From 8652a685e613efdab93ff6a6d5ffea7e0e2565cb Mon Sep 17 00:00:00 2001 From: yy-fighting Date: Fri, 17 Jul 2026 06:18:05 +0000 Subject: [PATCH] rewrite README, add new platform onboarding guide, fix broken scripts/common paths - rewrite README with project purpose, standard workflow, corrected index - add docs/NEW_PLATFORM_GUIDE.md (new GPU onboarding SOP, GLM5.2 reuse) - fix ../../scripts/common -> ../../../scripts/common in 42 experiment scripts - refresh stale docs (EXPERIMENT_GUIDE, H200_QUICKSTART, ADAPTIVE_CONCURRENCY_USAGE, BENCHMARK_WORKFLOW) - remove dead code (dp_proxy.py) and .bak leftovers - add p800 adaptive results (tp4_dp2/tp8_dp1 metrics + summary) - gitignore envs/charts and .tmp_charts --- .gitignore | 4 + BENCHMARK_WORKFLOW.md | 129 ++- README.md | 266 +++--- docs/EXPERIMENT_GUIDE.md | 44 +- docs/H200_QUICKSTART.md | 50 +- docs/NEW_PLATFORM_GUIDE.md | 139 ++++ envs/README.md | 1 + experiments/ADAPTIVE_CONCURRENCY_USAGE.md | 18 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- .../dsv4_h20_sglang_tp_dp_matrix/run_bench.sh | 6 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- ...e_concurrency_add16.sh.bak.20260716-084832 | 167 ---- .../dsv4_h20_vllm_tp_dp_matrix/run_bench.sh | 6 +- .../h200/dsv4_h200_256k_4k_probe/run_bench.sh | 10 +- .../dsv4_h200_64k_sglang_vs_vllm/run_bench.sh | 10 +- .../h200/dsv4_h200_dspark/run_bench.sh | 4 +- .../run_bench.sh | 8 +- .../dsv4_h200_max_context_length/run_bench.sh | 4 +- .../h200/dsv4_h200_sglang/run_bench.sh | 4 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- .../run_bench.sh | 6 +- .../dsv4_h200_sglang_vs_vllm/run_bench.sh | 10 +- experiments/h200/dsv4_h200_vllm/run_bench.sh | 4 +- .../run_bench.sh | 4 +- .../run_bench.sh | 8 +- .../run_bench.sh | 4 +- .../run_bench.sh | 4 +- .../run_bench.sh | 4 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- .../dsv4_h200_vllm_tp_dp_matrix/run_bench.sh | 6 +- .../p800/dsv4_p800_256k_4k_probe/run_bench.sh | 8 +- .../dsv4_p800_256k_4k_probe/start_server.sh | 2 +- .../run_bench.sh | 6 +- .../start_sglang.sh | 4 +- .../dsv4_p800_max_context_length/run_bench.sh | 4 +- .../start_server.sh | 4 +- .../p800/dsv4_p800_sglang/run_bench.sh | 8 +- .../p800/dsv4_p800_sglang/start_server.sh | 2 +- .../adaptive_points.jsonl | 120 +++ .../adaptive_shapes.jsonl | 46 ++ .../adaptive_summary.jsonl | 47 ++ .../adaptive_summary.md | 53 ++ .../run_manifest.json | 7 +- .../metrics/sglang_c16_i1024_o1024_a1.json | 25 + .../metrics/sglang_c16_i1024_o2048_a1.json | 25 + .../metrics/sglang_c16_i1024_o256_a1.json | 25 + .../metrics/sglang_c16_i1024_o4096_a1.json | 25 + .../metrics/sglang_c16_i1024_o512_a1.json | 25 + .../metrics/sglang_c16_i131072_o128_a1.json | 25 + .../metrics/sglang_c16_i16384_o1024_a1.json | 25 + .../metrics/sglang_c16_i16384_o128_a1.json | 25 + .../metrics/sglang_c16_i16384_o2048_a1.json | 25 + .../metrics/sglang_c16_i16384_o256_a1.json | 25 + .../metrics/sglang_c16_i16384_o512_a1.json | 25 + .../metrics/sglang_c16_i4096_o1024_a1.json | 25 + .../metrics/sglang_c16_i4096_o128_a1.json | 25 + .../metrics/sglang_c16_i4096_o2048_a1.json | 25 + .../metrics/sglang_c16_i4096_o256_a1.json | 25 + .../metrics/sglang_c16_i4096_o4096_a1.json | 25 + .../metrics/sglang_c16_i4096_o512_a1.json | 25 + .../metrics/sglang_c16_i65536_o1024_a1.json | 25 + .../metrics/sglang_c16_i65536_o128_a1.json | 25 + .../metrics/sglang_c16_i65536_o256_a1.json | 25 + .../metrics/sglang_c16_i65536_o512_a1.json | 25 + .../metrics/sglang_c1_i131072_o128_a1.json | 25 + .../metrics/sglang_c1_i16384_o1024_a1.json | 25 + .../metrics/sglang_c1_i16384_o128_a1.json | 25 + .../metrics/sglang_c1_i16384_o2048_a1.json | 25 + .../metrics/sglang_c1_i16384_o256_a1.json | 25 + .../metrics/sglang_c1_i16384_o512_a1.json | 25 + .../metrics/sglang_c1_i65536_o1024_a1.json | 25 + .../metrics/sglang_c1_i65536_o128_a1.json | 25 + .../metrics/sglang_c1_i65536_o256_a1.json | 25 + .../metrics/sglang_c1_i65536_o512_a1.json | 25 + .../metrics/sglang_c32_i1024_o1024_a1.json | 25 + .../metrics/sglang_c32_i1024_o2048_a1.json | 25 + .../metrics/sglang_c32_i1024_o256_a1.json | 25 + .../metrics/sglang_c32_i1024_o512_a1.json | 25 + .../metrics/sglang_c48_i1024_o1024_a1.json | 25 + .../metrics/sglang_c48_i1024_o2048_a1.json | 25 + .../metrics/sglang_c48_i1024_o256_a1.json | 25 + .../metrics/sglang_c48_i1024_o512_a1.json | 25 + .../metrics/sglang_c8_i1024_o4096_a1.json | 25 + .../metrics/sglang_c8_i131072_o128_a1.json | 25 + .../metrics/sglang_c8_i16384_o1024_a1.json | 25 + .../metrics/sglang_c8_i16384_o128_a1.json | 25 + .../metrics/sglang_c8_i16384_o2048_a1.json | 25 + .../metrics/sglang_c8_i16384_o256_a1.json | 25 + .../metrics/sglang_c8_i16384_o512_a1.json | 25 + .../metrics/sglang_c8_i4096_o1024_a1.json | 25 + .../metrics/sglang_c8_i4096_o128_a1.json | 25 + .../metrics/sglang_c8_i4096_o2048_a1.json | 25 + .../metrics/sglang_c8_i4096_o256_a1.json | 25 + .../metrics/sglang_c8_i4096_o4096_a1.json | 25 + .../metrics/sglang_c8_i4096_o512_a1.json | 25 + .../metrics/sglang_c8_i65536_o1024_a1.json | 25 + .../metrics/sglang_c8_i65536_o128_a1.json | 25 + .../metrics/sglang_c8_i65536_o256_a1.json | 25 + .../metrics/sglang_c8_i65536_o512_a1.json | 25 + .../metrics/sglang_c16_i1024_o1024_a1.json | 25 + .../metrics/sglang_c16_i1024_o128_a1.json | 25 + .../metrics/sglang_c16_i1024_o2048_a1.json | 25 + .../metrics/sglang_c16_i1024_o256_a1.json | 25 + .../metrics/sglang_c16_i1024_o4096_a1.json | 25 + .../metrics/sglang_c16_i1024_o512_a1.json | 25 + .../metrics/sglang_c16_i16384_o1024_a1.json | 25 + .../metrics/sglang_c16_i16384_o128_a1.json | 25 + .../metrics/sglang_c16_i16384_o2048_a1.json | 25 + .../metrics/sglang_c16_i16384_o256_a1.json | 25 + .../metrics/sglang_c16_i16384_o512_a1.json | 25 + .../metrics/sglang_c16_i4096_o1024_a1.json | 25 + .../metrics/sglang_c16_i4096_o128_a1.json | 25 + .../metrics/sglang_c16_i4096_o2048_a1.json | 25 + .../metrics/sglang_c16_i4096_o256_a1.json | 25 + .../metrics/sglang_c16_i4096_o4096_a1.json | 25 + .../metrics/sglang_c16_i4096_o512_a1.json | 25 + .../metrics/sglang_c16_i65536_o1024_a1.json | 25 + .../metrics/sglang_c16_i65536_o128_a1.json | 25 + .../metrics/sglang_c16_i65536_o256_a1.json | 25 + .../metrics/sglang_c16_i65536_o512_a1.json | 25 + .../metrics/sglang_c1_i16384_o1024_a1.json | 25 + .../metrics/sglang_c1_i16384_o128_a1.json | 25 + .../metrics/sglang_c1_i16384_o2048_a1.json | 25 + .../metrics/sglang_c1_i16384_o256_a1.json | 25 + .../metrics/sglang_c1_i16384_o512_a1.json | 25 + .../metrics/sglang_c1_i4096_o1024_a1.json | 25 + .../metrics/sglang_c1_i4096_o128_a1.json | 25 + .../metrics/sglang_c1_i4096_o2048_a1.json | 25 + .../metrics/sglang_c1_i4096_o256_a1.json | 25 + .../metrics/sglang_c1_i4096_o4096_a1.json | 25 + .../metrics/sglang_c1_i4096_o512_a1.json | 25 + .../metrics/sglang_c1_i65536_o1024_a1.json | 25 + .../metrics/sglang_c1_i65536_o128_a1.json | 25 + .../metrics/sglang_c1_i65536_o256_a1.json | 25 + .../metrics/sglang_c1_i65536_o512_a1.json | 25 + .../metrics/sglang_c32_i1024_o1024_a1.json | 25 + .../metrics/sglang_c32_i1024_o2048_a1.json | 25 + .../metrics/sglang_c32_i1024_o256_a1.json | 25 + .../metrics/sglang_c32_i1024_o4096_a1.json | 25 + .../metrics/sglang_c32_i1024_o512_a1.json | 25 + .../metrics/sglang_c48_i1024_o512_a1.json | 25 + .../metrics/sglang_c8_i1024_o128_a1.json | 25 + .../metrics/sglang_c8_i16384_o1024_a1.json | 25 + .../metrics/sglang_c8_i16384_o128_a1.json | 25 + .../metrics/sglang_c8_i16384_o2048_a1.json | 25 + .../metrics/sglang_c8_i16384_o256_a1.json | 25 + .../metrics/sglang_c8_i16384_o512_a1.json | 25 + .../metrics/sglang_c8_i4096_o1024_a1.json | 25 + .../metrics/sglang_c8_i4096_o128_a1.json | 25 + .../metrics/sglang_c8_i4096_o2048_a1.json | 25 + .../metrics/sglang_c8_i4096_o256_a1.json | 25 + .../metrics/sglang_c8_i4096_o4096_a1.json | 25 + .../metrics/sglang_c8_i4096_o512_a1.json | 25 + .../metrics/sglang_c8_i65536_o1024_a1.json | 25 + .../metrics/sglang_c8_i65536_o128_a1.json | 25 + .../metrics/sglang_c8_i65536_o256_a1.json | 25 + .../metrics/sglang_c8_i65536_o512_a1.json | 25 + .../tp8_dp1/server_cmd.txt | 1 + .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- .../run_bench.sh | 6 +- .../start_sglang_docker.sh | 4 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- .../run_bench.sh | 6 +- .../run_adaptive_concurrency.sh | 6 +- .../run_adaptive_concurrency_add16.sh | 6 +- platforms/README.md | 2 + .../adaptive_bench_lib.sh.bak.20260716-084832 | 778 ------------------ scripts/common/dp_proxy.py | 173 ---- 174 files changed, 3604 insertions(+), 1531 deletions(-) create mode 100644 docs/NEW_PLATFORM_GUIDE.md delete mode 100755 experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh.bak.20260716-084832 create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.jsonl create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.md create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i131072_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i131072_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i1024_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i131072_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c48_i1024_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i1024_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o2048_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o4096_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o1024_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o128_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o256_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o512_a1.json create mode 100644 experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/server_cmd.txt delete mode 100755 scripts/common/adaptive_bench_lib.sh.bak.20260716-084832 delete mode 100644 scripts/common/dp_proxy.py diff --git a/.gitignore b/.gitignore index 7ef7c28..3bcf22e 100644 --- a/.gitignore +++ b/.gitignore @@ -5,6 +5,10 @@ envs/modelscope-upload/ envs/sglang/ envs/sglang-src/ envs/vllm/ +envs/charts/ +# 其他一次性工具环境/临时产物 +.venv-charts/ +.tmp_charts/ # 其他通用忽略 deps/ tmp/ diff --git a/BENCHMARK_WORKFLOW.md b/BENCHMARK_WORKFLOW.md index ee298cc..c1e8df2 100644 --- a/BENCHMARK_WORKFLOW.md +++ b/BENCHMARK_WORKFLOW.md @@ -3,49 +3,51 @@ ## Directory Layout ``` -/data/user1/yy/ -├── platforms/ # chip/accelerator platform configs +/ # 本仓库根目录(如 /data1/yy/sskj,按机器调整) +├── platforms/ # chip/accelerator platform configs │ ├── kunlun_p800.env +│ ├── nvidia_h20.env │ ├── nvidia_h200.env -│ └── patches/kunlun_p800/ # runtime patches required by some images -├── scripts/ # shared benchmark/orchestrator/utility scripts -│ ├── common/ # reusable components (lib.sh, platform.sh, parse_backend.py, ...) -│ ├── benchmark_dspark_0707/ # legacy DSpark benchmark suite (仍被 experiments/dsv4_h200_dspark/ 引用) -│ └── SLO_STANDARDS.md # 推理服务 SLO 标准 -├── experiments/ # experiment-centric directories (preferred) -│ └── dsv4_p800_sglang/ -│ ├── README.md -│ ├── config.env # experiment-level configuration -│ ├── start_server.sh -│ ├── run_bench.sh -│ ├── parse_results.py -│ └── results/ -│ └── 20260708-XXXXXX/ -│ ├── report.md -│ ├── results.json -│ └── logs/ -├── bench_results/ # legacy benchmark outputs (read-only history) -│ ├── dspark_grid_20260707-132641/ -│ └── ... -├── logs/ # server logs (stdout/stderr from start scripts) -├── datasets/ # benchmark datasets -└── envs/ # Python virtual environments +│ ├── nvidia_rtx6000d.env +│ └── patches/kunlun_p800/ # runtime patches required by some images +├── scripts/ +│ └── common/ # reusable components (lib.sh, platform.sh, parse_backend.py, adaptive_bench_lib.sh, ...) +├── docs/ +│ └── SLO_STANDARDS.md # 推理服务 SLO 标准 +├── experiments/ # experiment-centric directories (preferred): experiments/// +│ ├── TEMPLATE/ # template for fixed-scenario sglang-vs-vllm experiments +│ └── / # h20 / h200 / p800 / pro6000 +│ └── dsv4_p800_sglang/ +│ ├── README.md +│ ├── config.env # experiment-level configuration +│ ├── start_server.sh +│ ├── run_bench.sh +│ ├── parse_results.py +│ └── results/ +│ └── 20260708-XXXXXX/ +│ ├── report.md +│ ├── results.json +│ └── logs/ +├── datasets/ # benchmark datasets +└── envs/ # Python virtual environments ``` +注:legacy 的 `scripts/benchmark_dspark_0707/` 已删除;顶层 `bench_results/` 与 `logs/` 已不在仓库中(历史结果已归档迁移至仓库外)。 + ## Rules 1. **Experiments are the primary organization unit.** - - Each experiment lives under `experiments//` and contains its scripts, configuration, and results. + - Each experiment lives under `experiments///` (platform ∈ {h20, h200, p800, pro6000}) and contains its scripts, configuration, and results. - Shared orchestration code lives in `scripts/common/`; do not copy server start / health check logic into every experiment. - - 新实验必须放在 `experiments//` 下;`scripts/` 下仅保留 `scripts/common/` 和被现有实验引用的 legacy 脚本。 + - 新实验必须放在 `experiments///` 下;`scripts/` 下仅保留 `scripts/common/`(原 legacy 套件 `scripts/benchmark_dspark_0707/` 已删除)。 2. **Benchmark outputs live with their experiment.