- Unité d'enseignement : Theory and practical applications of large language models
Nombre de crédits de l'UE : 3
Code APOGEE : INF2511M
Type d'enseignement
Nb heures *
Cours Magistraux (CM)
18 h
Travaux Pratiques (TP)
12 h
* Ces horaires sont donnés à titre indicatif.
Pré-requis :
M1 in computer science, having completed a traditional computer science program (incluging training in symbolic AI and machine learning).
Compétences attestées (transversales, spécifiques) :
Large language models (LLMs) have recently gained popularity due to their fluency and ability to generalise to many tasks and domains using in-context learning. This course explains the theory behind transformer-based LLMs, their training, and how they can be used and evaluated in practice. We cover a large variety of state-of-the art techniques as well as the ethical questions surrounding LLMs.
Programme de l'UE / Thématiques abordées :
- Part 1: Zoom on the transformer architecture
The RNN Architectures
The Transformer Model for NLG
- Part 2: A short history of large language models
A quick survey of existing LLMs
Running LLMs on Consumer-Grade Computers
- Part 3: Pre-training and Post-training
Pre-training
Post-training/Fine-tuning
- Part 4: Practical techniques and tools for large language models
Prompt engineering
Retrieval Augmented Generation
Useful tools for large language models
- Part 5: Evaluation and ethical questions of large language models
Evaluating LLMs
Security, privacy, and ethical questions of LLM
- Part 6: Exciting research of large language models (content of this section is updated continuously)
Parcours / Spécialité / Filière / Option utilisant cette UE :
Date de la dernière mise-à-jour : 24/10/2024
SELECT MEN_ID, `MEN_DIP_ABREVIATION`, `MEN_TITLE`, `PAR_TITLE`, `PAR_ID` FROM parcours INNER JOIN ue_parcours ON PAR_ID_FK=PAR_ID INNER JOIN mention ON MEN_ID = PAR_MENTION_FK WHERE PAR_ACTIVATE = 0 AND UE_ID_FK='31013' ORDER BY `MEN_DIP_ABREVIATION`, `MEN_TITLE`, `PAR_TITLE`
