Generative AI: Fine-Tuning LLMs and Diffusion Models
Adapt large language models with parameter-efficient fine-tuning - LoRA, QLoRA and SFT pipelines - and build controllable image generation with diffusion models and ControlNet.
Generative AI: Fine-Tuning LLMs and Diffusion Models
By Board Infinity · Offered on Coursera
- 4 modules · 45 videos · 16 assignments
- Hands-on project - a specialist LLM adapter
- Shareable certificate on Coursera
- Flexible schedule - learn at your own pace
Decoder-only transformers
Token flow, causal masking and the KV cache
Parameter-efficient fine-tuning
LoRA, QLoRA and SFT with peft and trl
Controllable diffusion
Diffusers pipelines, schedulers and ControlNet
Specialist LLM capstone
Train, export and evaluate your own adapter
Inside the model, then adapting it.
Decoder internals first, then parameter-efficient fine-tuning, controllable diffusion - and a specialist LLM you build yourself.
Generative AI: Fine-Tuning LLMs and Diffusion Models starts inside decoder-only transformers: token flow, self-attention, causal masking and why the KV cache makes inference faster.
Then you adapt large language models with LoRA and QLoRA, including 4-bit quantization, and build supervised fine-tuning pipelines with peft and trl. Module 3 moves to diffusion: the forward and reverse processes, diffusers pipelines, schedulers and ControlNet conditioning.
It closes with a hands-on project: a specialist LLM built through dataset creation, QLoRA training and adapter export, with rigorous evaluation.
For practitioners ready to go beyond calling an API.
A basic understanding of Python, PyTorch, and neural networks is helpful. Familiarity with transformers is a plus but not mandatory - the course covers decoder internals from scratch.
Exploring LLM fine-tuning
Moving into advanced LLM fine-tuning with parameter-efficient methods.
Applying PEFT
Using LoRA and QLoRA to adapt models on limited hardware.
Going under the hood
Working with decoder internals, causal masking and KV caching.
Images and adapters
Exploring controllable image generation and building domain-specialized LLM adapters.
Four capabilities, one per module.
The four learning outcomes Coursera lists for this course.
Decoder pipeline & KV cache
Build decoder-only transformer pipelines with KV caching optimizations
LoRA/QLoRA fine-tuning
Fine-tune 7B+ LLMs using LoRA and QLoRA on consumer GPUs
Diffusion & ControlNet
Configure diffusers pipelines with ControlNet for controllable images
Specialist LLM adapter
Train, export, and evaluate a domain-specialized LLM adapter
Fine-tune with someone in your corner.
Board Infinity 1:1 coaching pairs you with a practitioner while you take the course - a separate, optional add-on to your Coursera enrollment.
Book a demo session- A coach matched to your goal in 24 hours
- A written roadmap you keep
- Weekly sessions for as long as you want them
60-minute demo session · 100% refundable · no obligation to continue
Get unstuck faster
Talk through the training run that will not converge.
A second pair of eyes
Review your data, hyperparameters and evaluation.
A roadmap you keep
Structure what to learn next around your goals.
Aim it at a role
Discuss which generative AI path fits you.
Four modules, from decoder internals to a specialist LLM.
Two weeks at 10 hours a week, fully self-paced. Every module lists its description, what is inside it, and every lesson exactly as titled on Coursera.
The stack you work in, and the skills you leave with.
Official tools and skills are exactly as the Coursera course page lists them. The wider library stack is shown separately.
Hugging Face
transformers, peft, trl and diffusers - the libraries behind every module.
Generative AI
Decoder-only LLMs and diffusion models for image generation.
peft + bitsandbytes
LoRA and QLoRA adapters, with 4-bit quantization for large models.
diffusers
Diffusion pipelines, schedulers and ControlNet conditioning.
Libraries named in the course FAQ: transformers · peft · trl · bitsandbytes · diffusers. Also covered in the syllabus: LoRA, QLoRA, SFT pipelines, KV caching and ControlNet. Only Hugging Face and Generative AI are Coursera's official “Tools you'll learn”.
The nine skills the Coursera course page lists, grouped the way the work splits - the models themselves, training them, and measuring the result.
