What are Large Language Models(LLMs)?

Introduction

A large language model is a computer program that learns and generates human-like language using a transformer architecture trained on vast training data. Large Language Models (LLMs) are foundational machine learning models that use deep learning algorithms to process and understand natural language. These models are trained on massive amounts of text data to learn patterns and entity relationships in the language. LLMs can perform many types of language tasks, such as translating languages, analyzing sentiments, chatbot conversations, and more. They can understand complex textual data, identify entities and relationships between them, and generate new text that is coherent and grammatically accurate, making them ideal for sentiment analysis.In this article you will get to know about large language model with llm architecture, llm model architecture explained in this article.

Learning Objectives

• Understand the concept and meaning of Large Language Model (LLMs) and their importance in natural language processing.
• Know about different types of popular LLMs, such as BERT, GPT-3, GPT-4, and T5.
• Discuss the applications and use cases of Open Source LLMs.
• Hugging Face APIs for LLMs.
• Explore the future implications of LLMs, including their potential impact on job markets, communication, and society as a whole.

This article was published as a part of the Data Science Blogathon.

What is a Large Language Model (LLM)?

A large language model is an advanced type of language model that is trained using deep learning techniques on massive amounts of text data. These models are capable of generating human-like text and performing various natural language processing tasks.

In contrast, the definition of a language model refers to the concept of assigning probabilities to sequences of words, based on the analysis of text corpora. A language model can be of varying complexity, from simple n-gram models to more sophisticated neural network models. However, the term “large language model” usually refers to models that use deep learning techniques and have a large number of parameters, which can range from millions to billions. These AI models can capture complex patterns in language and produce text that is often indistinguishable from that written by humans.

Types of LLMs

• Autoregressive Models: These models generate text one token at a time based on the previously generated tokens. Examples include OpenAI’s GPT series and Google’s BERT.
• Conditional Generative Models: These models generate text conditioned on some input, such as a prompt or context. They are often used in applications like text completion and text generation with specific attributes or styles.

What are LLMs used for?

Large language models (LLMs) are finding application in a wide range of tasks that involve understanding and processing language. Here are some of the common uses:

• Content creation and communication: LLMs can be used to generate different creative text formats, like poems, code, scripts, musical pieces, emails, and letters. They can also be used to summarize information, translate languages, and answer your questions in an informative way.
• Analysis and insights: LLMs are capable of analyzing massive amounts of text data to identify patterns and trends. This can be useful for tasks like market research, competitor analysis, and legal document review.
• Education and training: LLMs can be used to create personalized learning experiences and provide feedback to students. They can also be used to develop chatbots that can answer student questions and provide support.pen_spark

How a Large Language Model (LLM) is Built?

A large-scale transformer model known as a “large language model” is typically too massive to run on a single computer and is, therefore, provided as a service over an API or web interface. These models are trained on vast amounts of text data from sources such as books, articles, websites, and numerous other forms of written content. By analyzing the statistical relationships between words, phrases, and sentences through this training process, the models can generate coherent and contextually relevant responses to prompts or queries. Also, Fine-tuning these models involves training them on specific datasets to adapt them for particular applications, improving their effectiveness and accuracy.

ChatGPT’s GPT-3, a large language model, was trained on massive amounts of internet text data, allowing it to understand various languages and possess knowledge of diverse topics. As a result, it can produce text in multiple styles. While its capabilities, including translation, text summarization, and question-answering, may seem impressive, they are not surprising, given that these functions operate using special “grammars” that match up with prompts.

How do LLMs work?

Large language models like GPT-3 (Generative Pre-trained Transformer 3) work based on a transformer architecture. Here’s a simplified explanation of how they Work:

1. Learning from Lots of Text: These models start by reading a massive amount of text from the internet. It’s like learning from a giant library of information.
2. Innovative Architecture: They use a unique structure called a transformer, which helps them understand and remember lots of information.
3. Breaking Down Words: They look at sentences in smaller parts, like breaking words into pieces. This helps them work with language more efficiently.
4. Understanding Words in Sentences: Unlike simple programs, these models understand individual words and how words relate to each other in a sentence. They get the whole picture.
5. Getting Specialized: After the general learning, they can be trained more on specific tasks to get good at certain things, like answering questions or writing about particular subjects.
6. Doing Tasks: When you give them a prompt (a question or instruction), they use what they’ve learned to respond. It’s like having an intelligent assistant that can understand and generate text.

Difference Between Large Language Models and Generative AI

So, generative AI is the whole playground, and LLMs are the language experts in that playground.

Also Read: Basic Tenets of Prompt Engineering in Generative AI

General Architecture

The architecture of Large Language Model primarily consists of multiple layers of neural networks, like recurrent layers, feedforward layers, embedding layers, and attention layers. These layers work together to process the input text and generate output predictions.

