[Docs] narrow Phase 3 artifact scope

.gitattributes

@@ -0,0 +1 @@
+*.nsys-rep filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,5 +1,9 @@
 # sskj — 多平台大模型推理性能基准测试项目
 
+> **更新（2026-08-03 10:02:07 CST）**
+>
+> 将 DeepSeek-V4-Pro 双机 Pro6000D SGLang Phase 3 正式 Run `dsv4pro-phase3-20260802-003213` 的六份原始 `.nsys-rep` 纳入 GitLab `hzy` 分支，并通过 Git LFS 保存；普通 Git 仅保存六份报告的 SHA256 清单。Bench 文本、请求响应、服务日志、环境快照不随原始 Trace 上传；约 2.44 GB 的临时 `.sqlite` 也不上传，因为它可由 `.nsys-rep` 重新导出且不构成独立证据。
+>
 > **更新（2026-08-02 01:25:30 CST）**
 >
 > 完成 DeepSeek-V4-Pro 双机 Pro6000D SGLang Phase 3 时间线 Profiling。正式 Run `dsv4pro-phase3-20260802-003213` 在 21 分 03 秒内完成 Decode Control、Decode+128K Prefill Mixed、独立 128K Prefill 三段 Capture；Head/Worker 各生成 3 份 Nsight Systems 报告及 Stats，合计 6/6 校验通过。Control 以 GEMM/MoE（Head 44.7%）和 NCCL（31.9%）为主；Mixed 中 NSA/MLA 升至 29.7%，NCCL 为 23.1%，说明 Chunked Prefill 的 Sparse Attention/MLA 计算进入同一 TP16 GPU 关键路径。Prefill Trace 漏采 3 个 Rank 的 CUDA 活动，因此暂不判断固定慢 Rank，也不将 NCCL 单独定性为根因。实验结束后两节点容器和 16 张 GPU 均已清理。

experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/README.md

@@ -68,3 +68,11 @@ results/<RUN_ID>/
 
 Profiler runs are diagnostic and must not replace the Phase 2 no-profiler TPS
 baseline.
+
+## Archived formal run
+
+The six raw reports for `dsv4pro-phase3-20260802-003213` are committed under
+`results/` on the `hzy` branch through Git LFS, together with their SHA256
+manifest. Benchmark text, request/response payloads, service logs, environment
+snapshots, and temporary `.sqlite` exports are intentionally excluded from the
+GitLab archive. The SQLite exports can be regenerated from the reports.

experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/REPORTS.sha256

@@ -0,0 +1,6 @@
+1b650ba174b8b9b5a66d43df148de1af8731061fcc097ebc1a4c35606cd366e9  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/head/nsys/timeline.1.nsys-rep
+4637abef08b2d0477e3bbc8003ac79ad6f0f047747e8cd160aa13b460fd8f6d6  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/head/nsys/timeline.2.nsys-rep
+9c1a678c4164bebecdb0607ab41ee819eaf81540725bb85b0ad89a94e953d8c2  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/head/nsys/timeline.3.nsys-rep
+272b8e37208ceba50ac03d85f120a9613ad6baac4911b2bc649324ded49fb2e4  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/worker/nsys/timeline.1.nsys-rep
+a845cb3cfe4e870658e77fa67deca848b36b71373dadb4b9c1353a8f70eea8d5  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/worker/nsys/timeline.2.nsys-rep
+b99603f20e25efc73360cc027fc7a529b4ea6f234dbafb95f54b3bb7d6f27d70  experiments/pro6000/dsv4pro_pro6000d_2node_sglang_timeline_profiling/results/dsv4pro-phase3-20260802-003213/profiles/worker/nsys/timeline.3.nsys-rep