OwnerSunshine530 6bb51a1057 revert+feat: triton退回contiguous(去contiguous非连续读更慢) + embedding_bag默认开(消unique同步) ...

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

2026-06-17 13:54:31 +08:00

..

tests

feat: F.embedding_bag 融合查表+池化(单kernel,免[M,512]中间) — 攻最大块(dedup index25%+segment11%=36%)

2026-06-17 13:30:47 +08:00

bench.py

feat: F.embedding_bag 融合查表+池化(单kernel,免[M,512]中间) — 攻最大块(dedup index25%+segment11%=36%)

2026-06-17 13:30:47 +08:00

build_env.sh

fix: build_env.sh 简化为纯净版本（避免 CUDA 预热导致异常）

2026-06-12 21:55:09 +08:00

EXPERIMENTS.md

docs: 收尾 — 最终67.998/记录RepEncoder预计算尝试与结论

2026-06-16 13:18:48 +08:00

infer.py

revert+feat: triton退回contiguous(去contiguous非连续读更慢) + embedding_bag默认开(消unique同步)

2026-06-17 13:54:31 +08:00

requirements.txt

revert: requirements.txt 还原为原始完整依赖列表

2026-06-12 21:24:22 +08:00

RISKS.md

docs: 潜在风险说明(RepEncoder预计算合规灰区/max_feasign一致性)与合规保底

2026-06-15 20:44:57 +08:00