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.
Recent articles
- Weeknotes: the aftermath of NICAR - 16th March 2024
- The GPT-4 barrier has finally been broken - 8th March 2024
- Prompt injection and jailbreaking are not the same thing - 5th March 2024
- Interesting ideas in Observable Framework - 3rd March 2024
- Weeknotes: Getting ready for NICAR - 27th February 2024
- The killer app of Gemini Pro 1.5 is video - 21st February 2024