPU), streamlining deployment.
Processor: The model's processor is loaded using the AutoProcessor class. This ensures inputs are formatted correctly for the model, whether it’s images, text, or a combination.
Groq Client: Here, we’re initializing the Groq client, which connects to Groq’s hardware accelerator. This is useful for speeding up inference, leveraging Groq’s AI hardware, and requires an API key to connect.
Sentence Embedder: We’re using the SentenceTransformer library with the "all-MiniLM-L6-v2" model to generate sentence embeddings. This model is small yet effective, making it ideal for creating vector representations of text for tasks like semantic search or clustering.
Qdrant Vector Database: Finally, we connect to a Qdrant instance, a vector database optimized for searching and storing embeddings. It’s running locally on port 6333 and will store our text embeddings, allowing us to efficiently search or compare them.

This setup enables robust natural language and vision tasks, high-performance inference, and fast vector-based searches—all in one workflow.


model_id = "meta-llama/Llama-3.2-11B-Vision-Instruct"

model = MllamaForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="auto",
)
processor = AutoProcessor.from_pretrained(model_id)

client = Groq(
    api_key="",
)

embedder = SentenceTransformer('all-MiniLM-L6-v2')
qdrant = QdrantClient("localhost", port=6333)

Qdrant Vector DB

The function create_qdrant_collection() sets up a new collection in the Qdrant vector database, specifically named "pdf_collection". This collection is configured to store vector embeddings with a size of 384 dimensions. The collection uses cosine similarity as the distance metric, which is ideal for comparing the similarity between vectors, commonly used in tasks like semantic search, recommendation systems, or clustering. The qdrant.create_collection method handles the creation and configuration of the collection.


def create_qdrant_collection():
    qdrant.create_collection(
        collection_name="pdf_collection",
        vectors_config=VectorParams(size=384, distance=Distance.COSINE),
    )

Text Extraction from Image (OCR)

The function convert_url_to_text(url) takes an image from a given URL, processes it, and extracts text from the image using a model.

Here's a breakdown:

Image Retrieval: The function fetches an image from a URL and opens it using PIL's Image.open with a stream request.
Message Setup: It prepares a message instructing the model to "Extract text from the image" and formats this input using a processor.
Processing Input: The image and text prompt are processed and tokenized to be compatible with the model.
Text Generation: The model generates a response based on the processed input (the image and the prompt), and it is configured to allow up to 28,000 tokens.
Decoding: The output is decoded into readable text and returned as the extracted text from the image.

This function is designed to convert an image from a URL into text using an AI model that processes both images and natural language instructions.


def convert_url_to_text(url):
    image = Image.open(requests.get(url, stream=True).raw)
    #image = Image.open(local_path)

    messages = [
        {"role": "user", "content": [
            {"type": "image"},
            {"type": "text", "text": "Extrcat text from the image, Respond with only extracted text"}
        ]}
    ]
    input_text = processor.apply_chat_template(messages, add_generation_prompt=True)
    inputs = processor(image, input_text, return_tensors="pt").to(model.device)

    output = model.generate(**inputs, max_new_tokens=28000)
    return processor.decode(output[0])

Index Text

The function index_texts(texts: str, max_token=100) is designed to encode and store text data into a Qdrant vector database for future search or retrieval.

Here's a summary of how it works:

Loop through Texts: The function iterates over each text in the provided texts list (or other iterable).
Generate Embeddings: For each text, it uses the embedder.encode() function to generate a vector (embedding) representing the text in a high-dimensional space. The vector is then converted to a list format.
Prepare Payload: It creates a payload (metadata) for each text, containing the page_num (which is the index of the text) and the actual text.
Upsert into Qdrant: The function inserts or updates (upsert) the text embedding into the "pdf_collection" in the Qdrant vector database. Each point is stored with an ID, its embedding, and the associated payload metadata (such as page number and text).

This function essentially indexes a series of text documents by generating embeddings and storing them in Qdrant for later vector-based search or retrieval.


def index_texts(texts: str, max_token = 100):
    for i, text in enumerate(texts):
        embedding = embedder.encode(text).tolist()
        payload = {"page_num": i, "text": text}
        qdrant.upsert(
            collection_name="pdf_collection",
            points=[{
                "id": i,
                "vector": embedding,
                "payload": payload
            }]
        )

Relevant Document Extraction

The function retrieve_relevant_text(query) is designed to search for and retrieve relevant text snippets from a Qdrant vector database based on a provided query. Here’s a brief overview of how it works:

Query Embedding: It first generates an embedding for the input query using the embedder.encode() method, converting the resulting vector to a list format.
Search in Qdrant: The function then performs a search in the "pdf_collection" of the Qdrant database using the query embedding. It retrieves a maximum of 3 relevant results based on similarity.
Extract and Return Text: Finally, it compiles a list of the text payloads from the results and returns them, providing the user with the most relevant text snippets related to their query.

