hands-on-LLMs

This repo covers the entire workflow from developing/fine-tuning Large Language models for text classification & summarization, making inferences and evaluation and then deploying to production using Mlflow as part of LLMOps

Enviornment:

Azure Databricks

Architecture

Align LLMs with Human Values: Reinforcement Learning from Human Feedback (RLHF)

Intelligent Document Search and Retrieval- Graph RAG (Knowledge Graphs + RAG)

MLOPs on Databricks System Design

Content:

• LLM introduction and Databricks
• Applications with LLMs- Classification and Generative AI using Hugging Face pre-trained models like T5-small/base
• Embeddings, Vector Databases, and Search
• Multi-Stage Reasoning (LLM chains, Prompts, Agents, Langchain)
• Fine-Tuning (Traditional fine-tuning, PEFT, Instruct-finetuning, RLHF)
• Task-Specific Evaluation (Accuracy, Precision, Recall, Rouge Score, BLEU)
• Ethical Evaluation & Biasness- Impact on society(Risks, Toxicity, Hallucinations, Mitigation)
• LLMOPs (MLOPs vs LLMOPs, model loading, Inference pipeline- Delta Tables, MLFLOW (experiment tracking & Model Registry), Orchestration using Delta Live Tables, Deployment to production from dev pipelines.)