** - - For `experiments//`, results go in `experiments//results//`. + - For `experiments///`, results go in `experiments///results//`. - Each run directory must contain `report.md` (human-readable) and `results.json` (structured data). - Raw outputs go in `raw_outputs/`; logs go in `logs/`. - - Legacy `bench_results/_/` directories remain valid for archived runs. + - Legacy `bench_results/_/` 目录已从仓库移除,历史结果已归档迁移至仓库外。 -3. **Each `bench_results//` directory must contain two final artifacts.** +3. **Each experiment run directory (`results//`) must contain two final artifacts.** - A Markdown report for human reading (e.g. `report.md`, `comparison_report.md`). - A JSON file with the complete structured result data for programmatic analysis (e.g. `results.json`). - The directory may also contain a `README.md` documenting provenance if the report alone does not cover it. @@ -59,8 +61,7 @@ - See [Final JSON Schema](#final-json-schema) below for the recommended structure. 5. **Scripts should default `RESULT_ROOT` to the experiment's results directory.** - - For `experiments//run_bench.sh`, default to `experiments//results/${RUN_ID}/`. - - For legacy scripts, default to `bench_results/_${RUN_ID}/`. + - For `experiments///run_bench.sh`, default to `experiments///results/${RUN_ID}/`. - Allow override via `RESULT_ROOT` env var. - Use `RUN_ID=$(date '+%Y%m%d-%H%M%S')` unless specified. @@ -96,30 +97,23 @@ For experiment-centric layout: ``` -experiments//results// +experiments///results// ``` Examples: -- `experiments/dsv4_p800_sglang/results/20260708-120000/` -- `experiments/dspark_grid/results/20260707-132641/` +- `experiments/p800/dsv4_p800_sglang/results/20260708-120000/` +- `experiments/h200/dsv4_h200_sglang_vs_vllm/results/20260707-132641/` The `results.json` metadata already records `chip`/`accelerator` and `engine`, so the directory path does not need to encode them. If a single experiment must distinguish across platforms in its directory tree, use: ``` -experiments//results/__/ +experiments///results/__/ ``` -### Legacy result directories +### Legacy result directories (removed) -``` -bench_results/_/ -``` - -Examples: - -- `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/`) +`bench_results/_/` 与 `experiments/legacy_bench_results/` 均已从仓库移除,历史结果已归档迁移至仓库外;保留此节仅说明历史命名格式。 ### Raw output files @@ -146,14 +140,14 @@ logs/_orchestrator_YYYYMMDD_HHMMSS.log ## Final JSON Schema -The JSON file inside each `bench_results//` directory should follow a stable schema so that downstream Python scripts can load every experiment the same way. The file is usually named `results.json`. +The JSON file inside each experiment run directory (`experiments///results//`) should follow a stable schema so that downstream Python scripts can load every experiment the same way. The file is usually named `results.json`. ### Required top-level fields ```json { "metadata": { - "experiment": "dspark_grid", + "experiment": "dsv4_h200_dspark", "run_id": "20260707-132641", "timestamp": "2026-07-07T13:26:41+08:00", "model": "/data/models/DeepSeek-V4-Flash-DSpark", @@ -162,8 +156,8 @@ The JSON file inside each `bench_results//` directory should follow a stabl "hardware": "8x NVIDIA H200 143GB", "accelerator": "NVIDIA H200", "chip": "NVIDIA H200", - "script": "scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh", - "env": "/data/user1/yy/envs/vllm-dspark", + "script": "experiments/h200/dsv4_h200_dspark/run_bench.sh", + "env": "/path/to/envs/vllm-dspark", "git_commit": "optional git sha", "description": "optional free-text note" }, @@ -241,37 +235,27 @@ The JSON file inside each `bench_results//` directory should follow a stabl ### Run P800 SGLang benchmark (Kunlun P800, Docker) ```bash -bash experiments/dsv4_p800_sglang/run_bench.sh +bash experiments/p800/dsv4_p800_sglang/run_bench.sh ``` ### Run H200 DSpark benchmark (NVIDIA H200, native venv) ```bash -bash experiments/dsv4_h200_dspark/run_bench.sh +bash experiments/h200/dsv4_h200_dspark/run_bench.sh ``` -### Run legacy DSpark grid benchmark +Note: the wrapper's default `LEGACY_GRID_SCRIPT` / `SERVER_START_SCRIPT` point at the removed legacy scripts; override them before running (see `docs/H200_QUICKSTART.md`). -```bash -bash scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh -``` +### Legacy DSpark scripts (removed) -### Run DSpark spec-tokens comparison - -```bash -bash scripts/benchmark_dspark_0707/run_dspark_st_comparison.sh -``` +The legacy suite `scripts/benchmark_dspark_0707/` (grid benchmark, spec-tokens comparison, etc.) has been removed from this repo. ### Parse results ```bash # Experiment-centric layout -python3 experiments/dsv4_p800_sglang/parse_results.py \ - experiments/dsv4_p800_sglang/results/ - -# Legacy layout -/data/user1/yy/envs/sglang/bin/python scripts/benchmark_dspark_0707/parse_results.py \ - /data/user1/yy/experiments/legacy_bench_results/dspark_grid_ +python3 experiments/p800/dsv4_p800_sglang/parse_results.py \ + experiments/p800/dsv4_p800_sglang/results/ ``` ## Adding a New Platform or Experiment @@ -297,7 +281,7 @@ For host-native platforms, set the relevant venv paths (see ### 2. Create an experiment directory ``` -experiments// +experiments/// ├── README.md # Purpose and usage ├── config.env # Model, port, scenarios, engine overrides ├── start_server.sh # (optional) platform-specific server launch @@ -309,7 +293,7 @@ At minimum, `run_bench.sh` should: 1. Source `scripts/common/lib.sh` and `scripts/common/platform.sh`. 2. Read `config.env`. -3. Create `experiments//results//`. +3. Create `experiments///results//`. 4. Call `write_metadata_json` to create `results.json`. 5. Record the exact server launch command/args in `results.json` `config.server_args` (or `phaseN_server_args`). 