Writings
- Attention is all you need, Unpacked 10.2026
A from-scratch walkthrough of the Transformer.
- 274 MNIST Runs 07.2026
A brute-force exploration of architecture, learning rate, dead ReLUs, and pruning across 274 MNIST training runs.
A from-scratch walkthrough of the Transformer.
A brute-force exploration of architecture, learning rate, dead ReLUs, and pruning across 274 MNIST training runs.