This function facilitates efficient retrieval of information from indexed texts by leveraging vector similarity search.


# Retrieve relevant text from Qdrant
def retrieve_relevant_text(query):
    query_embedding = embedder.encode(query).tolist()
    results = qdrant.search(
        collection_name="pdf_collection",
        query_vector=query_embedding,
        limit=3
    )
    return [result.payload['text'] for result in results]

‍

Generating Answers

The function generate_answer(context, query) is designed to generate a response to a user's query based on a given context using the Llama 3.2 model. Here’s a concise overview of its operation:

Prepare Messages: The function constructs a message format suitable for the chat model. It includes a system message that establishes the role of the agent as an answer generator, incorporating the provided context. It also includes a user message that specifies the user's query and instructs the model to answer it using the context.
Chat Completion: It sends these messages to the Llama model via the Groq Client to generate a chat completion, which is the model's response to the query.
Return Answer: Finally, the function retrieves and returns the generated answer from the model’s response, specifically the content of the first choice.

This function allows for dynamic question answering based on the context provided, leveraging the capabilities of the Llama 3.2 model.


# Generate an answer using LLaMA 3.2 11B
def generate_answer(context, query):
    chat_completion = client.chat.completions.create(
        messages=[
            {
                "role": "system",
                "content": "You are an answer generation angent, with context {context}",
                "role": "user",
                "content": f"your query was {query}, answer this query from the given context {context}",
            }
        ],
        model="llama-3.2-1b-preview",
    )

    return chat_completion.choices[0].message.content

Main Entry Point

This block of code serves as the main entry point for executing a series of tasks related to processing an image, indexing text, and generating a response to a query. Here’s a breakdown of its functionality:

URL and Query Setup: It defines a URL pointing to an image and a query asking for information about that image.
Text Extraction: The function convert_url_to_text(url) is called to extract text from the specified image URL. The extracted text is stored in the data variable.
Create Qdrant Collection: The create_qdrant_collection() function is called to create a new collection in Qdrant where the text data will be indexed.
Index Texts: The index_texts(context) function is used to index the extracted text into the Qdrant collection, allowing for future retrieval based on embeddings. (Note: There seems to be a missing definition of context before this call; it should likely reference the text extracted from the image.)
Retrieve Relevant Text: The retrieve_relevant_text(query) function retrieves text relevant to the provided query from the Qdrant collection. The results are stored in the context variable.
Generate Answer: Finally, the generate_answer(context, query) function is called to generate a response based on the retrieved context and the user's query.

This code effectively combines image processing, text indexing, and natural language understanding to answer user queries about the content of an image.


if __name__=="__main__":
    url = "https://resumegenius.com/wp-content/uploads/2019/08/Computer-Science-Cover-Letter-Example-Template.png"
    query = "What is this information about"
    data = convert_url_to_text(url)
    create_qdrant_collection()
    payload = index_texts(context)
    context = retrieve_relevant_text(query)
    response = generate_answer(context, query)

Gradio Interface

This code sets up a Gradio web application that processes an image URL and a user query.

Process Function: The process(url, query) function:some text
- Extracts text from the image at the specified URL.
- Creates a Qdrant collection.
- Indexes the extracted text.
- Retrieves relevant text based on the query.
- Generates and returns a response.
Gradio Interface: It defines a user interface with two text inputs for the URL and query, and outputs the generated response.
Launch: Finally, it launches the app, allowing users to interact with the model through a web interface.

This setup enables users to get relevant answers based on the content of an image simply by providing a URL and a question.


# Main function to handle the Gradio inputs
def process(url, query):
    context = convert_url_to_text(url)
    create_qdrant_collection()
    payload = index_texts(data)
    context = retrieve_relevant_text(query)
    response = generate_answer(context, query)
    return response

# Define the Gradio interface
interface = gr.Interface(
    fn=process,
    inputs=[gr.Textbox(label="URL"), gr.Textbox(label="Query")],
    outputs="text",
    title="URL to Text and Query Processing",
    description="Provide a URL and a query to get a relevant response."
)

# Launch the Gradio app
interface.launch(share = True)

Outputs

‍

‍

‍

‍

‍

Conclusion

Today, we delved into the fascinating world of object detection and demonstrated how powerful it can be in various applications.