• The embedding layer converts each word in the input text into a high-dimensional vector representation. These embeddings capture semantic and syntactic information about the words and help the model to understand the context.
• The feedforward layers of Large Language Models have multiple fully connected layers that apply nonlinear transformations to the input embeddings. These layers help the model learn higher-level abstractions from the input text.
• The recurrent layers of LLMs interpret information from the input text sequentially. These layers hold a hidden state that updates at each time step, enabling the model to capture dependencies between words in a sentence.
• The attention mechanism is another important part of LLMs, which allows the model to focus selectively on different parts of the input text. This self-attention helps the model attend to the input text’s most relevant parts and generate more accurate predictions.

Examples of LLMs

Let’s take a look at some popular large language models(LLM):

Llama 3.1

Another notable LLM is Llama 3.1, a family of language models released by Meta. Available in sizes of 8B, 70B, and 405B parameters, Llama 3.1 excels in multilingual capabilities, math, coding, and tool usage, establishing itself as a significant competitor in the AI landscape. Its open-source nature and diverse parameter options make it a valuable asset for research and experimentation.

How to use Llama 3.1?

Install Required Libraries

!pip install --upgrade transformers
!pip install --upgrade torch
!pip install --upgrade huggingface_hub

Set-up Hugging Face Access Token

!huggingface-cli login

Import Necessary Libraries and Load Llama 3.1

import transformers
import torch

# Define the model ID for LLaMA 3.1
model_id = "meta-llama/Meta-Llama-3.1-8B-Instruct"

# Initialize the text generation pipeline with LLaMA 3.1 model
llama3 = transformers.pipeline(
    "text-generation",
    model=model_id,
    model_kwargs={"torch_dtype": torch.bfloat16},  # Using bfloat16 for efficient computation
    device_map="auto",  # Automatically maps model to available GPU or CPU
)

Create LLaMA 3.1 Completion Access Function
def get_completion_llama(prompt, model_pipeline=llama3):
    messages = [{"role": "user", "content": prompt}]
    response = model_pipeline(
        messages,
        max_new_tokens=2000
    )
    return response[0]["generated_text"][-1]['content']
Let’s try out Llama 3.1


response = get_completion_llama(prompt='Explain Generative AI in 2 bullet points')
display(Markdown(response))

Output:

GPT 4o

In the rapidly evolving field of AI, several large language models (LLMs) stand out for their advanced capabilities and unique features. GPT 4o , one of the most advanced LLMs to date, impresses with its ability to generate human-quality text, translate languages, and write various forms of creative content. Its performance in coding and answering questions is noteworthy, though its large size and computational demands can be challenging for some applications.

How to use GPT 4o?

We will demonstrate a practical use case of GPT4o for text completion.

Install Required Libraries

!pip install openai

Enter Open AI API Key

We enter our Open AI key using the getpass() function so we don’t accidentally expose our key in the code:

from getpass import getpass

OPENAI_KEY = getpass('Enter Open AI API Key: ')

Setup Open AI API Key

Next, we setup our API key to use with the OpenAI library:

import openai

from IPython.display import HTML, Markdown, display

openai.api_key = openai_key

Create ChatGPT Completion Access Function

This function will use the Chat Completion API to access ChatGPT for us and return responses based on GPT-4o mini:

def get_completion_gpt(prompt, model="gpt-4o-mini"):

    messages = [{"role": "user", "content": prompt}]

    response = openai.chat.completions.create(

        model=model,

        messages=messages,

        temperature=0.0, # degree of randomness of the model's output

    )

    return response.choices[0].message.content

Let’s Try Out the GPT-4o Mini

We can quickly test the above function to see if our code can access OpenAI’s servers and use GPT-40 mini:

response = get_completion_gpt(prompt='Explain Generative AI in 2 bullet points')

display(Markdown(response))

Output:

Gemma 2

Gemma 2 from Google offers high efficiency and performance across its 27B and 9B parameter versions. This LLM model outperforms many larger models, including LLama 3, due to its optimization for diverse hardware, enhancing AI accessibility and customization.

How to use Gemma 2?

