diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-08-03 09:04:08 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-08-03 09:04:08 -0500 |
| commit | 2ec65555120378de3ff8298bed97852b143bd802 (patch) | |
| tree | 1358b1cde700daf563ff057c647e836ecf2e28f7 /TINYSTORIES_EQPROP_ANALOG_TRAINER_MVP.md | |
| parent | d8a9489f65a67ec26b56d9c0173f5f8f00aa47da (diff) | |
RESULT 85: EP成本定案 — 受控实测wall-clock 2.70×(纯训练步2.77×), 解析FLOPs 9F/3F=3.0×互证
口径清理: 项目页/call sheet的"2.0×"与给Suhas的"1.0-1.4×"均作废(后者疑将holofast+sdpa对EP
自身的~1.5×加速误作对BP比值); 七月brief的3.2× FLOPs一直正确。KHS对比(6.7-12×)不受影响,
EP家族内仍有2.5-4.4×成本优势。方法: SIGSTOP静音同卡, 同模型同batch各400步, 两窗口复核。
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'TINYSTORIES_EQPROP_ANALOG_TRAINER_MVP.md')
0 files changed, 0 insertions, 0 deletions