To get started with Flux.1-dev, sign up to E2E Cloud today, and launch a cloud GPU node, or head to TIR. E2E Cloud offers the most price-performant cloud GPUs in the Indian market, and enables developers to use advanced GPUs like H200, H100, A100 for application development.

Sign up for Free Trial

Latest Blogs

October 21, 2024

A Complete Guide To Customer Acquisition For Startups

Any business is enlivened by its customers. Therefore, a strategy to constantly bring in new clients is an ongoing requirement. In this regard, having a proper customer acquisition strategy can be of great importance.

So, if you are just starting your business, or planning to expand it, read on to learn more about this concept.

The problem with customer acquisition

As an organization, when working in a diverse and competitive market like India, you need to have a well-defined customer acquisition strategy to attain success. However, this is where most startups struggle. Now, you may have a great product or service, but if you are not in the right place targeting the right demographic, you are not likely to get the results you want.

To resolve this, typically, companies invest, but if that is not channelized properly, it will be futile.

So, the best way out of this dilemma is to have a clear customer acquisition strategy in place.

How can you create the ideal customer acquisition strategy for your business?

Define what your goals are

You need to define your goals so that you can meet the revenue expectations you have for the current fiscal year. You need to find a value for the metrics –

MRR – Monthly recurring revenue, which tells you all the income that can be generated from all your income channels.
CLV – Customer lifetime value tells you how much a customer is willing to spend on your business during your mutual relationship duration.
CAC – Customer acquisition costs, which tells how much your organization needs to spend to acquire customers constantly.
Churn rate – It tells you the rate at which customers stop doing business.

All these metrics tell you how well you will be able to grow your business and revenue.

Identify your ideal customers

You need to understand who your current customers are and who your target customers are. Once you are aware of your customer base, you can focus your energies in that direction and get the maximum sale of your products or services. You can also understand what your customers require through various analytics and markers and address them to leverage your products/services towards them.

Choose your channels for customer acquisition

How will you acquire customers who will eventually tell at what scale and at what rate you need to expand your business? You could market and sell your products on social media channels like Instagram, Facebook and YouTube, or invest in paid marketing like Google Ads. You need to develop a unique strategy for each of these channels.

Communicate with your customers

If you know exactly what your customers have in mind, then you will be able to develop your customer strategy with a clear perspective in mind. You can do it through surveys or customer opinion forms, email contact forms, blog posts and social media posts. After that, you just need to measure the analytics, clearly understand the insights, and improve your strategy accordingly.

Combining these strategies with your long-term business plan will bring results. However, there will be challenges on the way, where you need to adapt as per the requirements to make the most of it. At the same time, introducing new technologies like AI and ML can also solve such issues easily. To learn more about the use of AI and ML and how they are transforming businesses, keep referring to the blog section of E2E Networks.

‍

Reference Links

https://www.helpscout.com/customer-acquisition/

https://www.cloudways.com/blog/customer-acquisition-strategy-for-startups/

https://blog.hubspot.com/service/customer-acquisition

This is a decorative image for: Constructing 3D objects through Deep Learning

October 18, 2022

Image-based 3D Object Reconstruction State-of-the-Art and trends in the Deep Learning Era

3D reconstruction is one of the most complex issues of deep learning systems. There have been multiple types of research in this field, and almost everything has been tried on it — computer vision, computer graphics and machine learning, but to no avail. However, that has resulted in CNN or convolutional neural networks foraying into this field, which has yielded some success.

The Main Objective of the 3D Object Reconstruction

Developing this deep learning technology aims to infer the shape of 3D objects from 2D images. So, to conduct the experiment, you need the following:

Highly calibrated cameras that take a photograph of the image from various angles.
Large training datasets can predict the geometry of the object whose 3D image reconstruction needs to be done. These datasets can be collected from a database of images, or they can be collected and sampled from a video.

By using the apparatus and datasets, you will be able to proceed with the 3D reconstruction from 2D datasets.

State-of-the-art Technology Used by the Datasets for the Reconstruction of 3D Objects

The technology used for this purpose needs to stick to the following parameters:

Input

Training with the help of one or multiple RGB images, where the segmentation of the 3D ground truth needs to be done. It could be one image, multiple images or even a video stream.

The testing will also be done on the same parameters, which will also help to create a uniform, cluttered background, or both.

