للحصول على شهادة
This course “LLMs from Scratch – Practical Engineering from Base Model to PPO RLHF” is an advanced, hands-on learning path focused on building and aligning large language models from the ground up. It is designed for learners who want to understand not only how LLMs are built, but also how they are improved and aligned with human preferences.
The course begins with the fundamentals of constructing a base language model from scratch, including data preparation, tokenization, model architecture, and training workflows. Learners gain a clear understanding of how modern transformer-based models are built and optimized.
It then moves into model training strategies, explaining how large-scale models learn language patterns through pretraining on massive datasets.
A key focus of the course is reinforcement learning from human feedback (RLHF), which is used to align models with human intent and improve response quality. Learners explore how human preferences are incorporated into training pipelines.
The course also introduces PPO (Proximal Policy Optimization), a core algorithm used in RLHF to fine-tune language models safely and effectively.
In addition, learners study the full alignment pipeline, from base model creation to reward modeling and policy optimization.
By the end of this course, learners will understand both how LLMs are built from scratch and how they are aligned to behave more usefully and sa