Classroom availability and venue
Classroom training and venue arrangements depend on the availability of a suitable conference hall in Manchester, United Kingdom. Dates, seating capacity, facilities and the final venue are confirmed before booking.
Manchester, United Kingdom
Generative AI
Fine-tuning Open Source LLMsin Manchester
Explore Fine-Tuning Open Source LLMs training to learn model customisation, supervised fine-tuning, parameter-efficient techniques, and dataset preparation. Gain practical experience with Python, Hugging Face Transformers, PyTorch, and…
40 Hours/5 daysLive virtual trainingIntermediate
Classroom training and venue arrangements depend on the availability of a suitable conference hall in Manchester, United Kingdom. Dates, seating capacity, facilities and the final venue are confirmed before booking.
Learning in Manchester, United Kingdom
Build your understanding
Explore Fine-tuning Open Source LLMs
Explore Fine-Tuning Open Source LLMs training to learn model customisation, supervised fine-tuning, parameter-efficient techniques, and dataset preparation. Gain practical experience with Python, Hugging Face Transformers, PyTorch, and Low-Rank Adaptation (LoRA) to adapt language models for specialised applications. Understand model evaluation, performance optimisation, and responsible AI practices. Join StepMerit's Fine-Tuning Open Source LLMs course in Manchester, United Kingdom, to strengthen your Generative AI development skills and build customised AI solutions.
View the complete course informationCourse curriculum
- Module 1: Introduction to Open Source Large Language Models
- Understanding large language models and their capabilities
- Overview of open source and openly available LLMs
- Popular models and ecosystems, including Llama, Mistral, and Hugging Face
- Use cases and limitations of open source LLMs
- Module 2: Transformer Architecture and LLM Fundamentals
- Transformer architecture and attention mechanisms
- Tokenisation, embeddings, and positional information
- Pretraining, inference, and text generation
- Understanding model parameters and context windows
- Module 3: Development Environment Setup
- Setting up Python, Jupyter Notebook, and required libraries
- Working with PyTorch and Hugging Face Transformers
- Configuring GPU, cloud notebook, or local computing environments
- Managing dependencies and model files
- Module 4: Working with Pretrained Open Source Models
- Accessing pretrained models from Hugging Face Hub
- Loading tokenisers and model checkpoints
- Selecting models based on task requirements and hardware capacity
- Running inference and evaluating baseline outputs
- Module 5: Dataset Preparation and Preprocessing
- Collecting and validating task-specific training data
- Cleaning, deduplication, and text normalisation
- Formatting instruction-response datasets
- Splitting datasets into training, validation, and test sets
- Identifying privacy risks and sensitive information in datasets
- Module 6: Fundamentals of LLM Fine-Tuning
- Understanding transfer learning and supervised fine-tuning
- Comparing prompting, retrieval-augmented generation (RAG), and fine-tuning
- Defining training objectives and evaluation criteria
- Understanding loss functions, epochs, batch sizes, and learning rates
- Module 7: Supervised Fine-Tuning (SFT)
- Preparing instruction-tuning datasets
- Configuring training arguments and model checkpoints
- Training a pretrained model using supervised examples
- Tracking training loss and validation performance
- Saving and loading fine-tuned model checkpoints
- Module 8: Parameter-Efficient Fine-Tuning (PEFT)
- Introduction to parameter-efficient adaptation techniques
- Understanding Low-Rank Adaptation (LoRA)
- Exploring Quantised LoRA (QLoRA)
- Configuring adapters and trainable parameters
- Comparing full fine-tuning with parameter-efficient approaches
- Module 9: Fine-Tuning with Hugging Face Tools
- Using Transformers and the Trainer API
- Working with the Datasets and PEFT libraries
- Configuring tokenisers, collators, and training pipelines
- Managing checkpoints, logs, and model artefacts
- Module 10: GPU Memory and Training Optimisation
- Understanding GPU memory consumption during training
- Using quantisation and gradient accumulation
- Exploring mixed-precision training and gradient checkpointing
- Adjusting batch sizes and sequence lengths
- Identifying common training bottlenecks
- Module 11: Model Evaluation and Quality Assessment
- Establishing baseline and post-training performance
- Evaluating generated responses for relevance and accuracy
- Using task-specific metrics and human evaluation
- Testing for hallucinations, bias, and inconsistent outputs
- Performing regression tests across representative prompts
- Module 12: Experiment Tracking and Troubleshooting
- Monitoring training metrics and experiment configurations
- Identifying overfitting and underfitting
