К содержимому
learnspaceYOUR NEXT CHAPTER
ПРОСТРАНСТВО ОБУЧЕНИЯ
ГлавнаяКаталог курсовМоё обучениеCoursera

Знания без границ

Учитесь у лучших университетов и компаний мира.

Открыть Coursera
Интеграция
Пространство университета
Моё пространствоСтраница курса
↵
ЯЛичный кабинетСтудент
© 2026 LearnSpaceКаждый день — возможность узнать больше.Помощь
Generative AI: Fine-Tuning LLMs and Diffusion Models · LearnSpace
Назад в каталог
courseraАнализ данных

Generative AI: Fine-Tuning LLMs and Diffusion Models

Курс от Board Infinity
Средний≈ 19.2 чАнглийский
О курсеНавыкиПрограммаПреподаватели

О курсе

"Master Generative AI with hands-on training in Large Language Models (LLMs), PEFT techniques (LoRA, QLoRA), and Diffusion Models using Hugging Face, diffusers, peft, trl, and bitsandbytes. This course takes you from the internals of decoder-only transformers to building a specialist fine-tuned LLM and generating high-quality, controllable images with ControlNet. In Module 1, explore decoder-only transformer architectures, self-attention, causal masking, KV caching, and token flow mechanics. Module 2 focuses on Parameter-Efficient Fine-Tuning (PEFT), where you'll implement LoRA, QLoRA, and 4-bit quantization to fine-tune large models on consumer GPUs using SFT pipelines. Module 3 dives into diffusion models, covering forward/reverse processes, UNet, schedulers (DDIM, Euler, DPM++), and ControlNet conditioning. Module 4 is a capstone where you'll build a Specialist LLM — from dataset creation to adapter export and evaluation. By the end of this course, you will: - Build and optimize decoder-only transformer pipelines with KV caching - Fine-tune 7B+ LLMs using LoRA, QLoRA, and SFT pipelines on limited hardware - Configure diffusers pipelines with ControlNet for controllable image generation - Train, export, and evaluate a domain-specialized LLM adapter end-to-end" Disclaimer: This is an independent educational resource created by Board Infinity for informational and educational purposes only. This course is not affiliated with, endorsed by, sponsored by, or officially associated with any company, organization, or certification body unless explicitly stated. The content provided is based on industry knowledge and best practices but does not constitute official training material for any specific employer or certification program. All company names, trademarks, service marks, and logos referenced are the property of their respective owners and are used solely for educational identification and comparison purposes.

Навыки, которые вы освоите

Model OptimizationFine-tuningGenerative Model ArchitecturesModel EvaluationLarge Language ModelingGenerative AITransfer LearningData ScienceLLM ApplicationHugging FaceModel DeploymentToken OptimizationMachine LearningModel Training

Программа курса

4 модулей · 73 учебных материалов

01Transformer Internals & Decoder-Only Architectures18 материалов

Career Scope in LLM Engineering

Where LLM Engineers Work TodayВидеоWhy Decoder-Only Models Dominate Industry Part 1ВидеоWhy Decoder-Only Models Dominate Industry part -2ВидеоSkills Employers WantВидеоSkills Employers Want Part 2

Учитесь у экспертов

Board Infinity

Instructor

Generative AI: Fine-Tuning LLMs and Diffusion Models
В каталоге вашей программы

Инвестируйте в себя

Новые знания — в удобное для вас время.

Начать на Coursera

Обучение откроется на Coursera
в новой вкладке

Обучение на Coursera

≈ 19.2 ч

4 модулей

Язык: Английский

Субтитры: Арабский, Французский, Итальянский, Бразильский португальский, Корейский, Немецкий, Индонезийский, Испанский, Японский

Часть программы вашего университета
Видео
The LLM Engineering Landscape (2026)Чтение
Career Scope in LLM EngineeringЗадание

Anatomy of Decoder-Only Transformers

What Happens When You Type a PromptВидеоDecoder Block Components (Mental Model)ВидеоHow Tokens Flow Through LayersВидеоInside the Decoder Block: Architecture & Data FlowЧтениеAnatomy of Decoder-Only TransformersЗадание