Output

The volumetric output will be done in both high and low resolution, and the surface output will be generated through parameterisation, template deformation and point cloud. Moreover, the direct and intermediate outputs will be calculated this way.

Network architecture used

The architecture used in training is 3D-VAE-GAN, which has an encoder and a decoder, with TL-Net and conditional GAN. At the same time, the testing architecture is 3D-VAE, which has an encoder and a decoder.

Training used

The degree of supervision used in 2D vs 3D supervision, weak supervision along with loss functions have to be included in this system. The training procedure is adversarial training with joint 2D and 3D embeddings. Also, the network architecture is extremely important for the speed and processing quality of the output images.

Practical applications and use cases

Volumetric representations and surface representations can do the reconstruction. Powerful computer systems need to be used for reconstruction.

Given below are some of the places where 3D Object Reconstruction Deep Learning Systems are used:

3D reconstruction technology can be used in the Police Department for drawing the faces of criminals whose images have been procured from a crime site where their faces are not completely revealed.
It can be used for re-modelling ruins at ancient architectural sites. The rubble or the debris stubs of structures can be used to recreate the entire building structure and get an idea of how it looked in the past.
They can be used in plastic surgery where the organs, face, limbs or any other portion of the body has been damaged and needs to be rebuilt.
It can be used in airport security, where concealed shapes can be used for guessing whether a person is armed or is carrying explosives or not.
It can also help in completing DNA sequences.

So, if you are planning to implement this technology, then you can rent the required infrastructure from E2E Networks and avoid investing in it. And if you plan to learn more about such topics, then keep a tab on the blog section of the website.

‍

Reference Links

https://tongtianta.site/paper/68922

https://github.com/natowi/3D-Reconstruction-with-Deep-Learning-Methods

This is a decorative image for: Comprehensive Guide to Deep Q-Learning for Data Science Enthusiasts

October 18, 2022

A Comprehensive Guide To Deep Q-Learning For Data Science Enthusiasts

For all data science enthusiasts who would love to dig deep, we have composed a write-up about Q-Learning specifically for you all. Deep Q-Learning and Reinforcement learning (RL) are extremely popular these days. These two data science methodologies use Python libraries like TensorFlow 2 and openAI’s Gym environment.

So, read on to know more.

What is Deep Q-Learning?

Deep Q-Learning utilizes the principles of Q-learning, but instead of using the Q-table, it uses the neural network. The algorithm of deep Q-Learning uses the states as input and the optimal Q-value of every action possible as the output. The agent gathers and stores all the previous experiences in the memory of the trained tuple in the following order:

State> Next state> Action> Reward

The neural network training stability increases using a random batch of previous data by using the experience replay. Experience replay also means the previous experiences stocking, and the target network uses it for training and calculation of the Q-network and the predicted Q-Value. This neural network uses openAI Gym, which is provided by taxi-v3 environments.

Now, any understanding of Deep Q-Learning is incomplete without talking about Reinforcement Learning.

What is Reinforcement Learning?

Reinforcement is a subsection of ML. This part of ML is related to the action in which an environmental agent participates in a reward-based system and uses Reinforcement Learning to maximize the rewards. Reinforcement Learning is a different technique from unsupervised learning or supervised learning because it does not require a supervised input/output pair. The number of corrections is also less, so it is a highly efficient technique.

Now, the understanding of reinforcement learning is incomplete without knowing about Markov Decision Process (MDP). MDP is involved with each state that has been presented in the results of the environment, derived from the state previously there. The information which composes both states is gathered and transferred to the decision process. The task of the chosen agent is to maximize the awards. The MDP optimizes the actions and helps construct the optimal policy.

For developing the MDP, you need to follow the Q-Learning Algorithm, which is an extremely important part of data science and machine learning.

What is Q-Learning Algorithm?

The process of Q-Learning is important for understanding the data from scratch. It involves defining the parameters, choosing the actions from the current state and also choosing the actions from the previous state and then developing a Q-table for maximizing the results or output rewards.

The 4 steps that are involved in Q-Learning:

Initializing parameters – The RL (reinforcement learning) model learns the set of actions that the agent requires in the state, environment and time.
Identifying current state – The model stores the prior records for optimal action definition for maximizing the results. For acting in the present state, the state needs to be identified and perform an action combination for it.
Choosing the optimal action set and gaining the relevant experience – A Q-table is generated from the data with a set of specific states and actions, and the weight of this data is calculated for updating the Q-Table to the following step.
Updating Q-table rewards and next state determination – After the relevant experience is gained and agents start getting environmental records. The reward amplitude helps to present the subsequent step.

