Welcome to CHAT GI-IADS’s documentation!

• Introduction to the documentation
• Overview of Large Language Models
  • Applications of Large Language Models
  • The Predecessors of Large Language Models
  • Rule-Based Systems
    • Advantages
    • Drawbacks
  • Statistical Models
    • Advantages
    • Drawbacks
    • Historical Significance
  • Machine Learning in NLP
  • Word Embeddings in NLP
    • Advantages
    • Drawbacks
    • Historical Significance
  • Neural Networks in NLP
    • Recurrent Neural Networks (RNNs)
    • Convolutional Neural Networks (CNNs)
  • The Emergence of Large Language Models
  • Rise of Transformer-Based Models
    • Impact:
  • Breakthroughs with BERT and GPT
    • BERT
    • GPT
  • Recent Advances
    • Notable Models:
• Understanding the Architecture of Transformers
  • Transformer Architecture Overview
  • Introduction
  • The Components of Transformer Architecture
  • Input Processing
    • Input
    • Tokenization
    • Embedding Layer
    • Positional Encoding
  • Core Mechanisms
    • Self-Attention Layer
    • Multi-Head Attention
  • Network Layers
    • Feedforward Neural Network
    • Residual Connections
    • Layer Normalization
    • Output Layer
  • Types of Transformer Architectures
  • Encoder-Only Models
    • Overview
    • Masked Language Modeling in Encoder-Only Models
      • Masking Tokens
      • Bidirectional Context
      • Objective function
      • Denoising
  • Decoder-Only Models
    • Overview
    • Causal Language Modeling in Decoder-Only Models
      • Sequential Prediction
      • Unidirectional Context
      • Objective Function
      • Text Generation
  • Encoder-Decoder Models
    • Overview
• Applications of LLMs
  • Natural Language Processing (NLP) Tasks
  • Sentiment Analysis
    • Graded Sentiment Analysis (Fine-Grained Sentiment Analysis)
    • Emotion Detection
    • Aspect-Based Sentiment Analysis
  • Named Entity Recognition (NER)
  • Part-Of-Speech Tagging (PoS)
    • Rule-Based POS Tagging
    • Artificial Neural Networks (ANN)
    • Transformation-Based Tagging (TBT)
    • Machine Translation
    • Rule-based Machine Translation or RBMT
    • Corpus-based Machine Translation Approach or CBMT
    • Example-based Machine Translation or EBMT
  • Text Generation and Summarization
    • Text Summarization Techniques ( extractive and abstractive summarization methods enabled by LLMs)
      • Extractive Summarization
      • Abstractive Summarization
    • LLMs relevance in context-aware Text Generation
      • How AI Uses Context for Text Generation?
      • Preceding Text
      • Subsequent Text
      • Training Data
      • Prompt Design
  • Question Answering and Chatbots
    • Examples of Existing LLM-powered chatbots
    • Applications of LLM-powered Chatbots
      • Customer Support
      • Education and Tutoring
      • Healthcare Assistance
      • Information Retrieval
      • E-commerce and Shopping
• Synthetic Data Generation & Data preprocessing
  • Synthetic data generation for industrial applications
    • Importance of synthetic data
    • 1. Overcoming Data Scarcity
    • 2. Improving Model Robustness
    • 3. Accelerating Development
    • 4. Data Protection and Privacy:
    • 5. Data Augmentation:
    • 6. Data Availability and Accessibility:
  • Techniques for generating synthetic data
  • Drawing Numbers from a Distribution
  • Agent-based Modeling (ABM):
  • Prompts:
  • Data preprocessing for LLMs
  • Data: What’s It All About?
  • Data Preprocessing and LLMs
  • Data cleaning
  • Data cleaning methods :
  • 1. Handling missing values
  • 2. Noise reduction
  • 3. Consistency checks
  • 4. Deduplication
  • Text cleaning and normalization
  • Removing noise and normalization
  • Tokenization and word embedding
  • 1.Tokenization
  • 2.Word Embedding
  • 1. One-Hot Encoding
  • 2. Word2Vec
  • 3. Bag of Words (BOW)
  • 4. GloVe (Global Vectors for Word Representation)
  • 5. FastText
  • 6. Papers related to Data cleaning
• Computational Challenges in Training Large Language Models
  • The computational requirements for Training LLMs
    • Graphics Processing Unit (GPU)
    • Central Processing Unit (CPU)
    • Random Access Memory(RAM)
    • Storage
    • Networking
    • Distributed Computing
  • Existing Optimization techniques
    • Model pruning
    • Quantization
    • Parallelization
• Fine-tuning techniques
  • How to Fine-tune a Large Language Model (LLM)
  • Instruction Fine-Tuning :
  • Catasrophic Forgetting :
  • How to prevent Catastrophic Forgetting :
  • Instruction Multi-Fine-Tuning :
  • PEFT
  • Why PEFT?
  • High performance on consumer hardware
  • Low-Rank Adaptation (LoRA)
  • Why LoRA?
  • Implementation in Transformers
  • Qlora
  • Implementing QLoRA
  • Putting everything together
• Benchmarking and Evaluation
  • Metrics for Evaluating The LLM-Powered Chatbot
    • 1. Accuracy
    • 2. Intent Recognition Accuracy
    • 3. Fallback
    • 4. Task Completion Rate
    • 5. Engagement Rate
    • 6. Response Time
  • Benchmarking LLM Performance
    • Do’s and dont’s section
    • 1. Comparison with Baselines
    • 2. Task-Specific Benchmarks
    • 3. LLM’s efficiency in resource utilization
  • Similarity Search
    • Semantic Similarity in Similarity Search
    • Cosine Similarity
    • BLEU Score
    • METEOR Score
  • Using Similarity Metrics to Validate Fine-Tuning
• Prompt engineering
  • Zero shot :
  • One shot :
  • Few shot :
  • Chain of Thouths :
  • self-consistency :
  • Active prompt :
• Retrieval-Augmented Generation (RAG)
  • Introduction
  • Methodology
  • Retriever
  • Generator
  • Training
  • Putting it all together:
• langchain
• Reinforcement Learning from Human Feedback
  • Reinforcement Learning
  • Practical example
  • Reinforcement Learning For Language Models
  • Automated Decision-Making
  • PPO Algorithm
  • Conclusion
• Mistral 7B:Use and fine-tune
  • Understanding Mistral 7B
  • Accessing Mistral 7B
  • Mistral 7B Fine-tuning
    • Setting up
    • Data loading
    • Loading the Mistral 7B model
    • Loading the Tokenizer
    • Adding the adopter to the layer
  • Hyperparmeters
    • Model training
    • Saving the fine-tuned model
    • Model evaluation