[issue8] Alltoall combine 三模式位级仿真与理论性能上限决策表 - #31
Open
shsaihdsaiudh wants to merge 1 commit into
Open
Conversation
shsaihdsaiudh
force-pushed
the
issue8-combine-modes
branch
from
July 26, 2026 09:08
df5415a to
689c1a3
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
关联 Issue
#8
提交内容
MoE combine 阶段三种归约模式的精度/流量权衡分析,拆成两层:
src/code/issue8/combine_modes.py:对应 DeepEP direct combine 三种模式语义的 BF16/FP16 位级精度仿真与闭式流量模型;src/code/issue8/run_analysis.py:扫描 top-k 的 rank 重复率,生成 JSON / CSV 与模式决策表。三种模式按 issue 描述建模:
精度侧以 FP32 全量回传后归约为高精度基线,报告各模式相对基线的数值误差;流量侧给出闭式的网络返回字节数;时延侧使用
固定时延 + 返回净荷 / 有效带宽模型。验证
python -m pytest src/code/issue8/test_combine_modes.py -q # 3 passed python src/code/issue8/run_analysis.py默认扫描 6 档 rank 重复率,结果写入
src/code/issue8/results/,其中decision_table.md是按"top-k 重复率 / 消息大小 / 精度要求"选择 combine 模式的决策表,结论由同目录生成的数据支撑。当前边界
本 PR 完整覆盖了流量与数值精度两个维度的对比,以及验收要求 2 的决策表。
通信完成时间是模型值,不是实测值——使用上述解析时延模型算出,文件内已显式标注为理论值。验收要求 1 中的"combine 通信完成时间"若要给出真机数字,需要多 rank GPU 环境跑通 DeepEP combine 后回填;同样,三种模式的真实 kernel 路径与 NVLink/RDMA 流量也需在多 rank 环境复核。提交者当前为单卡环境。