Models & architectures
What generative models are made of.
- Large Language Modeling
- Generative Model Architectures
- LLM Application
Training & tuning
Adapting a model to a domain.
- Fine-tuning
- Model Training
- Model Optimization
Evaluation & data science
Knowing whether it actually worked.
- Model Evaluation
- Machine Learning
- Data Science
Proof you can link to.
Coursera lists a shareable certificate for this course, with Board Infinity as the provider.
Not a university degree, university credit or regulated professional qualification.
Add it to LinkedIn
Coursera says the certificate can be added to your LinkedIn profile.
On your Accomplishments page
From there you can print it, as Coursera describes.
Earned, not attended
Backed by 16 assignments, including the hands-on project.
Part of a specialization
Course 2 of the Advanced Deep Learning Architectures Specialization.
Fine-tuning is where generic models become useful ones.
These are directions where the skills from this course apply - not guaranteed job outcomes.
LLM engineering
Decoder internals, KV caching and the inference behaviour of large language models.
Generative AI development
Working with both LLMs and diffusion models in one toolset.
Model fine-tuning
LoRA, QLoRA and SFT pipelines for adapting large models efficiently.
AI application development
Domain-specialized adapters loaded into base models for inference.
Diffusion image workflows
Diffusers pipelines, schedulers and ControlNet-conditioned generation.
ML experimentation & evaluation
Monitoring training loss, detecting overfitting and evaluating with perplexity and scenarios.
A focused course, not the whole stack.
This course goes deep on four things. It is one part of a three-course specialization, and some topics deliberately live elsewhere.
Course 2 of 3. Here is the full path.
Advanced Deep Learning Architectures: advanced architectures → fine-tuning and generative AI → optimized inference and deployment.
Deep Learning: Advanced Backbones and Efficient GPU Training
ConvNeXt and Vision Transformers, RMSNorm, SwiGLU and RoPE, mixed precision and distributed training.
View course Course 2 · You are here · 19 hoursGenerative AI: Fine-Tuning LLMs and Diffusion Models
Decoder internals and KV caching, LoRA and QLoRA, SFT pipelines, diffusion with ControlNet and a specialist LLM project.
Enroll on Coursera Course 3 · 21 hoursDeploying Deep Learning: Quantization, Serving, and Edge AI
INT4/INT8 quantization, AWQ, GPTQ and GGUF, vLLM and NVIDIA Triton, ONNX Runtime, Llama.cpp and Docker.
View courseHours are the specialization page's listing. Board Infinity runs 270+ courses and 57 specializations on Coursera - browse the full catalogue.
Know exactly what you're signing up for.
Common questions
Answered from the Coursera course page.
A basic understanding of Python, PyTorch, and neural networks is helpful. Familiarity with transformers is a plus but not mandatory — we cover decoder internals from scratch.
You'll work with Hugging Face transformers, peft, trl, bitsandbytes, and diffusers for LLM fine-tuning and image generation.
The course page lists 2 weeks to complete at 10 hours a week, on a flexible schedule. The specialization page separately lists this course at 19 hours.
To access course materials, assignments, and earn a Certificate, you'll need to purchase the Certificate experience when you enroll in a course. Eligible learners may also have the option to start with a Free Trial. Some courses may also offer a Full Course, No Certificate option. This lets you access course materials, submit required assessments, and receive a final grade, but you won't be able to earn or purchase a Certificate.
When you enroll in the course, you get access to all of the courses in the Specialization, and you earn a certificate when you complete the work. Your electronic Certificate will be added to your Accomplishments page - from there, you can print your Certificate or add it to your LinkedIn profile.
Yes. In select learning programs, you can apply for financial aid or a scholarship if you can’t afford the enrollment fee. If fin aid or scholarship is available for your learning program selection, you’ll find a link to apply on the description page.
Want someone building this with you?
Board Infinity 1:1 coaching pairs you with a practitioner while you take the course.
Adapt the model. Control the output.
Decoder internals, LoRA and QLoRA, diffusion with ControlNet, and a specialist LLM of your own.
Enroll on Coursera