6. Run the benchmark scenarios. @@ -327,16 +311,15 @@ Docker/XPU 平台可以在 `scripts/common/` 下新增自己的 helper,但不 ### 4. Example experiments -- Docker / XPU: `experiments/dsv4_p800_sglang/` -- Native / H200: `experiments/dsv4_h200_dspark/` +- Docker / XPU: `experiments/p800/dsv4_p800_sglang/` +- Native / H200: `experiments/h200/dsv4_h200_dspark/` See `docs/H200_QUICKSTART.md` for a concrete H200 porting walkthrough. ## Checklist Before Committing / Archiving - [ ] No `.jsonl`, `.json`, `.log`, or `.md` files left in the project root. -- [ ] For `experiments//`, outputs live in `experiments//results//`. -- [ ] For legacy runs, outputs live in `bench_results/_/`. +- [ ] For `experiments///`, outputs live in `experiments///results//`. - [ ] `/report.md` (or equivalent human-readable `.md`) exists. - [ ] `/results.json` exists and follows the [Final JSON Schema](#final-json-schema). - [ ] `/results.json` metadata records the `chip`/`accelerator` and `engine`/`backend` used. @@ -344,6 +327,6 @@ See `docs/H200_QUICKSTART.md` for a concrete H200 porting walkthrough. - [ ] `/results.json` each scenario records `slo_status` (overall pass/fail/partial against the relevant SLO). - [ ] Raw output filenames include the chip/accelerator and engine when cross-platform runs may collide. - [ ] `/README.md` exists and documents provenance (or the report itself covers provenance). -- [ ] Scripts either live under `experiments//` or in `scripts/` (or `scripts//`). +- [ ] Scripts either live under `experiments///` or in `scripts/common/`. - [ ] Script path references updated after moving. - [ ] Commits include the experiment code plus final `results.json` and `.md` reports; logs/raw outputs are left ignored. diff --git a/README.md b/README.md index 132b23b..f6e3977 100644 --- a/README.md +++ b/README.md @@ -1,190 +1,124 @@ -# DSV4 / DeepSeek-V4-Flash 推理测速项目 +# sskj — 多平台大模型推理性能基准测试项目 -> 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`](docs/H200_QUICKSTART.md)。 +> **项目目的**:当新显卡(GPU/NPU)到货时,用最短时间完成大模型在该平台上的推理性能评估与部署配置选型。 +> 当前模型:DeepSeek-V4-Flash(FP8 / INT8);后续接入 GLM5.2,**完全复用**本项目的实验与报告流程。 +> 新平台接入 SOP:[`docs/NEW_PLATFORM_GUIDE.md`](docs/NEW_PLATFORM_GUIDE.md)。 -## 目录说明 +## 这个项目解决什么问题 -| 目录/文件 | 说明 | +1. **新卡快速评估**:新平台到货后,按 SOP 复制实验目录、改少量配置,即可跑起与既有平台口径一致的 benchmark,横向可比。 +2. **部署配置选型**:通过 TP×DP 并行矩阵 + 自适应并发搜索,回答"该平台上哪种并行配置在什么并发下满足 SLO、吞吐最高"。 +3. **数据资产化**:每轮实验产出结构化结果(jsonl/csv/manifest),汇总成飞书性能报告并回填多维表格,形成可检索的历史数据库。 + +## 标准工作流(全流程) + +``` +① 跑实验 experiments///run_adaptive_concurrency*.sh + └─ 公共搜索逻辑: scripts/common/adaptive_bench_lib.sh +② 落盘 adaptive_results// + └─ adaptive_points.jsonl / adaptive_shapes.jsonl / adaptive_summary.* / run_manifest.json +③ 分析 用 Python 从 jsonl 实算(禁止目测),SLO 口径见 docs/SLO_STANDARDS.md +④ 飞书报告 按飞书 wiki「性能报告编写指南(README)」写到「显卡性能报告/<平台>」节点下 +⑤ 回填多维表格 飞书「模型推理适配 Bench 迭代跟踪」:每成功探测点一行 +``` + +- SLO 口径(S2 层,DSV4-Flash / GLM5.2 均适用):**TTFT P95 < 3s,TPOT P95 < 50ms**,严格小于。详见 [`docs/SLO_STANDARDS.md`](docs/SLO_STANDARDS.md)。 +- 报告编写规范(章节骨架、分析口径、发布流程)以飞书 wiki「显卡性能报告 / 性能报告编写指南(README)」为准。 +- 多维表格每行 = 一个成功探测点(config、Concurrency、Throughput、Input/Output TPS、TTFT/TPOT/E2E P95、SLO达标状态、开发文档链接)。 + +## 目录结构 + +``` +platforms/ # 平台配置(每平台一个 .env)+ 平台补丁(patches/) +scripts/common/ # 跨实验复用的编排组件(见下) +experiments/ # 实验目录(三层结构) +├── TEMPLATE/ # 老式固定场景对比实验模板(sglang vs vllm) +├── h20/ # NVIDIA H20 96GB +├── h200/ # NVIDIA H200 143GB +├── p800/ # Kunlun P800 XPU +└── pro6000/ # NVIDIA RTX 6000D +docs/ # 规范与指南(SLO、实验规范、新卡接入) +envs/ # 环境搭建文档(具体 venv 不提交,见 .gitignore) +datasets/ # benchmark 数据集(不提交) +BENCHMARK_WORKFLOW.md # 结果目录与命名规范、results.json schema +``` + +### scripts/common/ 组件 + +| 文件 | 职责 | |---|---| -| `platforms/` | 芯片/加速器平台配置(`*.env`),如 P800、H200、RTX 6000D | -| `scripts/common/` | 跨实验复用的 orchestration 组件(server 启停、health check、元数据生成、结果解析) | -| `scripts/` | 仅保留仍在使用的 legacy DSpark 脚本与 SLO 标准 | -| `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 | -| `docs/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。 +| `lib.sh` | 日志、health check、metadata JSON 生成 | +| `platform.sh` | 平台自动探测并加载 `platforms/.env` | +| `adaptive_bench_lib.sh` | **自适应并发搜索核心库**(引擎无关,靠 5 个 `engine_*` 回调注入平台逻辑) | +| `adaptive_concurrency.py` | 上者的 Python 助手(shapes 生成、结果解析、汇总) | +| `parse_backend.py` | 固定并发实验:raw jsonl → results.json + report.md | +| `compare.py` | SGLang vs vLLM 横向对比表(老式实验用) | +| `warmup.py` | 服务预热 | +| `server_docker.sh` / `bench_client_docker.sh` | P800 专用 Docker 服务端/客户端生命周期 | +| `adaptive_heartbeat.sh` | 长跑实验的心跳监控(手动独立运行) | ## 实验索引 -### 新结构(experiments/) +### TP/DP matrix + 自适应并发(当前主流形态) -| 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 | -| DSV4 H200 max context length | `experiments/dsv4_h200_max_context_length/run_bench.sh` | 探索 SGLang / vLLM 在 H200 上能支持的最大输入长度 | -| DSV4 H200 SGLang TP/DP Matrix | `experiments/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh` | H200 上 SGLang 不同 TP×DP 配置的固定并发矩阵测试 | -| DSV4 H200 SGLang TP/DP Adaptive | `experiments/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh` | H200 上 SGLang 不同 TP×DP 配置的自适应并发饱和点搜索 | -| DSV4 H200 vLLM TP/DP Matrix | `experiments/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh` | H200 上 vLLM 不同 TP×DP 配置的固定并发矩阵测试 | -| DSV4 H200 vLLM TP/DP Adaptive | `experiments/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh` | H200 上 vLLM 不同 TP×DP 配置的自适应并发饱和点搜索 | -| DSV4 H200 64k SGLang vs vLLM | `experiments/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh` | H200 上 64k 上下文 SGLang 与 vLLM 对比 | -| DSV4 H200 256k 4k Probe | `experiments/dsv4_h200_256k_4k_probe/run_bench.sh` | H200 上 256k 输入/4k 输出长上下文探测 | -| DSV4 H200 vLLM TP2 Custom Bench | `experiments/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh` | H200 上 vLLM TP=2 自定义压测客户端(多服务负载均衡) | -| DSV4 H200 vLLM TP4 Custom Bench | `experiments/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh` | H200 上 vLLM TP=4 自定义压测客户端 | -| DSV4 H200 vLLM TP8 Custom Bench | `experiments/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh` | H200 上 vLLM TP=8 自定义压测客户端 | -| DSV4 H200 vLLM MTP vs Default | `experiments/dsv4_h200_vllm_mtp_vs_default/run_bench.sh` | H200 上 vLLM MTP(Multi-Token Prediction)vs 默认配置对比 | -| DSV4 H200 Long Context Matrix | `experiments/dsv4_h200_long_context_matrix/run_bench.sh` | H200 上长上下文矩阵测试(多后端对比) | -| **DSV4 RTX 6000D SGLang TP/DP Adaptive** | `experiments/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh` | **RTX 6000D 上 SGLang 自适应并发饱和点搜索** | -| **DSV4 RTX 6000D vLLM TP/DP Adaptive** | `experiments/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh` | **RTX 6000D 上 vLLM 自适应并发饱和点搜索** | +| 实验 | 说明 | +|---|---| +| `experiments/h20/dsv4_h20_vllm_tp_dp_matrix/` | H20 + vLLM,TP2/DP4、TP4/DP2、TP8/DP1 | +| `experiments/h20/dsv4_h20_sglang_tp_dp_matrix/` | H20 + SGLang,同上 | +| `experiments/h200/dsv4_h200_vllm_tp_dp_matrix/` | H200 + vLLM | +| `experiments/h200/dsv4_h200_sglang_tp_dp_matrix/` | H200 + SGLang | +| `experiments/p800/dsv4_p800_sglang_tp_dp_matrix/` | P800 + SGLang(INT8;TP2/DP4 启动 OOM 无数据,见 config.env 注释) | +| `experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/` | RTX 6000D + vLLM | +| `experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/` | RTX 6000D + SGLang | -### 旧结构(scripts/ + bench_results/) +每个目录内:`run_bench.sh` 跑固定并发矩阵;`run_adaptive_concurrency.sh` 从 C=1 指数倍增搜饱和点;`run_adaptive_concurrency_add16.sh` 从 C=16 线性 +16 步进、带 TTFT SLO 停止与回退(当前主力用法,见 `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`)。 -仅保留还在被 `experiments/dsv4_h200_dspark/` 引用的 legacy DSpark 脚本,其余旧脚本已删除。 +### 其他实验(H200 老形态 & 专项) -| 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` | +| 实验 | 说明 | +|---|---| +| `experiments/h200/dsv4_h200_sglang/` `…/dsv4_h200_vllm/` | H200 单引擎 baseline | +| `experiments/h200/dsv4_h200_sglang_vs_vllm/` | H200 引擎控制变量对比 | +| `experiments/h200/dsv4_h200_dspark/` `…/dsv4_h200_vllm_dspark_vs_default/` | DSpark 投机解码相关 | +| `experiments/h200/dsv4_h200_vllm_mtp_vs_default/` | vLLM MTP 对比 | +| `experiments/h200/dsv4_h200_vllm_tp{2,4,8}_custom_bench/` | 自定义压测客户端(多服务负载均衡) | +| `experiments/h200/dsv4_h200_max_context_length/` `…/dsv4_h200_long_context_matrix/` `…/dsv4_h200_256k_4k_probe/` `…/dsv4_h200_64k_sglang_vs_vllm/` | 长上下文专项 | +| `experiments/p800/dsv4_p800_sglang/` `…/dsv4_p800_max_context_length/` `…/dsv4_p800_long_context_matrix/` `…/dsv4_p800_256k_4k_probe/` | P800 baseline 与长上下文专项 | +| `experiments/TEMPLATE/` | 老式固定场景实验模板 | ## 快速复现 -### P800 SGLang +以 H20 vLLM 自适应并发搜索为例(其他平台同理,换目录即可): ```bash -bash experiments/dsv4_p800_sglang/run_bench.sh +# 1. dry-run:只打印搜索计划,不加载模型 +cd experiments/h20/dsv4_h20_vllm_tp_dp_matrix +DRY_RUN=1 bash run_adaptive_concurrency_add16.sh + +# 2. 单组合冒烟:只测 TP=8、1K/128,并发上限 8 +RUN_ID=smoke-$(date +%Y%m%d-%H%M%S) \ +TP_LIST="8" ISL_LIST="1024" OSL_LIST="128" GRID_LIMIT=1 SEARCH_MAX_CONCURRENCY=8 \ +bash run_adaptive_concurrency_add16.sh + +# 3. 正式跑(放 tmux) +tmux new-session -d -s h20-vllm-adaptive \ + "cd $(pwd) && bash run_adaptive_concurrency_add16.sh" ``` -### H200 DSpark +更多用法(断点续跑 RESUME_RUN_ID、常用覆盖参数)见 `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`。 -```bash -bash experiments/dsv4_h200_dspark/run_bench.sh -``` +## 新平台 / 新模型接入 -### H200 SGLang baseline - -```bash -bash experiments/dsv4_h200_sglang/run_bench.sh -``` - -### H200 SGLang vs vLLM 对比 - -```bash -bash experiments/dsv4_h200_sglang_vs_vllm/run_bench.sh -``` - -### H200 vLLM DSpark vs default - -```bash -bash experiments/dsv4_h200_vllm_dspark_vs_default/run_bench.sh -``` - -### H200 SGLang TP/DP Matrix - -```bash -bash experiments/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh -``` - -### H200 SGLang TP/DP 自适应并发搜索 - -```bash -bash experiments/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh -``` - -### H200 vLLM TP/DP Matrix - -```bash -bash experiments/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh -``` - -### H200 vLLM TP/DP 自适应并发搜索 - -```bash -bash experiments/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh -``` - -### H200 64k SGLang vs vLLM - -```bash -bash experiments/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh -``` - -### H200 256k 4k Probe - -```bash -bash experiments/dsv4_h200_256k_4k_probe/run_bench.sh -``` - -### H200 vLLM TP2/TP4/TP8 Custom Bench - -```bash -bash experiments/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh -bash experiments/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh -bash experiments/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh -``` - -### H200 vLLM MTP vs Default - -```bash -bash experiments/dsv4_h200_vllm_mtp_vs_default/run_bench.sh -``` - -### H200 Long Context Matrix - -```bash -bash experiments/dsv4_h200_long_context_matrix/run_bench.sh -``` - -### RTX 6000D SGLang TP/DP 自适应并发搜索 - -```bash -bash experiments/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh -``` - -### RTX 6000D vLLM TP/DP 自适应并发搜索 - -```bash -bash experiments/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh -``` - -### DSpark grid / spec-tokens(旧结构) - -```bash -bash scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh -bash scripts/benchmark_dspark_0707/run_dspark_st_comparison.sh -``` - -### 解析已有结果 - -```bash -# 新结构 -python3 experiments/dsv4_h200_dspark/parse_results.py \ - experiments/dsv4_h200_dspark/results/ - -# 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 `docs/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` +- **新显卡**:按 [`docs/NEW_PLATFORM_GUIDE.md`](docs/NEW_PLATFORM_GUIDE.md) 执行,核心动作是复制 `experiments/h20/dsv4_h20__tp_dp_matrix`、改 `config.env` / `adaptive_config.env` / `matrix.json` 三件套,再加 `platforms/.env`。 +- **新模型(GLM5.2)**:复用同一实验目录结构,新增 `experiments//glm52___tp_dp_matrix/`,改模型路径、量化与引擎启动参数即可;SLO 口径、报告流程、多维表格结构不变。 ## 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`。 -- **NVIDIA RTX 6000D**:Python env `/root/.miniconda3/envs/sglang`(压测客户端);模型 `/data/hf_models/DeepSeek-V4-Flash`;Docker 镜像 `lmsysorg/sglang:latest`、`vllm/vllm-openai:latest`。 -- **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。 +- 公共:bash、Python 3、`sglang.bench_serving` 压测客户端(Docker 或 venv);绘图分析用 `envs/charts/`(matplotlib)。 +- 各平台的镜像 / venv / 模型路径见对应实验目录的 `config.env` 与 `platforms/.env`;**模型与数据集路径是机器相关的**,换机器时按实际路径调整。 +- 环境搭建规范见 `envs/README.md` 与 `envs/UV_ENV_SETUP.md`。 -平台相关常量见 `platforms/*.env`。 +## 注意事项 + +- 提交内容:实验代码 + 最终产物(results.json / report.md / adaptive 的 jsonl 与 summary);日志、raw_outputs、gpu_logs 不入库(`.gitignore` 已排除)。 +- 仓库会clone到多台机器,实验脚本内引用公共组件一律用 `${SCRIPT_DIR}/../../../scripts/common`(TEMPLATE 为 `../../`),不要写绝对路径。 diff --git a/docs/EXPERIMENT_GUIDE.md b/docs/EXPERIMENT_GUIDE.md index a772634..2a6b2bd 100644 --- a/docs/EXPERIMENT_GUIDE.md +++ b/docs/EXPERIMENT_GUIDE.md @@ -4,8 +4,10 @@ ## 1. 目录结构 +实验目录为三层结构:`experiments///`,其中 platform ∈ {h20, h200, p800, pro6000};另有 `experiments/TEMPLATE/` 作为新实验模板。 + ``` -experiments// +experiments/// ├── config.env # 实验参数唯一来源 ├── run_bench.sh # 编排入口 ├── start_sglang.sh # SGLang server 启动脚本 @@ -24,22 +26,27 @@ experiments// └── logs/ # .gitignore 忽略 ``` -通用工具集中放在 `scripts/common/`,不要复制到每个实验: +上述是 TEMPLATE 式的固定场景对比实验布局(sglang vs vllm,老形态,仍可用)。当前主流实验形态是 TP/DP matrix + 自适应并发搜索:实验目录内由 `matrix.json` + `generate_scenarios.py` 生成场景,`adaptive_config.env` 配置搜索参数,`run_adaptive_concurrency.sh` 驱动,结果落盘 `adaptive_results//`(`adaptive_points.jsonl`、`adaptive_summary.{jsonl,md,csv}`、`run_manifest.json`、`tp*_dp*/` 等),用法见 `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`。 + +通用工具集中放在 `scripts/common/`,不要复制到每个实验。三层实验目录以 `../../../scripts/common` 引用(`experiments/TEMPLATE/` 为两层,用 `../../scripts/common`): - `scripts/common/lib.sh`:日志、metadata、JSON 工具 - `scripts/common/platform.sh`:平台自动检测 - `scripts/common/warmup.py`:server 预热 - `scripts/common/parse_backend.py`:解析 raw jsonl -> results.json + report.md - `scripts/common/compare.py`:生成 SGLang vs vLLM 对比表 +- `scripts/common/adaptive_bench_lib.sh` / `adaptive_concurrency.py`:自适应并发搜索 +- `scripts/common/server_docker.sh` / `bench_client_docker.sh`:Docker server / client 管理 +- `scripts/common/adaptive_heartbeat.sh`:自适应测试心跳保活 ## 2. 