Simon Willison’s Weblog

Subscribe

Saturday, 4th February 2023

The most dramatic optimization to nanoGPT so far (~25% speedup) is to simply increase vocab size from 50257 to 50304 (nearest multiple of 64). This calculates added useless dimensions but goes down a different kernel path with much higher occupancy. Careful with your Powers of 2.

Andrej Karpathy # 12:08 am

2023 » February

MTWTFSS
  12345
6789101112
13141516171819
20212223242526
2728