Record TP2 prefill serving-path verdict

This commit is contained in:
2026-07-23 18:08:32 +08:00
parent cf610003ed
commit 4f22688bfd
9 changed files with 2013 additions and 5 deletions

View File

@@ -1,6 +1,6 @@
# 实验 EXP-TP2-PREFILL-SERVINGTP2 base-prefill residual 是否来自 serving-path MoE
> **状态:** GPU smoke harness 已冻结,待远端执行
> **状态:** 完成;机制 PASSglobal constant injection FAIL
>
> Parent campaign[`../frontier-simulator-gap-campaign-v0/README.md`](../frontier-simulator-gap-campaign-v0/README.md)
@@ -84,10 +84,26 @@
real=`410 ms`bias=`13.11%`TP4=`231.33 vs 231 ms`。TP2
component ledger MoE=`171.12 ms`,若单独解释 residual 需增至约
`224.9 ms`+31.4%)。
- **观察事实** GPU
- **观察事实**
- TP2 q8192 profile execute=`408.19 ms`,冻结 real anchor=`410 ms`
两个 TP rank 分别为 `408.19/407.34 ms`
- critical-rank attention=`99.67 ms`sim attention execution=`100.24 ms`
serving MoE=`214.52 ms`sim MoE=`171.12 ms`
- MoE delta=`43.41 ms`,解释 total residual `80.75%`MoE-only
counterfactual total=`399.65 ms`(相对 real `2.52%`)。三个事前 smoke
gate 全部 PASS
- 但把 q8k ratio `1.25366×` 用作所有 TP2 prefill shape 的常数 scale
ρ=0.0025 TTFT mean `4.54%→+5.23%`p99
`7.67%→+0.89%`E2E mean `+12.70%→+15.55%`ρ=0.005
TTFT mean `7.08%→+3.13%`E2E mean `+5.02%→+9.14%`
- **Fleet preflight2026-07-23** dash1--dash4 均为 8×H2032 张卡
`memory.used=0 MiB``utilization=0%` compute processdry-run 选择
`dash1:[0,1]`正式 job pin dash1
- **含义** GPU
- **Claim update** unchanged
- **下一步** probe fleet运行 TP2 serving-path prefill smoke
- **含义** TP2 base residual 的主要机制确实是 serving-path MoE而不是
attention real-anchor 噪声但单个 q8k 点不能外推成全 prefill-domain
constant calibration
- **Claim update** TP2 prefill MoE 尚有明显工程可优化 gap得到支持
用一个 TP2 常数 scale 即可修复 trace fidelity被否定
- **下一步** 将后续工程项收窄为 TP2 token/routing-conditioned serving
MoE curve至少 q2k/q4k/q8k 与真实 routing allocation不合入当前
global scalecampaign 继续实验 3