Mastering Generative AI and LLM Deployment.

Overview

Proficiency on OpenAI,LangChain,MidJourney,LLama3,;Javascript Applications for 20X Fast Inference Prototypes.Get Hired What you'll learn:

What is Docker and How to use DockerAdvance Docker UsageWhat are OpenCL and OpenGL and when to use ?(LAB) Tensorflow and Pytorch Installation, Configuration with Docker(LAB)DockerFile, Docker Compile and Docker Compose Debug file configuration(LAB)Different YOLO version, comparisons, and when to use which version of YOLO according to your problem(LAB)Jupyter Notebook Editor as well as Visual Studio Coding Skills(LAB)Learn and Prepare yourself for full stack and c++ coding exercies(LAB)TENSORRT PRECISION FLOAT 32/16 MODEL QUANTIZIATIONKey Differences:

Explicit vs. Implicit Batch Size(LAB)TENSORRT PRECISION INT8 MODEL QUANTIZIATION(LAB) Visual Studio Code Setup and Docker Debugger with VS and GDB Debugger(LAB) what is ONNX framework C Plus and how to use apply onnx to your custom C ++ problems(LAB) What is TensorRT Framework and how to use apply to your custom problems(LAB) Custom Detection, Classification, Segmentation problems and inference on images and videos(LAB) Advance C ++ Object Oriented Programming(LAB) Deep Learning Problem Solving Skills on Edge Devices, and Cloud Computings with C++ Programming Language(LAB) How to generate High Performance Inference Models on Embedded Device, in order to get high precision, FPS detection as well as less gpu memory consumption(LAB) Visual Studio Code with Docker(LAB) GDB Debugger with SonarLite and SonarCube Debuggers(LAB) yolov4 onnx inference with opencv c++ dnn libraries(LAB) yolov5 onnx inference with opencv c++ dnn libraries(LAB) yolov5 onnx inference with Dynamic C++ TensorRT Libraries(LAB) C++(11/14/17) compiler programming exerciesKey Differences:

OpenCV AND CUDA/ OPENCV AND TENSORRT(LAB) Deep Dive on React Development with Axios Front End Rest API(LAB) Deep Dive on Flask Rest API with REACT with MySqlUnderstand model optimization techniques:

Pruning, Distillation, and Quantization(LAB) Deep Dive on Text Summarization Inference on Web App(LAB) Prompt Penetration Testing(LAB) Deep Dive on BERT (LLM) Fine tunning and Emotion Analysis on Web App(LAB) Deep Dive On Distributed GPU Programming with Natural Language Processing (Large Language Models))(LAB) Deep Dive on BERT (LLM) Fine tunning and Emotion Analysis on Web App(LAB) Prompt Engineering from basics to advance(LAB) Deep Dive on Generative AI use cases, project lifecycle, and model pre-training(LAB) OPENAI GPT models with Specific prompt Engineering techniques(LAB) Fine-tuning and evaluating large language models(LAB) Reinforcement learning and LLM-powered applications, ALIGN Fine tunning with User Feedback(LAB) Quantization of Large Language Models with Modern Nvidia GPU's(LAB) C++ OOP TensorRT Quantization and Fast Inference(LAB) Deep Dive on Hugging FACE Library(LAB)Translation ● Text summarization ● Question answering(LAB)Sequence-to-sequence models, ONLY Encoder Based Models, Only Decoder Based Models(LAB)Define the terms Generative AI, large language models, prompt, and describe the transformer architecture that powers LLMs(LAB)Discuss computational challenges during model pre-training and determine how to efficiently reduce memory footprint(LAB)Describe how fine-tuning with instructions using prompt datasets can improve performance on one or more tasks(LAB)Explain how PEFT decreases computational cost and overcomes catastrophic forgetting(LAB)Describe how RLHF uses human feedback to improve the performance and alignment of large language models(LAB)Discuss the challenges that LLMs face with knowledge cut-offs, and explain how information retrieval and augmentation techniques can overcome these challenRecognize and understand the various strategies and techniques used in fine-tuning language models for specialized applications.Master the skills necessary to preprocess datasets effectively, ensuring they are in the ideal format for AI training.Investigate the vast potential of fine-tuned AI models in practical, real-world scenarios across multiple industries.Acquire knowledge on how to estimate and manage the costs associated with AI model training, making the process efficient and economicDistributing Computing for (DDP) Distributed Data Parallelization and Fully Shared Data Parallel across multi GPU/CPU with Pytorch together with Retrieval AugmeThe RoBERTa model was proposed in RoBERTa:

A Robustly Optimized BERT Pretraining ApproachMaster downcasting from FP32 to BF16 and FP32 to INT8Learn the difference between symmetric and asymmetric quantizationImplement quantization techniques in Python with real examplesApply quantization to make models more efficient and deployment-readyLearn the basics of data types like FP32, FP16, BFloat16, and INT8Gain practical skills to optimize models for edge devices and resource-constrained environmentsAdvance Image Generation and Editing This course is diving into Generative AI State-Of-Art Scientific Challenges.

It helps to uncover ongoing problems and develop or customize your Own Large Models Applications. Course mainly is suitable for any candidates(students, engineers,experts) that have great motivation to Large Language Models with Todays-Ongoing Challenges as well as their deeployment with Python Based and Javascript Web Applications, as well as with C/C++ Programming Languages.

Candidates will have deep knowledge on TensorFlow , Pytorch, Keras models, HuggingFace with Docker Service. In addition, one will be able to optimize and quantize TensorRT frameworks for deployment in variety of sectors.

