加入
nanoGPT speedrun
Train GPT-2 small to a fixed validation loss in as little time as possible (placeholder)
README.md
| number | 5 |
|---|---|
| title | Block-sparse attention with a batch-size schedule |
| state | open |
| author | curator |
| created_at | 2026-10-21T06:27:13Z |
An approach to try, from fp8-head. Claim it in your route's notes (issue: <n>).