diff --git a/BENCHMARK_WORKFLOW.md b/BENCHMARK_WORKFLOW.md index ff15ad0..6a2f352 100644 --- a/BENCHMARK_WORKFLOW.md +++ b/BENCHMARK_WORKFLOW.md @@ -108,7 +108,7 @@ bench_results/_/ Examples: -- `bench_results/dspark_grid_20260707-132641/` (legacy result kept in original location) +- `experiments/legacy_bench_results/dspark_grid_20260707-132641/` (legacy result moved under `experiments/`) - `experiments/legacy_bench_results/dsv4_backend_comparison_20260707/` (legacy result moved under `experiments/`) ### Raw output files diff --git a/README.md b/README.md index 0be6c4f..e7c5804 100644 --- a/README.md +++ b/README.md @@ -12,7 +12,7 @@ | `scripts/analysis/` | 跨实验对比分析工具 | | `scripts/` | 旧版/legacy benchmark 脚本 | | `experiments/` | 以实验为单位的自包含目录（脚本 + 配置 + 结果） | -| `bench_results/` | 历史归档的实验产物（只读），目前仅保留 `dspark_grid_20260707-132641` | +| `experiments/legacy_bench_results/` | 历史归档的实验产物（只读），从原 `bench_results/` 迁移而来 | | `BENCHMARK_WORKFLOW.md` | benchmark 目录与命名规范 | | `scripts/SLO_STANDARDS.md` | 推理服务 SLO 标准（TTFT/TPOT） | @@ -32,7 +32,7 @@ | 实验 | 脚本 | 最终报告 | 说明 | |---|---|---|---| -| DSpark grid benchmark | `scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh` | `bench_results/dspark_grid_20260707-132641/report.md` | P1/P2/P3 全量网格（唯一保留在 `bench_results/` 的结果） | +| DSpark grid benchmark | `scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh` | `experiments/legacy_bench_results/dspark_grid_20260707-132641/report.md` | P1/P2/P3 全量网格 | | DSpark spec-tokens 对比 | `scripts/benchmark_dspark_0707/run_dspark_st_comparison.sh` | `experiments/legacy_bench_results/dspark_st_comparison_20260707-150649/comparison_report.md` | `--spec-tokens 3` vs `5` | | SGLang EAGLE vs DSpark | `scripts/start_sglang_dsv4_8card.sh` + 对比解析脚本 | `experiments/legacy_bench_results/eagle_grid/dspark_vs_eagle_report.md` | EAGLE 投机解码对比 | | SGLang vs vLLM 后端对比（0707） | `scripts/benchmark_dsv4_backend_comparison.sh` | `experiments/legacy_bench_results/dsv4_backend_comparison_20260707/README.md` | 遗留 raw outputs 清单 | @@ -79,7 +79,7 @@ python3 experiments/dsv4_h200_dspark/parse_results.py \ # 旧结构 /data/user1/yy/envs/sglang/bin/python scripts/benchmark_dspark_0707/parse_results.py \ - /data/user1/yy/bench_results/dspark_grid_20260707-132641 + /data/user1/yy/experiments/legacy_bench_results/dspark_grid_20260707-132641 ``` ## SLO 参考 diff --git a/experiments/legacy_bench_results/README.md b/experiments/legacy_bench_results/README.md index e5b6594..6a9698c 100644 --- a/experiments/legacy_bench_results/README.md +++ b/experiments/legacy_bench_results/README.md @@ -2,12 +2,13 @@ This directory contains historical benchmark results that were originally stored under `bench_results/` before the `experiments/` layout was introduced. -Only `bench_results/dspark_grid_20260707-132641/` remains in the original location per project decision; everything else was moved here for archival. +All legacy results have been moved here; `bench_results/` itself no longer exists. ## Contents | Directory | Original location | Description | |---|---|---| +| `dspark_grid_20260707-132641/` | `bench_results/dspark_grid_20260707-132641/` | P1/P2/P3 DSpark grid benchmark | | `dspark_st_comparison_20260707-150649/` | `bench_results/dspark_st_comparison_20260707-150649/` | `--spec-tokens 3` vs `5` comparison | | `dsv4_backend_comparison_20260707/` | `bench_results/dsv4_backend_comparison_20260707/` | SGLang vs vLLM backend comparison raw outputs manifest | | `dsv4_comparison_20260705_152221/` | `bench_results/dsv4_comparison_20260705_152221/` | Early vLLM-dspark configuration comparison | diff --git a/bench_results/dspark_grid_20260707-132641/evaluation.md b/experiments/legacy_bench_results/dspark_grid_20260707-132641/evaluation.md similarity index 98% rename from bench_results/dspark_grid_20260707-132641/evaluation.md rename to experiments/legacy_bench_results/dspark_grid_20260707-132641/evaluation.md index 70d0cf7..d0663c9 100644 --- a/bench_results/dspark_grid_20260707-132641/evaluation.md +++ b/experiments/legacy_bench_results/dspark_grid_20260707-132641/evaluation.md @@ -1,6 +1,6 @@ # vllm-dspark Benchmark 结果评估 -> 对应结果：`/data/user1/yy/bench_results/dspark_grid_20260707-132641` +> 对应结果：`/data/user1/yy/experiments/legacy_bench_results/dspark_grid_20260707-132641` > 运行时间：2026-07-07 > 硬件：8× NVIDIA H200 143GB > 模型：`/data/models/DeepSeek-V4-Flash-DSpark` diff --git a/bench_results/dspark_grid_20260707-132641/report.md b/experiments/legacy_bench_results/dspark_grid_20260707-132641/report.md similarity index 99% rename from bench_results/dspark_grid_20260707-132641/report.md rename to experiments/legacy_bench_results/dspark_grid_20260707-132641/report.md index 6f8dc81..6b0ce76 100644 --- a/bench_results/dspark_grid_20260707-132641/report.md +++ b/experiments/legacy_bench_results/dspark_grid_20260707-132641/report.md @@ -1,6 +1,6 @@ # vllm-dspark Benchmark Grid Report -- Result root: `/data/user1/yy/bench_results/dspark_grid_20260707-132641` +- Result root: `/data/user1/yy/experiments/legacy_bench_results/dspark_grid_20260707-132641` - Model: `/data/models/DeepSeek-V4-Flash-DSpark` - Backend: vllm-dspark (TP=8, FP8 KV cache, spec-method=dspark, spec-tokens=5) - Benchmark client: `sglang.bench_serving --backend vllm` diff --git a/scripts/benchmark_dspark_0707/parse_results.py b/scripts/benchmark_dspark_0707/parse_results.py index 1d177cd..d85fefa 100644 --- a/scripts/benchmark_dspark_0707/parse_results.py +++ b/scripts/benchmark_dspark_0707/parse_results.py @@ -51,7 +51,7 @@ def parse_log(log_path: Path) -> dict: def main(): - result_root = Path(sys.argv[1]) if len(sys.argv) > 1 else Path("/data/user1/yy/bench_results/dspark_grid_20260707-132641") + result_root = Path(sys.argv[1]) if len(sys.argv) > 1 else Path("/data/user1/yy/experiments/legacy_bench_results/dspark_grid_20260707-132641") if not result_root.exists(): print(f"Result root not found: {result_root}", file=sys.stderr) sys.exit(1)