Self-Attention, Causal Masking & KV-Cache Behavior

Causal Masking & "No Peeking"ВидеоWhy Attention Is ExpensiveВидеоKV Cache: Why Inference Gets FasterВидеоKV Cache Optimization & Attention Behavior AnalysisЧтениеSelf-Attention, Causal Masking & KV-Cache BehaviorЗаданиеTransformer Internals & Decoder-Only ArchitecturesЗадание
02PEFT - LoRA, QLoRA, & SFT Pipelines19 материалов

Introduction to PEFT & Low-Rank Adaptation

The Core Idea Behind PEFTВидеоThe Core Idea Behind PEFT part 2ВидеоThe Core Idea Behind PEFT Part3ВидеоImplementing LoRA Layers in Attention BlocksВидеоWhere PEFT Works (and Where It Doesn't)ВидеоLoRA Fundamentals & Design DecisionsЧтениеIntroduction to PEFT & Low-Rank AdaptationЗадание

QLoRA & 4-Bit Quantization Pathway

LoRA ExplainedВидеоQLoRA: Training Big Models on Small GPUsВидеоSetting QLoRA Hyperparameters for StabilityВидеоQLoRA Implementation Guide for Large ModelsЧтениеQLoRA & 4-Bit Quantization PathwayЗадание

Building SFT Pipelines with peft + trl

What Good SFT Data Looks Like Part 1ВидеоWhat Good SFT Data Looks Like Part2ВидеоBuilding Full SFT Pipelines Using TRLВидеоEarly Evaluation & Failure SignalsВидеоConstructing Reliable SFT Datasets for Behavior ModelingЧтениеBuilding SFT Pipelines with peft + trlЗадание
03Diffusion Models & Image Generation17 материалов

The Forward & Reverse Diffusion Process

Adding Noise: The Forward ProcessВидеоRemoving Noise: The Reverse ProcessВидеоWhy Timesteps MatterВидеоMathematical Intuition Behind Diffusion TimestepsЧтениеThe Forward & Reverse Diffusion ProcessЗадание

Configuring diffusers Pipelines

Diffusers Pipeline ComponentsВидеоDiffusers Pipeline Components Part2ВидеоChoosing the Right SchedulerВидеоStyle, Guidance & Sampling TricksВидеоScheduler Comparison & Practical RecommendationsЧтениеConfiguring diffusers PipelinesЗадание

ControlNets & Conditioning Techniques

Why Prompting Alone Is Not EnoughВидеоControlNet Concepts (Visual)ВидеоPractical: Conditioning an Image with ControlNetВидеоControlNet for Structured Image GenerationЧтениеControlNets & Conditioning TechniquesЗаданиеDiffusion Models & Image GenerationЗадание
04The Hands-On Project - The Specialist LLM19 материалов

Dataset Creation & Prompt-Style Engineering

Converting Logs Into SFT-Ready Training DataВидеоDesigning Prompt Templates & Chat FormatsВидеоCleaning, Normalizing & Validating Training DataВидеоCleaning, Normalizing & Validating Training Data Part 2ВидеоCleaning, Normalizing & Validating Training Data Part 3ВидеоDataset Quality Standards for Persona-Level LLMsЧтениеDataset Creation & Prompt-Style EngineeringЗадание

Training the Specialist LoRA/QLoRA Adapter

Training Workflow Using LoRA/QLoRAВидеоMonitoring Training Loss & Detecting OverfittingВидеоMemory-Saving TechniquesВидеоTroubleshooting PEFT Training FailuresЧтениеTraining the Specialist LoRA/QLoRA AdapterЗадание

Exporting, Loading & Evaluating the Adapter

Exporting LoRA/QLoRA Adapters for DeploymentВидеоExporting LoRAQLoRA Adapters for Deployment part 2ВидеоLoading the Adapter Into Base Models for InferenceВидеоEvaluating Alignment: Perplexity, Behavior, ScenariosВидеоEvaluation Framework for Domain-Specialized LLMsЧтениеExporting, Loading & Evaluating the AdapterЗадание
PEFT - LoRA, QLoRA, & SFT PipelinesЗадание
The Hands-On Project - The Specialist LLMЗадание