Sahar Banisafar

A deep dive into the architecture that revolutionized NLP. We break down attention mechanisms, positional encodings, and the mathematical foundations that make transformers work, with practical examples using PyTorch.

A practical guide to fine-tuning pretrained language models, covering dataset preparation, handling class imbalance, custom data collators, and training loops in PyTorch with HuggingFace Transformers. Includes full code and explanations of the mechanics beyond high-level APIs.