DSV4 / DeepSeek-V4-Flash 推理测速项目

记录 vllm-dspark、SGLang、vLLM 等后端在 DeepSeek-V4-Flash 上的 benchmark 脚本与最终结果。 H200 快速上手指南见 docs/H200_QUICKSTART.md。

目录说明

目录/文件	说明
platforms/	芯片/加速器平台配置（*.env），如 P800、H200
scripts/common/	跨实验复用的 orchestration 组件（server 启停、health check、元数据生成）
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）
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）

注：.gitignore 已排除 envs/、deps/、tmp/、datasets/、logs/、*.zip、原始 raw_outputs/*.jsonl 等大文件/中间文件。

实验索引

新结构（experiments/）

实验	脚本	说明
DSV4 P800 SGLang	experiments/dsv4_p800_sglang/run_bench.sh	Kunlun P800 + SGLang + DeepSeek-V4-Flash-INT8
DSV4 H200 DSpark	experiments/dsv4_h200_dspark/run_bench.sh	NVIDIA H200 + vllm-dspark + DeepSeek-V4-Flash-DSpark
DSV4 H200 vLLM baseline	experiments/dsv4_h200_vllm/run_bench.sh	NVIDIA H200 + vLLM + DeepSeek-V4-Flash
DSV4 H200 SGLang baseline	experiments/dsv4_h200_sglang/run_bench.sh	NVIDIA H200 + native SGLang + DeepSeek-V4-Flash
DSV4 H200 SGLang vs vLLM	experiments/dsv4_h200_sglang_vs_vllm/run_bench.sh	NVIDIA H200 上 SGLang 与 vLLM 控制变量对比（TP=8，最长 200k 上下文）
DSV4 H200 vLLM DSpark vs default	experiments/dsv4_h200_vllm_dspark_vs_default/run_bench.sh	vLLM 开启 DSpark 投机解码 vs 默认配置，验证 TTFT 差异

旧结构（scripts/ + bench_results/）

实验	脚本	最终报告	说明
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
SGLang 8-card systematic	scripts/run_sglang_benchmark.sh	experiments/legacy_bench_results/sglang_8card_systematic_20260704_120819/report.md	早期 SGLang 系统扫描
SGLang 8-card max throughput	scripts/run_sglang_max_throughput.sh	experiments/legacy_bench_results/sglang_8card_max_throughput_20260705_030839/summary.json	最大吞吐扫描
vLLM-dspark Qwen3	scripts/bench_vllm_dspark_qwen3.py	experiments/legacy_bench_results/vllm_dspark_qwen3_20260705_121256/summary.json	Qwen3 模型测速
DSV4 后端对比（0705）	scripts/bench_dsv4_comparison.py	experiments/legacy_bench_results/dsv4_comparison_20260705_152221/summary.json	早期 DSpark 配置对比

快速复现

P800 SGLang

bash experiments/dsv4_p800_sglang/run_bench.sh

H200 DSpark

bash experiments/dsv4_h200_dspark/run_bench.sh

H200 SGLang baseline

bash experiments/dsv4_h200_sglang/run_bench.sh

H200 SGLang vs vLLM 对比

bash experiments/dsv4_h200_sglang_vs_vllm/run_bench.sh

H200 vLLM DSpark vs default

bash experiments/dsv4_h200_vllm_dspark_vs_default/run_bench.sh

DSpark grid（旧结构）

bash scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh

Spec-tokens 对比

bash scripts/benchmark_dspark_0707/run_dspark_st_comparison.sh

解析已有结果

# 新结构
python3 experiments/dsv4_p800_sglang/parse_results.py \
  experiments/dsv4_p800_sglang/results/<run_id>

python3 experiments/dsv4_h200_dspark/parse_results.py \
  experiments/dsv4_h200_dspark/results/<run_id>

# 旧结构
/data/user1/yy/envs/sglang/bin/python scripts/benchmark_dspark_0707/parse_results.py \
  /data/user1/yy/experiments/legacy_bench_results/dspark_grid_20260707-132641

SLO 参考

详见 scripts/SLO_STANDARDS.md。

主要关注指标：

• TTFT P95：S2（DeepSeek-V4-Flash 所在层）目标 < 3s
• TPOT：S2 目标 < 50ms

环境要求

• NVIDIA H200：Python env /data/user1/yy/envs/vllm-dspark（服务端）、/data/user1/yy/envs/sglang（压测客户端）；模型 /data/models/DeepSeek-V4-Flash、/data/models/DeepSeek-V4-Flash-DSpark。
• Kunlun P800：Docker 镜像 iregistry.baidu-int.com/xpu/sglang-p800-pd-disagg-0510:20260511_4202；模型 /data1/models/DeepSeek-V4-Flash-INT8；压测客户端在容器内运行。

平台相关常量见 platforms/*.env。