Fine-tuning of LLM models: adaptation to a business use case

This two-day training guides you step by step in adapting large language models to your business use cases. You will learn to prepare your data, configure and launch fine-tuning sessions, and evaluate and deploy models tailored to your specific context.

Learning Objectives:

• Understand the differences between pre-training and fine-tuning LLMs.
• Prepare and annotate a relevant dataset for your business scenario.
• Configure fine-tuning parameters (learning rate, batch size, epochs).
• Evaluate and compare the performance of your customized models.

Target Audience:
Developers, data scientists, and ML engineers looking to adapt an LLM to precise business needs.

Duration:
2 days (9:00–12:30 & 13:30–17:00).

Number of Participants:
Maximum 12 participants.

Prerequisites:
Proficiency in Python, basic ML knowledge, and initial experience with an LLM.

Program:

Day 1: Preparation & Configuration

• Overview of fine-tuning techniques.
• Data collection and cleaning for training.
• Dataset annotation and structuring.
• Tool setup (Hugging Face Transformers, OpenAI API).

Day 2: Execution & Evaluation

• Launching fine-tuning sessions and monitoring logs.
• Model evaluation (metrics and qualitative tests).
• Choosing deployment strategies (inference, hosting).
• Workshop: deploy a custom fine-tuned model.

Pricing:

• Open enrollment: €1,400 excl. VAT per participant
• In-house: on request, tailored to your needs

Methods Employed:

• Theoretical presentations and demos
• Guided practical workshops
• Fine-tuning mini-project

Evaluation:

• Knowledge validation quiz
• Mini-project review and recommendations

Accessibility:
Our sessions are accessible to people with disabilities. Please contact accessibility@eurekia-learning.com to arrange accommodations.

Contacts:

• Quality referent: Jihane Khouzaimi (06 69 53 77 75 – contact@eurekia-learning.com)
• Pedagogical referent: Hatim Khouzaimi

Note:
This training provides a comprehensive framework to fine-tune and deploy high-performance language models adapted to your sector.