Install Necessary Libraries

!pip install -q  -U transformers accelerate bitsandbytes huggingface_hub

 Import Necessary Libraries

import torch

from transformers import AutoTokenizer, AutoModelForCausalLM, BitsAndBytesConfig

Set-up Quantization Configuration

quantization_config = BitsAndBytesConfig(
    load_in_4bit=True,
    bnb_4bit_quant_type="nf4",
    bnb_4bit_use_double_quant=True,
    bnb_4bit_compute_dtype=torch.bfloat16
)

Load Tokenizer and Model

tokenizer = AutoTokenizer.from_pretrained("google/gemma-2-9b-it", device="cuda")

model = AutoModelForCausalLM.from_pretrained(

    "google/gemma-2-9b-it",

    quantization_config=quantization_config,

    device_map="cuda")

Prepare Input Text and Tokenize

input_text = "For the below sentence extract the names and \

organizations in a json format\nElon Musk is the CEO of SpaceX"

input_ids = tokenizer(input_text, return_tensors="pt").to("cuda")

Generate Output

outputs = model.generate(**input_ids, max_length = 512)

Decode and Print Output

print(tokenizer.decode(outputs[0], skip_special_tokens=True))

Output:

Claude 3.5 Sonnet

Claude 3.5 Sonnet, Anthropic’s latest generative AI model, excels in reasoning, coding, multilingual tasks, and visual capabilities. With robust safety measures and promises of future advancements through models like Haiku and Opus, Claude 3.5 Sonnet contributes significantly to the ongoing development of AI.

We can also access the model through Anthropic API. It costs $3 / 1 Million tokens, and $15 / 1 Million tokens for input and output respectively.

How to use Claude 3.5 Sonnet?

Let’s see this with Customer Support Chatbot”

Installation of the Anthropic Python Package

pip install anthropic

Import the Anthropic Module

import anthropic

Create an Instance of the Anthropic API Client

client = anthropic.Anthropic(api_key='your_api_key_here')  

# Define a customer support inquiry 

customer_message = "Hi, I need help with resetting my password. Can you guide me?"

# Send the customer support inquiry to the Claude 

model response = client.messages.create( model="claude-3-5-sonnet-20240620", max_tokens=150, messages=[{"role": "user", "content": customer_message}] ) 

# Print the response from the model 

print("AI Response:", response['completion'])

Output:

AI Response: Sure, I can help you with that! Here are the steps to reset your password:

1. Go to the login page of our website.

2. Click on the "Forgot Password?" link below the login form.

3. Enter your registered email address in the provided field.

4. Check your email inbox for a password reset link. If you don't 
see it, check your spam or junk folder.

5. Click on the link and follow the instructions to create 
a new password.

6. Make sure your new password is strong and unique, 
combining letters, numbers, and special characters.

If you encounter any issues or don't receive the reset email, 
please let me know, and I'll assist you further!

Open Source Large Language Model(LLM)

The availability of open-source LLMs has revolutionized the field of natural language processing, making it easier for researchers, developers, and businesses to build applications that leverage the power of these models to build products at scale for free. One such example is Bloom. It is the first multilingual Large Language Model (LLM) trained in complete transparency by the largest collaboration of AI researchers ever involved in a single research project.

With its 176 billion parameters (larger than OpenAI’s GPT-3), BLOOM can generate text in 46 natural languages and 13 programming languages. It is trained on 1.6TB of text data, 320 times the complete works of Shakespeare.

Bloom Architecture

The architecture of BLOOM shares similarities with GPT3 (auto-regressive model for next token prediction), but has been trained in 46 different languages and 13 programming languages. It consists of a decoder-only architecture with several embedding layers and multi-headed attention layers.

Bloom’s architecture is suited for training in multiple languages and allows the user to translate and talk about a topic in a different language. We will look at these examples below in the code.

Other LLMs

We can utilize the APIs connected to pre-trained models of many of the widely available LLMs through Hugging Face.

Hugging Face APIs

Let’s look into how Hugging Face APIs can help generate text using LLMs like Bloom, Roberta-base, etc. First, we need to sign up for Hugging Face and copy the token for API access. After signup, hover over to the profile icon on the top right, click on settings, and then Access Tokens.

Use Case 1: Sentence Completion

Let’s look at how we can use Bloom for sentence completion. The code below uses the hugging face token for API to send an API call with the input text and appropriate parameters for getting the best response.

import requests
from pprint import pprint

API_URL = 'https://api-inference.huggingface.co/models/bigscience/bloomz'
headers = {'Authorization': 'Entertheaccesskeyhere'}
# The Entertheaccesskeyhere is just a placeholder, which can be changed according to the user's access key


def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()
  
params = {'max_length': 200, 'top_k': 10, 'temperature': 2.5}
output = query({
    'inputs': 'Sherlock Holmes is a',
    'parameters': params,
})

print(output)

Temperature and top_k values can be modified to get a larger or smaller paragraph while maintaining the relevance of the generated text to the original input text. We get the following output from the code:

[{'generated_text': 'Sherlock Holmes is a private investigator whose cases '
                    'have inspired several film productions'}]

Let’s look at some more examples using other LLMs.

Use Case 2: Question Answers

We can use the API for the Roberta-base model which can be a source to refer to and reply to. Let’s change the payload to provide some information about myself and ask the model to answer questions based on that.

