8: Deep Learning for Natural Language – Transformers, Self-Supervised Learning

Name: 8: Deep Learning for Natural Language – Transformers, Self-Supervised Learning
Uploaded: 2026-01-07T00:00:00.000Z
Duration: 1 h 16 min 46 s
Description: This lecture completes the 2017 transformer encoder: it recasts last class’s parameter-free self-attention as GPU-friendly matrix math, then adds three “industrial” pieces—learnable Q/K/V matrices, residual connection...

MIT OpenCourseWare · 76:46

This lecture completes the 2017 transformer encoder: it recasts last class’s parameter-free self-attention as GPU-friendly matrix math, then adds three “industrial” pieces—learnable Q/K/V matrices, residual connection...

Read the full summary on tuber