PodcastIntel
Sign in Get Started Free
Andrej Karpathy
Andrej Karpathy

Building makemore Part 4: Becoming a Backprop Ninja

Oct 11, 2022
Listen to Episode
✨ AI Summary
  • Manually backpropagate through MLP, cross-entropy loss, and BatchNorm.
  • Understand gradient flow through compute graphs and tensors.
  • Build competence in deep learning mechanics without autograd.

More from Andrej Karpathy

Andrej Karpathy
How I use LLMs
Feb 27, 2025
Andrej Karpathy
Deep Dive into LLMs like ChatGPT
Feb 5, 2025
Andrej Karpathy
Let's reproduce GPT-2 (124M)
Jun 9, 2024
Andrej Karpathy
Let's build the GPT Tokenizer
Feb 20, 2024
View all episodes →

Get AI Summaries for Every New Episode

Subscribe to Andrej Karpathy and get AI summaries, guest tracking, and email digests delivered automatically.

Sign Up Free →