Intro to Hugging Face: Fine-tuning BERT for NLP tasks

on YouTube ↗02h 10m

" summary

Shows step‑by‑step fine‑tuning of a public BERT checkpoint via Hugging Face + PyTorch in Jupyter notebooks: tokenization, masked‑LM and next‑sentence objectives, then fine‑tuning on Rotten Tomatoes dataset (they downsample to ~100 train / 20 test for demos) using DistilBERT as a faster base; baseline performance is poor with tiny samples but improves with more data. Includes demos of GPT‑2 (repetitive) vs chatGPT (better via RLHF), and Stable Diffusion image prompts; notebooks run in Google Colab and expose clear dataset and bias failure modes (e.g. gender and name stereotypes).

AI-written summary · about

" chapters (12)

1. 0:00Introduction
2. 0:46Embeddings and deep learning
3. 10:39Transformer history
4. 21:58Hugging Face overview
5. 23:40Audience NLP interests
6. 30:43Transfer learning and fine-tuning
7. 40:57Fine-tuning setup with BERT
8. 56:00BERT for sentiment analysis
9. 1:28:48GPT-2 text generation demo
10. 1:45:24Stable Diffusion images
11. 1:59:12AI bias and risks
12. 2:09:14Q&A

" related talks

• 32mAn Introduction to Transfer Learning in NLP and HuggingFaceThomas Wolf · ML Conf EU 2020
• 20mGetting Started with Pytorch 2.0 and Hugging Face TransformersPhilipp Schmid · PyTorch Conference 2023
• 58mAn Introduction to Transfer Learning and Hugging FaceThomas Wolf · Data Science fwdays'20 online
• 49mMachine Learning 2.0 With Hugging FaceJulien Simon · PyData London 2022
• 43mDiscovering the BERT family of modelsAlexia Audevart · DevFest Toulouse 2025
• 25mEasy Access to SOTA NLP Models with Ray and Hugging FaceThomas Wolf · Ray Summit 2020
• 33mMachine Learning 2.0 with Hugging FaceJulien Simon · Conf42 Machine Learning 2022
• 33mBuild Better AI with First-Party Data and Fine-TuningPaul Yang · AI By the Bay 2025