Moreover, They will learn deployment of LLM quantized model to Web Pages developed with React, Javascript and FLASKHere you will also learn how to integrate Reinforcement Learning(PPO) to Large Language Model, in order to fine them with Human Feedback based. Candidates will learn how to code and debug in C/C++ Programming languages at least in intermediate level.LLM Models used:

The Falcon, LLAMA2, BLOOM, MPT, Vicuna,FLAN-T5, GPT2/GPT3, GPTNEOXBERT 101, Distil BERTFINE-Tuning Small Models under supervision of BIG ModelsImage Generation :

LLAMA modelsGemini Dall-E OpenAIHugging face ModelsLearning and Installation of Docker from scratchKnowledge of Javscript, HTML ,CSS, BootstrapReact Hook,DOM and Javacscript Web DevelopmentDeep Dive on Deep Learning Transformer based Natural Language ProcessingPython FLASK Rest API along with MySqlPreparation of DockerFiles, Docker Compose as well as Docker Compose Debug fileConfiguration and Installation of Plugin packages in Visual Studio CodeLearning, Installation and Confguration of frameworks such as Tensorflow, Pytorch, Kears with docker images from scratchPreprocessing and Preparation of Deep learning datasets for training and testingOpenCV DNNwith C++ InferenceTraining, Testing and Validation of Deep Learning frameworksConversion of prebuilt models to Onnx and Onnx Inference on images with C++ ProgrammingConversion of onnx model to TensorRT engine with C++ RunTime and Compile Time APITensorRT engine Inference on images and videosComparison of achieved metrices and result between TensorRT and Onnx InferencePrepare Yourself for C++ Object Oriented Programming Inference!Ready to solve any programming challenge with C/C++ Read to tackle Deployment issues on Edge Devices as well as CloudAreasLarge Language Models Fine TunningLarge Language Models Hands-On-Practice:

BLOOM, GPT3-GPT3.5, FLAN-T5 familyLarge Language Models Training, Evaluation and User-Defined Prompt IN-Context Learning/On-Line LearningHuman FeedBack Alignment on LLM with Reinforcement Learning (PPO) with Large Language Model :

BERT and FLAN-T5How to Avoid Catastropich Forgetting Program on Large Multi-Task LLM Models.How to prepare LLM for Multi-Task Problems such as Code Generation, Summarization, Content Analizer, Image Generation.Quantization of Large Language Models with various existing state-of-art techniquesImportante Note:

In this course, there is not nothing to copy &paste, you will put your hands in every line of project to be successfully LLM and Web Application Developer!You DO NOT need any Special Hardware component.

You will be delivering project either on CLOUD or on Your Local Computer.

Syllabus

• Introduction to Generative AI and LLM Deployment
Overview of Generative AI Concepts
Introduction to Large Language Models (LLMs)
Course Objectives and Outcomes
• Docker Basics and Advanced Usage
What is Docker and How to Use It
DockerFile, Docker Compile, and Docker Compose
Configuration and Debugging of Docker Compose Files
• OpenCL, OpenGL, and Deep Learning Frameworks
Overview of OpenCL and OpenGL
Installation and Configuration of TensorFlow and PyTorch with Docker
Jupyter Notebook and Visual Studio Code for Development
• Deployment and Optimization Techniques
TensorRT Precision: FP32/FP16 and INT8 Model Quantization
Model Optimization: Pruning, Distillation, and Quantization
ONNX Framework: Application in Custom C++ Problems
• Advanced C++ Programming for AI
C++ Object-Oriented Programming
Fast Inference on Edge Devices and Cloud Computing
OpenCV DNN with C++ Inference
• YOLO Models for Object Detection
Comparisons of YOLO Versions
YOLOv4 and YOLOv5 ONNX Inference with C++ and TensorRT
• Large Language Models and Fine-Tuning
Overview of Models: Falcon, LLAMA2, BLOOM, MPT, Vicuna, FLAN-T5, GPT-2/3
Fine-tuning and Evaluating Large Models with User Feedback
Quantization Techniques for LLMs
• Frontend and Backend Development for AI Deployment
React Development with Axios and REST APIs
Flask REST API Development with MySQL
Integration and Deployment of Models to Web Applications
• Natural Language Processing with Transformers
Deep Dive on Transformer Architecture
Prompt Engineering from Basics to Advanced
Sequence-to-sequence Models: Encoder-Based and Decoder-Based
• Advanced Use Cases in Generative AI
Generative AI in Image Generation and Editing
State-of-the-Art Challenges in Generative AI
Project Lifecycles and Model Pre-training
• Distributed Computing and Multi-GPU Processing
Distributed Data Parallelization and Fully Shared Data Parallel
Retrieval Augmentation Techniques with Distributed Computing
• Practical Application and Hands-On Labs
C++ Programming Exercises
Deep Dive on Text Summarization and BERT Fine-tuning
Debugging in Visual Studio Code with GDB and SonarCube
• Reinforcement Learning with Human Feedback
Applying PPO to LLMs
Addressing Catastrophic Forgetting in Multi-Task Models
• Real-World Scenarios and Industry Applications
Preprocessing Data for AI Training
Optimizing Models for Edge Devices
Economic Management of Model Training Costs
• Conclusion and Next Steps
Course Recap and Key Takeaways
Paths for Further Learning and Development
Final Projects and Capstone Presentations

---

Taught by

PhD Researcher AI & Robotics Scientist Fikrat Gasimov

---

Subjects

Computer Science