- Troubleshooting dataset, tokenisation, and memory errors
- Comparing model checkpoints and training runs
- Documenting experiments for reproducibility
- Module 13: Fine-Tuned Model Inference and Deployment
- Loading fine-tuned models and adapter weights
- Configuring generation parameters and inference settings
- Serving models through Python applications or APIs
- Understanding model quantisation for inference
- Reviewing deployment, latency, and resource requirements
- Module 14: Responsible AI, Security, and Model Governance
- Reviewing training data quality, consent, and licensing considerations
- Assessing bias, privacy, and potential misuse risks
- Protecting model artefacts and sensitive datasets
- Documenting model limitations and intended use
- Establishing human review and ongoing evaluation practices
- Module 15: Capstone Project and Practical Assessment
- Selecting an open source LLM for a defined use case
- Preparing a domain-specific dataset and baseline model
- Implementing supervised fine-tuning or LoRA/QLoRA
- Evaluating and comparing model performance before and after fine-tuning
- Demonstrating the adapted model and documenting results
- Presenting project findings, limitations, and future improvements
- Training Delivery and Learning Support
- Instructor-led learning with practical demonstrations and guided exercises
- Hands-on labs using suitable open source models and tools
- Project-based learning focused on real-world LLM customisation
- StepMerit training support for learners in Manchester, United Kingdom
Exam details
- Exam Level: Intermediate – Fine-Tuning Open Source LLMs
- Exam Type: Online computer-based examination consisting of multiple-choice questions and practical, scenario-based questions.
- Number of Questions: 40 questions.
- Exam Duration: 90 minutes.
- Passing Score: 70% (28 out of 40 questions).
- Exam Delivery: Online assessment through the StepMerit learning and assessment platform.
- Practical Assessment: Learners may be assessed on dataset preparation, fine-tuning configuration, model evaluation, and interpretation of training results.
- Certification: Learners who meet the assessment requirements may receive a StepMerit Certificate of Course Completion. This is a StepMerit course certificate, not an official certification issued by a model provider.
- Exam Preparation: Review course modules, practise fine-tuning workflows, complete hands-on exercises, and understand model evaluation techniques.
- Training and Assessment Support: StepMerit provides course-related learning support for participants in Manchester, United Kingdom.
Plan your learning
Who this course is for
AI Engineers Machine Learning Engineers Generative AI Developers Data Scientists Python Developers NLP Engineers and Practitioners Deep Learning Professionals LLM Application Developers AI Researchers and Research Scholars Software Developers Exploring Open Source LLMs IT Professionals Upskilling in Generative AI Technical Professionals seeking Fine-Tuning Open Source LLMs training in Manchester, United Kingdom
Before you begin
Python Programming: Basic knowledge of Python, including functions, classes, libraries, and data handling. Machine Learning Fundamentals: Understanding of model training, validation, evaluation, and basic machine learning concepts. Deep Learning Knowledge: Familiarity with neural networks, gradient descent, loss functions, and backpropagation is recommended. Large Language Model Fundamentals: Basic understanding of LLMs, transformer architecture, tokens, embeddings, and text generation. Transformer Models: Familiarity with attention mechanisms, pretrained models, and tokenization is beneficial. Python AI Libraries: Basic experience with PyTorch and familiarity with the Hugging Face…
Learning outcomes
Problem Statement: Healthcare organisations manage large volumes of patient education materials, hospital guidelines, frequently asked questions, clinical information documents, and administrative resources. General-purpose large language models may not consistently follow healthcare-specific terminology, response formats, or approved communication guidelines. Healthcare teams need AI systems that can provide clear, accessible, and consistent general health information while protecting sensitive data and avoiding unsupported medical claims. The organisation wanted to explore fine-tuning open source LLMs to support healthcare education and information services. How We Overcame the Problem…
Continue exploring
Insights for Fine-tuning Open Source LLMs
Fine-Tuning Open Source LLMs: A Complete Guide to Custom AI Model Development
Large Language Models (LLMs) are transforming how organisations develop intelligent applications, automate workflows, process information, and deliver personalised user experiences. Although pretrained LLMs can perform a wide range of tasks, businesses often require models that follow specific instructions, understand…
Read the full article