API_URL = 'https://api-inference.huggingface.co/models/deepset/roberta-base-squad2'
headers = {'Authorization': 'Entertheaccesskeyhere'}


def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()
  
params = {'max_length': 200, 'top_k': 10, 'temperature': 2.5}
output = query({
    'inputs': {
            "question": "What's my profession?",
            "context": "My name is Suvojit and I am a Senior Data Scientist"
        },
    'parameters': params
})

pprint(output)

The code prints the below output correctly to the question – What is my profession?:

{'answer': 'Senior Data Scientist',
 'end': 51,
 'score': 0.7751647233963013,
 'start': 30}

Use Case 3: Summarization

We can summarize using Large Language Models. Let’s summarize a long text describing large language models using the Bart Large CNN model. We modify the API URL and added the input text below:

API_URL = "https://api-inference.huggingface.co/models/facebook/bart-large-cnn"
headers = {'Authorization': 'Entertheaccesskeyhere'}


def query(payload):
    response = requests.post(API_URL, headers=headers, json=payload)
    return response.json()
    
params = {'do_sample': False}

full_text = '''AI applications are summarizing articles, writing stories and 
engaging in long conversations — and large language models are doing 
the heavy lifting.

A large language model, or LLM, is a deep learning model that can 
understand, learn, summarize, translate, predict, and generate text and other 
content based on knowledge gained from massive datasets.

Large language models - successful applications of 
transformer models. They aren’t just for teaching AIs human languages, 
but for understanding proteins, writing software code, and much, much more.

In addition to accelerating natural language processing applications — 
like translation, chatbots, and AI assistants — large language models are 
used in healthcare, software development, and use cases in many other fields.'''

output = query({
    'inputs': full_text,
    'parameters': params
})

print(output)

The output will print the summarized text about LLMs:

[{'summary_text': 'Large language models - most successful '
                  'applications of transformer models. They aren’t just for '
                  'teaching AIs human languages, but for understanding '
                  'proteins, writing software code, and much, much more. They '
                  'are used in healthcare, software development and use cases '
                  'in many other fields.'}]

These were some of the examples of using Hugging Face API for common large language models.

Also Read: How to Build Your AI Chatbot with Conversational AI and NLP in Python?

LLMs vs SLMs

Here’s the information in a table format:

Future Implications of LLMs

In recent years, there has been specific interest in large language model (LLMs) like GPT-3, and chatbots like ChatGPT, which can generate natural language text that has very little difference from that written by humans. These foundation models have seen a breakthrough in the field of artificial intelligence (AI). While LLMs have seen a breakthrough in the field of artificial intelligence (AI), there are concerns about their impact on job markets, communication, and society.

One major concern about LLMs is their potential to disrupt job markets. Large Language Model, with time, will be able to perform tasks by replacing humans like legal documents and drafts, customer support chatbots, writing news blogs, etc. This could lead to job losses for those whose work can be easily automated.

However, it is important to note that LLMs are not a replacement for human workers. They are simply a tool that can help people to be more productive and efficient in their work through automation. While some jobs may be automated, new jobs will also be created as a result of the increased efficiency and productivity enabled by LLMs. For example, businesses may be able to create new products or services that were previously too time-consuming or expensive to develop. By leveraging LLMs, they can optimize processes and improve efficiency, leading to innovation and growth.

LLMs have the potential to impact society in several ways. For example, LLMs could be used to create personalized education or healthcare plans, leading to better patient and student outcomes. LLMs can be used to help businesses and governments make better decisions by analyzing large amounts of data and generating insights.

Conclusion

Large Language Model (LLMs) have revolutionized the field of natural language processing, allowing for new advancements in text generation and understanding. LLMs can learn from big data, understand its context and entities, and answer user queries. This makes them a great alternative for regular usage in various tasks in several industries. However, there are concerns about the ethical implications and potential biases associated with these models. It is important to approach LLMs with a critical eye and evaluate their impact on society. With careful use and continued development, LLMs have the potential to bring about positive changes in many domains, but we should be aware of their limitations and ethical implications.

Hope you like the article and get clear understanding about the llm architecture , llm architecture is explained with these concepts of Large Language models you will clear your doubts.

Key Takeaways

• Large Language Models (LLMs) can understand complex sentences, understand relationships between entities and user intent, and generate new text that is coherent and grammatically correct
• The article explores the architecture of some LLMs, including embedding, feedforward, recurrent, and attention layers.
• The article discusses some of the popular LLMs like BERT, BERT, Bloom, and GPT3 and the availability of open-source LLMs.
• Hugging Face APIs can be helpful for users to generate text using LLMs like Bart-large-CNN, Roberta, Bloom, and Bart-large-CNN.
• LLMs are expected to revolutionize certain domains in the job market, communication, and society in the future.

Frequently Asked Questions

The media shown in this article is not owned by Analytics Vidhya and is used at the Author’s discretion.

Suvojit Hore 01 Sep, 2024