新增一个实验 最快方式: ```bash -cp -r experiments/TEMPLATE experiments/ +cp -r experiments/TEMPLATE experiments// # 修改 config.env、start_*.sh -bash experiments//run_bench.sh +bash experiments///run_bench.sh ``` ### 2.1 config.env 必备字段 @@ -52,8 +59,9 @@ MODEL_PATH="/data/models/DeepSeek-V4-Flash" SGLANG_PORT="${SGLANG_PORT:-30006}" VLLM_PORT="${VLLM_PORT:-30005}" -VENV_SGLANG="${VENV_SGLANG:-/data/user1/yy/envs/sglang}" -VENV_VLLM="${VENV_VLLM:-/data/user1/yy/envs/vllm}" +# venv 路径按机器实际配置调整 +VENV_SGLANG="${VENV_SGLANG:-/path/to/envs/sglang}" +VENV_VLLM="${VENV_VLLM:-/path/to/envs/vllm}" export CUDA_VISIBLE_DEVICES="${CUDA_VISIBLE_DEVICES:-0,1,2,3,4,5,6,7}" TP="${TP:-8}" @@ -89,9 +97,11 @@ VLLM_START_SCRIPT="${SCRIPT_DIR:-.}/start_vllm.sh" ### 3.1 芯片级默认配置 -芯片相关默认值放到 `platforms/.env`,例如: +芯片相关默认值放到 `platforms/.env`,现有: +- `platforms/nvidia_h20.env` - `platforms/nvidia_h200.env` +- `platforms/nvidia_rtx6000d.env` - `platforms/kunlun_p800.env` 实验级覆盖通过 `config.env` 实现。 @@ -110,7 +120,7 @@ VLLM_START_SCRIPT="${SCRIPT_DIR:-.}/start_vllm.sh" ## 4. SLO 标准 -默认使用 S2 层级: +默认使用 S2 层级(完整分级定义见 `docs/SLO_STANDARDS.md`): - TTFT P95 < 3000 ms - TPOT mean < 50 ms @@ -142,14 +152,14 @@ VLLM_START_SCRIPT="${SCRIPT_DIR:-.}/start_vllm.sh" ### 6.3 提交示例 ```bash -git add experiments//config.env \ - experiments//run_bench.sh \ - experiments//start_*.sh \ - experiments//results//comparison.md \ - experiments//results//sglang/results.json \ - experiments//results//sglang/report.md \ - experiments//results//vllm/results.json \ - experiments//results//vllm/report.md +git add experiments///config.env \ + experiments///run_bench.sh \ + experiments///start_*.sh \ + experiments///results//comparison.md \ + experiments///results//sglang/results.json \ + experiments///results//sglang/report.md \ + experiments///results//vllm/results.json \ + experiments///results//vllm/report.md git commit -m "results: run " git push origin main @@ -174,7 +184,7 @@ declare -a SCENARIOS=( ## 8. 命名约定 -- 实验目录:`{model}_{chip}_{scenario}_{backend_vs_backend}`,例如 `dsv4_h200_sglang_vs_vllm` +- 实验目录:`experiments//{model}_{chip}_{scenario}_{backend_vs_backend}`,例如 `experiments/h200/dsv4_h200_sglang_vs_vllm` - scenario 名称:`c{concurrency}_i{input_len}_o{output_len}` - raw 输出:`{backend}_{label}_{MMDD}_{concurrency}_{input_len}_{output_len}.jsonl` - run_id:`YYYYMMDD-HHMMSS` diff --git a/docs/H200_QUICKSTART.md b/docs/H200_QUICKSTART.md index dc83a90..41b89e7 100644 --- a/docs/H200_QUICKSTART.md +++ b/docs/H200_QUICKSTART.md @@ -4,7 +4,7 @@ This guide covers how to use this benchmark repository on an NVIDIA H200 machine ## Current state -The H200 path is currently based on **native host virtual environments** (not Docker): +The older H200 experiments (e.g. `experiments/h200/dsv4_h200_dspark/`) are based on **native host virtual environments** (not Docker): - Server engine: `vllm-dspark` - Server env: `envs/vllm-dspark` @@ -12,13 +12,15 @@ The H200 path is currently based on **native host virtual environments** (not Do - Default model: `/data/models/DeepSeek-V4-Flash-DSpark` - Default port: `30004` -The legacy benchmark scripts live under `scripts/benchmark_dspark_0707/`. A migration wrapper experiment `experiments/dsv4_h200_dspark/` is also provided so the results follow the new `experiments//results//` layout. +Note: the current mainstream H200 experiments — the TP/DP matrix adaptive-concurrency suites `experiments/h200/dsv4_h200_vllm_tp_dp_matrix/` and `experiments/h200/dsv4_h200_sglang_tp_dp_matrix/` — are Docker-based (`USE_DOCKER=1` for the server, benchmark client via `DOCKER_CLIENT_IMAGE`); see `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`. + +The legacy benchmark suite `scripts/benchmark_dspark_0707/` has been removed from this repo. The migration wrapper experiment `experiments/h200/dsv4_h200_dspark/` still exists and produces the `experiments/h200//results//` layout, but its default `LEGACY_GRID_SCRIPT` / `SERVER_START_SCRIPT` point at the removed scripts and must be overridden to run. ## 1. Pull and verify ```bash -git clone /data/user1/yy -cd /data/user1/yy +git clone sskj # clone path: adjust to your machine +cd sskj # Platform should auto-detect as nvidia_h200 source scripts/common/platform.sh @@ -30,36 +32,28 @@ If auto-detection fails, set it explicitly: PLATFORM=nvidia_h200 source scripts/common/platform.sh ``` -## 2. Run the existing DSpark benchmark grid +## 2. Legacy DSpark benchmark grid (removed) -This uses the legacy scripts and writes results to `bench_results/dspark_grid_/`. - -```bash -bash scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh -``` - -To reuse an already-running server: - -```bash -SKIP_MANAGE_SERVER=1 bash scripts/benchmark_dspark_0707/run_dspark_benchmark_grid.sh -``` +The legacy suite `scripts/benchmark_dspark_0707/` has been removed from this repo, along with its `bench_results/dspark_grid_/` output location (historical results were archived elsewhere). Use the wrapper experiment below, or the current tp_dp matrix experiments (`experiments/h200/dsv4_h200_{vllm,sglang}_tp_dp_matrix/`), instead. ## 3. Run the migration wrapper experiment -This produces results in the new experiment-centric layout: +This produces results in the experiment-centric layout: ```bash -bash experiments/dsv4_h200_dspark/run_bench.sh +bash experiments/h200/dsv4_h200_dspark/run_bench.sh ``` -Results land in `experiments/dsv4_h200_dspark/results//`. +Results land in `experiments/h200/dsv4_h200_dspark/results//`. + +Caveat: the wrapper's default `LEGACY_GRID_SCRIPT` and `SERVER_START_SCRIPT` refer to the removed `scripts/benchmark_dspark_0707/` suite and `scripts/start_dsv4_dspark_8card.sh`; set both explicitly (or restore equivalent scripts) before running. ## 4. Create a new H200 experiment To add a new H200 benchmark (for example a different model or engine), create: ``` -experiments// +experiments/h200// ├── README.md # What this experiment measures ├── config.env # Model, port, scenarios, venv paths ├── start_server.sh # (optional) native server launch @@ -77,9 +71,9 @@ PORT="${PORT:-30004}" BACKEND="${BACKEND:-vllm}" ENGINE="${ENGINE:-vllm-dspark}" -# Native venvs -VENV_SERVER="${VENV_SERVER:-/data/user1/yy/envs/vllm-dspark}" -VENV_CLIENT="${VENV_CLIENT:-/data/user1/yy/envs/sglang}" +# Native venvs (paths are per-machine; adjust to yours) +VENV_SERVER="${VENV_SERVER:-/path/to/envs/vllm-dspark}" +VENV_CLIENT="${VENV_CLIENT:-/path/to/envs/sglang}" SCENARIOS=( "32 512 256" @@ -89,11 +83,11 @@ SCENARIOS=( Then source the platform loader and shared helpers at the top of `run_bench.sh`: ```bash -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" ``` -For