diff options
| author | Yuren Hao <yurenh2@illinois.edu> | 2026-07-13 06:53:00 -0500 |
|---|---|---|
| committer | Yuren Hao <yurenh2@illinois.edu> | 2026-07-13 06:53:00 -0500 |
| commit | f71f449e89b505286861021bde1af8faa8e41660 (patch) | |
| tree | 5e417c29df8d3b694f7b5b2ea18d82130c3e9a05 /ep_run/dip_screen.py | |
| parent | 4404b945a7cccc52b8a69c0fbbd96776d4a967b0 (diff) | |
Manual DDP for cascade-EP: guard-synced (drift/kretry via global max, gn on reduced grads), sigma broadcast, per-rank data RNG, None-preserving flat all-reduce, bitwise param sync-check, grad-equivalence test PASSED (cos 0.999999999 vs single-GPU big batch)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_014FAPDWQ49M5Ye3NpTndTpn
Diffstat (limited to 'ep_run/dip_screen.py')
0 files changed, 0 insertions, 0 deletions