In case the Q-table size is huge, then the generation of the model is a time-consuming process. This situation requires Deep Q-learning.

Hopefully, this write-up has provided an outline of Deep Q-Learning and its related concepts. If you wish to learn more about such topics, then keep a tab on the blog section of the E2E Networks website.

Reference Links

https://analyticsindiamag.com/comprehensive-guide-to-deep-q-learning-for-data-science-enthusiasts/

https://medium.com/@jereminuerofficial/a-comprehensive-guide-to-deep-q-learning-8aeed632f52f

October 13, 2022

GAUDI: A Neural Architect for Immersive 3D Scene Generation

The evolution of artificial intelligence in the past decade has been staggering, and now the focus is shifting towards AI and ML systems to understand and generate 3D spaces. As a result, there has been extensive research on manipulating 3D generative models. In this regard, Apple’s AI and ML scientists have developed GAUDI, a method specifically for this job.

An introduction to GAUDI

The GAUDI 3D immersive technique founders named it after the famous architect Antoni Gaudi. This AI model takes the help of a camera pose decoder, which enables it to guess the possible camera angles of a scene. Hence, the decoder then makes it possible to predict the 3D canvas from almost every angle.

What does GAUDI do?

GAUDI can perform multiple functions –

The extensions of these generative models have a tremendous effect on ML and computer vision. Pragmatically, such models are highly useful. They are applied in model-based reinforcement learning and planning world models, SLAM is s, or 3D content creation.
Generative modelling for 3D objects has been used for generating scenes using graf, pigan, and gsn, which incorporate a GAN (Generative Adversarial Network). The generator codes radiance fields exclusively. Using the 3D space in the scene along with the camera pose generates the 3D image from that point. This point has a density scalar and RGB value for that specific point in 3D space. This can be done from a 2D camera view. It does this by imposing 3D datasets on those 2D shots. It isolates various objects and scenes and combines them to render a new scene altogether.
GAUDI also removes GANs pathologies like mode collapse and improved GAN.
GAUDI also uses this to train data on a canonical coordinate system. You can compare it by looking at the trajectory of the scenes.

How is GAUDI applied to the content?

The steps of application for GAUDI have been given below:

Each trajectory is created, which consists of a sequence of posed images (These images are from a 3D scene) encoded into a latent representation. This representation which has a radiance field or what we refer to as the 3D scene and the camera path is created in a disentangled way. The results are interpreted as free parameters. The problem is optimized by and formulation of a reconstruction objective.
This simple training process is then scaled to trajectories, thousands of them creating a large number of views. The model samples the radiance fields totally from the previous distribution that the model has learned.
The scenes are thus synthesized by interpolation within the hidden space.
The scaling of 3D scenes generates many scenes that contain thousands of images. During training, there is no issue related to canonical orientation or mode collapse.
A novel de-noising optimization technique is used to find hidden representations that collaborate in modelling the camera poses and the radiance field to create multiple datasets with state-of-the-art performance in generating 3D scenes by building a setup that uses images and text.

To conclude, GAUDI has more capabilities and can also be used for sampling various images and video datasets. Furthermore, this will make a foray into AR (augmented reality) and VR (virtual reality). With GAUDI in hand, the sky is only the limit in the field of media creation. So, if you enjoy reading about the latest development in the field of AI and ML, then keep a tab on the blog section of the E2E Networks website.

‍

Reference Links

https://www.researchgate.net/publication/362323995_GAUDI_A_Neural_Architect_for_Immersive_3D_Scene_Generation

https://www.technology.org/2022/07/31/gaudi-a-neural-architect-for-immersive-3d-scene-generation/

https://www.patentlyapple.com/2022/08/apple-has-unveiled-gaudi-a-neural-architect-for-immersive-3d-scene-generation.html

Build on the most powerful infrastructure cloud

Contact Sales Get Started for Free

A vector illustration of a tech city using latest cloud technologies & infrastructure

Building a Llama 3.2-11B-Based RAG System with Vision Model Integration

Table of Contents

Introduction

Our Aim in This Tutorial

Let’s Code

The Installations

Imports

Hugging Face Login

Setup

Qdrant Vector DB

Text Extraction from Image (OCR)

Index Text

Relevant Document Extraction

Generating Answers

Main Entry Point

Gradio Interface

Outputs

Conclusion