a concrete example, see `experiments/dsv4_h200_dspark/`. +For a concrete example, see `experiments/h200/dsv4_h200_dspark/`. For the current mainstream form (TP/DP matrix + adaptive concurrency search, Docker-based), copy `experiments/h200/dsv4_h200_vllm_tp_dp_matrix/` instead; usage in `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`. ## 5. Cross-platform comparison @@ -101,5 +95,5 @@ For a concrete example, see `experiments/dsv4_h200_dspark/`. ## Notes -- `platforms/nvidia_h200.env` assumes the legacy directory layout under `/data/user1/yy`. If your H200 machine uses different paths, override `VENV_VLLM_DSPARK`, `VENV_SGLANG`, `MODEL_ROOT`, or `SERVER_START_SCRIPT`. -- Native server management helpers are not yet as mature as the Docker helpers in `scripts/common/server_docker.sh`. The H200 wrapper currently delegates to the existing `scripts/start_dsv4_dspark_8card.sh`. +- `platforms/nvidia_h200.env` still carries venv defaults under `/data/user1/yy` (paths from another machine). Override `VENV_VLLM_DSPARK`, `VENV_SGLANG`, `MODEL_ROOT`, or `SERVER_START_SCRIPT` to match your machine. +- Native server management helpers are not as mature as the Docker helpers in `scripts/common/server_docker.sh`. The wrapper's default `SERVER_START_SCRIPT` (`scripts/start_dsv4_dspark_8card.sh`) no longer exists in this repo and must be overridden. diff --git a/docs/NEW_PLATFORM_GUIDE.md b/docs/NEW_PLATFORM_GUIDE.md new file mode 100644 index 0000000..c46084a --- /dev/null +++ b/docs/NEW_PLATFORM_GUIDE.md @@ -0,0 +1,139 @@ +# 新平台接入指南(New Platform Onboarding SOP) + +> 目标:新显卡(GPU/NPU)到货后,用最短时间跑起与既有平台口径一致的推理性能实验。 +> 读者:执行接入的人或 Agent。带 🤖 的步骤 Agent 可自动完成;带 🧑 的步骤需要人确认或提供信息。 + +## 0. 前置准备 🧑 + +接入前确认以下信息就绪,缺一项都会卡住后续步骤: + +| 项目 | 说明 | 示例 | +|---|---|---| +| 机器与驱动 | 8 卡整机,驱动/固件可用,`nvidia-smi` 或厂商等价工具(`xpu-smi`)能看到全部卡 | — | +| 推理镜像 | 该平台可用的 vLLM / SGLang 镜像或 native 环境 | `vllm/vllm-openai:latest`、`iregistry.baidu-int.com/xpu/sglang-p800-...` | +| 模型权重 | 模型在本机的路径与精度格式 | `/data1/hf_models/DeepSeek-V4-Flash`(FP8)、`/data1/models/DeepSeek-V4-Flash-INT8` | +| 上下文上限 | 该平台显存能支撑的 max context | H20:1M;P800 INT8:140k | +| 数据集 | ShareGPT 种子文件(random workload 用) | `datasets/ShareGPT_V3_unfiltered_cleaned_split.json` | +| 显存约束 | 单卡显存,预判 TP/DP 可行性(见 §4.2) | P800 96 GiB/卡 → TP2 放不下 INT8 专家权重 | + +## 1. 平台配置 🤖 + +1. 新建 `platforms/.env`(参考 `platforms/nvidia_h20.env` / `platforms/kunlun_p800.env`),至少包含 `CHIP`、`ACCELERATOR`、`HARDWARE`、`ENGINE`、`DEFAULT_PORT`、`MODEL_ROOT`;Docker 平台再加 `DOCKER_IMAGE`、`CONTAINER_NAME`;需要运行时补丁的加 `PATCH_ROOT` 并把补丁放到 `platforms/patches//`。 +2. 在 `scripts/common/platform.sh` 的自动探测分支里加上新平台(当前只识别 P800 / H20 / 其他 NVIDIA→H200,**新 NVIDIA 卡不改会被误判成 H200**)。 +3. 在 `platforms/README.md` 的平台表中加一行。 + +## 2. 复制实验目录 🤖 + +**NVIDIA 平台**(有官方 vLLM/SGLang 镜像)直接复制 H20 目录——它是路径正确、维护最新的范本,且 NVIDIA 平台间 `start_*_docker.sh` 零改动可复用: + +```bash +cp -r experiments/h20/dsv4_h20_vllm_tp_dp_matrix experiments//dsv4__vllm_tp_dp_matrix +cp -r experiments/h20/dsv4_h20_sglang_tp_dp_matrix experiments//dsv4__sglang_tp_dp_matrix +# 清理复制过来的结果与缓存 +rm -rf experiments//*/adaptive_results experiments//*/results experiments//*/__pycache__ +``` + +**非 NVIDIA 平台**(XPU 等)同样从 H20 目录复制,但需按 §4 重写启动脚本与平台回调,以 `experiments/p800/dsv4_p800_sglang_tp_dp_matrix` 为范例。 + +> 不要抄 `experiments/h200/` 或 `experiments/pro6000/` 的目录结构组织方式以外的内容:那里存在 `dsl`(应为 `osl`)的批量误改名残留,新目录以 H20 版为准。 + +## 3. 修改三件套配置 🤖(🧑 确认关键参数) + +### 3.1 `config.env`(核心参数面板) + +| 必改项 | 说明 | +|---|---| +| `EXPERIMENT` | 实验名,与目录名一致 | +| `MODEL_NAME` / `MODEL_PATH` / `SERVED_MODEL_NAME` | 本机模型路径(机器相关) | +| `DOCKER_IMAGE` / `CONTAINER_NAME` | 平台镜像;容器名专用化避免冲突 | +| `VLLM_PORT` / `SGLANG_PORT` | 避开已占用端口 | +| `CUDA_VISIBLE_DEVICES` / `XPU_VISIBLE_DEVICES` | 设备选择 | +| `MAX_MODEL_LEN` / `CONTEXT_LENGTH` | 按平台显存与需求设定 | +| `GPU_MEMORY_UTILIZATION` / `MEM_FRACTION_STATIC` | 显存水位(H20 0.9 / P800 0.8) | +| `KV_CACHE_DTYPE`、`BLOCK_SIZE` / `PAGE_SIZE` | KV cache 精度与块大小 | +| `PARALLEL_CONFIGS` | 按显存裁剪(见 §4.2) | +| `DATASET_PATH` | 本机数据集路径 | + +### 3.2 `adaptive_config.env` + +一般保持默认即可(起步 C=16、×2 或 +16 步进、TTFT_SLO_MS=4000、TTFT_GROUP_SKIP_MS=8000、num_prompts=5×C)。显存小或单请求 prefill 慢的平台,可适当降低 `SEARCH_MAX_CONCURRENCY`。 + +### 3.3 `matrix.json` + +按 `CONTEXT_LENGTH` 和显存砍 shape:超出上下文上限的 ISL 删掉,显存放不下的组合标 `N`。参照 H20(1M 上下文,7 个 ISL 档)与 P800(140k 上下文,5 个 ISL 档)两版的取舍。 + +## 4. 非 NVIDIA 平台的额外工作(以 P800 为范例)🧑🤖 + +### 4.1 需要重写的地方 + +- **`start_*_docker.sh`**:设备映射(如 `/dev/xpu*`)、厂商 env 变量、引擎 launch args(attention backend、量化、kv-cache dtype、cuda graph 等价物开关)、补丁挂载(如 P800 挂 `bench_serving.py` 补丁补 P95 指标)、容器内 bootstrap。 +- **`run_adaptive_concurrency*.sh` 的 5 个 `engine_*` 回调**:`engine_start_server` / `engine_stop_server`(pid kill 还是 `docker rm -f`)、`engine_run_bench`(独立客户端容器还是 `docker exec` 进服务端容器)、`engine_detect_oom`(日志文件还是 `docker logs`)、`engine_build_server_args`。 +- **GPU 监控**:公共库的 `adaptive_start_gpu_monitor` 写死 `nvidia-smi`,非 NVIDIA 平台在 run 脚本里用厂商工具重写该函数(P800 即用 `xpu-smi` 覆盖)。 +- **warmup 能力差异**:部分厂商镜像的 bench_serving 不支持 `--warmup-requests`,需设 `BENCH_WARMUP_MAX_REQUESTS=0` 并在 adaptive_config.env 里注明。 + +### 4.2 显存可行性预判(避免白跑) + +MoE 模型的专家权重在未开 EP 时按 TP 组切分、DP 副本间不共享:每卡专家权重 ≈ 专家总权重 / TP。实例:DSV4-Flash-INT8 路由专家约 264 GiB,P800 单卡 96 GiB → TP=2 每卡需 137 GiB,启动即 OOM,TP≥4 才可行。接入时先算一遍,把不可行的配置从 `PARALLEL_CONFIGS` 去掉,并把根因写进 config.env 注释(照 P800 的做法)。 + +## 5. 冒烟 🤖 + +```bash +cd experiments//dsv4___tp_dp_matrix + +# 1. dry-run:只打印计划,不起服务 +DRY_RUN=1 bash run_adaptive_concurrency_add16.sh + +# 2. 单 shape 小并发:验证起服务、发压测、出指标全链路 +RUN_ID=smoke-$(date +%Y%m%d-%H%M%S) \ +TP_LIST="8" ISL_LIST="1024" OSL_LIST="128" GRID_LIMIT=1 SEARCH_MAX_CONCURRENCY=8 \ +bash run_adaptive_concurrency_add16.sh + +# 3. 检查 smoke 产物:adaptive_results/smoke-*/adaptive_points.jsonl 应有 COMPLETED 点 +``` + +冒烟失败的常见原因:镜像 launch args 不被支持(查服务端日志)、端口冲突、数据集路径不存在、显存不足(回 §4.2)。 + +## 6. 正式跑 🤖 + +```bash +tmux new-session -d -s --adaptive \ + "cd $(pwd) && bash run_adaptive_concurrency_add16.sh" +``` + +- 心跳监控:`scripts/common/adaptive_heartbeat.sh` 或 `tail -f adaptive_results//logs/orchestrator.log`。 +- 中断了用 `RESUME_RUN_ID=` 断点续跑(按已终态 shape 去重)。 +- 完整参数说明见 `experiments/ADAPTIVE_CONCURRENCY_USAGE.md`。 + +## 7. 分析、报告与归档 🤖(🧑 确认发布) + +1. **分析**:用 Python 从 `adaptive_points.jsonl` / `adaptive_shapes.jsonl` 实算(SLO:TTFT P95 < 3s 且 TPOT P95 < 50ms,严格小于),禁止目测估算。 +2. **飞书报告**:按飞书 wiki「显卡性能报告 / 性能报告编写指南(README)」的骨架写到「显卡性能报告/<平台>」节点下;图表用 `envs/charts/bin/python` + matplotlib 生成。 +3. **多维表格**:在「模型推理适配 Bench 迭代跟踪」的对应平台表(没有则按现有 schema 新建)按"每成功探测点一行"回填;`SLO达标状态` 按上述口径判定。 +4. **提交**:实验代码 + `adaptive_results//` 的 jsonl/summary/manifest 入库;日志与 raw_outputs 不入库。 + +## 8. 完成检查清单 + +- [ ] `platforms/.env` 已建,`platform.sh` 探测分支已加,`platforms/README.md` 已更新 +- [ ] `config.env` / `adaptive_config.env` / `matrix.json` 三件套已按平台调整 +- [ ] smoke 通过(有 COMPLETED 探测点) +- [ ] 正式 run 的 `run_manifest.json` 参数正确 +- [ ] 飞书报告已发布到对应平台节点下 +- [ ] 多维表格已回填 +- [ ] git 提交不含日志、raw_outputs、gpu_logs + +## 9. 已知坑 + +1. **公共库路径层级**:`experiments///` 下的脚本引用公共组件必须是 `${SCRIPT_DIR}/../../../scripts/common`;`experiments/TEMPLATE/` 下是 `../../`。写错会在 source 阶段直接失败。 +2. **机器相关绝对路径**:模型、数据集、venv 路径随机器不同(`/data1/...`、`/data3/...`),换机器先改 config.env,不要在脚本里写死。 +3. **TTFT SLO 停止 ≠ 饱和**:add16 流派的搜索在 P95 TTFT 超 4000ms 即停,绝大多数 shape 不会到达 TPS 平台期,报告中的"最佳 TPS"是 TTFT 边界内最高值,分析时不要当成吞吐上限。 +4. **`dsl` 命名坑**:h200/pro6000 目录部分文件把 `osl` 误改名成 `dsl`(含 CSV 表头),复制时以 h20 目录为准。 +5. **结果结构以 jsonl 为准**:`.gitignore` 排除 csv/logs,跨机器汇总数据时用 `adaptive_points.jsonl`、`adaptive_shapes.jsonl`、`adaptive_summary.*`、`run_manifest.json`。 + +## 10. 新模型接入(GLM5.2) + +GLM5.2 **完全复用**本 SOP 的实验与报告流程,差异只在配置层: + +1. 新目录命名 `experiments//glm52___tp_dp_matrix/`; +2. `config.env` 改模型路径、`SERVED_MODEL_NAME`、量化方式(按发布精度)、`CONTEXT_LENGTH`(按模型实际上限)、引擎 launch args(attention backend / rope / 量化 flag 等按模型要求); +3. `matrix.json` 按模型上下文上限调整 ISL 档; +4. SLO 口径、报告骨架、多维表格结构不变;多维表格的 `模型名称` 选/新增 GLM5.2 对应选项。 diff --git a/envs/README.md b/envs/README.md index 5ecd51f..cd2c426 100644 --- a/envs/README.md +++ b/envs/README.md @@ -40,3 +40,4 @@ envs/ - `sglang/` - `sglang-src/` - `vllm/` +- `charts/`(matplotlib 绘图环境,用于生成性能报告图表) diff --git a/experiments/ADAPTIVE_CONCURRENCY_USAGE.md b/experiments/ADAPTIVE_CONCURRENCY_USAGE.md index 3a1093b..93964f6 100644 --- a/experiments/ADAPTIVE_CONCURRENCY_USAGE.md +++ b/experiments/ADAPTIVE_CONCURRENCY_USAGE.md @@ -6,16 +6,18 @@ ## 1. 进入目录 +以下路径均相对仓库根目录(如 H200 机器上的 `experiments/h200/...`)。 + vLLM: ```bash -cd /data3/yy/sskj/experiments/dsv4_h200_vllm_tp_dp_matrix +cd experiments/h200/dsv4_h200_vllm_tp_dp_matrix ``` SGLang: ```bash -cd /data3/yy/sskj/experiments/dsv4_h200_sglang_tp_dp_matrix +cd experiments/h200/dsv4_h200_sglang_tp_dp_matrix ``` 两个目录的入口都是: @@ -43,19 +45,19 @@ ISL_LIST="1024" \ OSL_LIST="128" \ GRID_LIMIT=1 \ SEARCH_MAX_CONCURRENCY=8 \ -RESULT_BASE=/data3/yy/sskj_adaptive_results/smoke \ +RESULT_BASE="$PWD/adaptive_results/smoke" \ bash run_adaptive_concurrency.sh ``` ## 4. 正式运行 -长任务放进 tmux。vLLM 示例: +长任务放进 tmux。以下命令假设在仓库根目录下执行。vLLM 示例: ```bash tmux new-session -d -s vllm-adaptive \ - "cd /data3/yy/sskj/experiments/dsv4_h200_vllm_tp_dp_matrix && \ + "cd experiments/h200/dsv4_h200_vllm_tp_dp_matrix && \ RUN_ID=vllm-\$(date +%Y%m%d-%H%M%S) \ - RESULT_BASE=/data3/yy/sskj_adaptive_results/vllm \ + RESULT_BASE=\$PWD/adaptive_results/vllm \ bash run_adaptive_concurrency.sh" ``` @@ -63,9 +65,9 @@ SGLang 示例: ```bash tmux new-session -d -s sglang-adaptive \ - "cd /data3/yy/sskj/experiments/dsv4_h200_sglang_tp_dp_matrix && \ + "cd experiments/h200/dsv4_h200_sglang_tp_dp_matrix && \ RUN_ID=sglang-\$(date +%Y%m%d-%H%M%S) \ - RESULT_BASE=/data3/yy/sskj_adaptive_results/sglang \ + RESULT_BASE=\$PWD/adaptive_results/sglang \ bash run_adaptive_concurrency.sh" ``` diff --git a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency.sh index 7972a9e..8f48c5f 100755 --- a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh index c98b531..d7b26a0 100755 --- a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_bench.sh b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_bench.sh index 3fbb7b3..948aa52 100755 --- a/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_bench.sh +++ b/experiments/h20/dsv4_h20_sglang_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -524,7 +524,7 @@ run_parallel_config() { # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ + "$PYTHON" "${SCRIPT_DIR}/../../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency.sh index f76a93d..3454a44 100755 --- a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh index fb18812..7c905e8 100755 --- a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh.bak.20260716-084832 b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh.bak.20260716-084832 deleted file mode 100755 index 1a4b3bc..0000000 --- a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh.bak.20260716-084832 +++ /dev/null @@ -1,167 +0,0 @@ -#!/usr/bin/env bash -# Find the Total-TPS saturation concurrency for each vLLM TP/DP/ISL/OSL shape. -set -Eeuo pipefail - -SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" -EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" - -# shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" -# shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" -# shellcheck source=/dev/null -source "${SCRIPT_DIR}/config.env" -# shellcheck source=/dev/null -source "${SCRIPT_DIR}/adaptive_config.env" -# shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" - -ENGINE="vllm" -ENGINE_PORT="$VLLM_PORT" -RESULT_BASE="${RESULT_BASE:-${SCRIPT_DIR}/adaptive_results}" -ACTIVE_ENGINE_SERVER_LOG="" - -if [[ -x "${VENV_CLIENT}/bin/python" ]]; then - PYTHON="${VENV_CLIENT}/bin/python" -else - PYTHON="$(command -v python3)" -fi - -DOCKER_IMAGE="${DOCKER_IMAGE:-vllm/vllm-openai:latest}" -DOCKER_CLIENT_IMAGE="${DOCKER_CLIENT_IMAGE:-lmsysorg/sglang:latest}" - -engine_is_healthy() { - curl --fail --silent --show-error --max-time 5 \ - "http://127.0.0.1:${ENGINE_PORT}/health" >/dev/null 2>&1 -} - -engine_stop_server() { - local tp="$1" - local dp="$2" - local pid_file="${RUNTIME_BASE}/${EXPERIMENT}_vllm_tp${tp}_dp${dp}.pid" - - if [[ -f "$pid_file" ]]; then - local pid - pid="$(cat "$pid_file")" - if kill -0 "$pid" 2>/dev/null; then - log "stopping vllm server pid=${pid} tp=${tp} dp=${dp}" - kill "$pid" 2>/dev/null || true - sleep 5 - kill -9 "$pid" 2>/dev/null || true - fi - rm -f "$pid_file" - fi - docker rm -f "${EXPERIMENT}_vllm_tp${tp}_dp${dp}" >/dev/null 2>&1 || true - ACTIVE_ENGINE_SERVER_LOG="" - sleep 2 -} - -engine_build_server_args() { - local tp="$1" - local dp="$2" - local -a args=( - vllm serve "$MODEL_PATH" - --trust-remote-code - --kv-cache-dtype "$KV_CACHE_DTYPE" - --block-size "$BLOCK_SIZE" - --tensor-parallel-size "$tp" - --gpu-memory-utilization "$GPU_MEMORY_UTILIZATION" - --max-model-len "$MAX_MODEL_LEN" - --max-num-seqs "$MAX_NUM_SEQS" - --no-enable-flashinfer-autotune - --host 0.0.0.0 - --port "$ENGINE_PORT" - ) - if (( dp > 1 )); then - args+=(--data-parallel-size "$dp") - fi - printf '%q ' "${args[@]}" -} - -engine_start_server() { - local tp="$1" - local dp="$2" - local outer_log="${ADAPTIVE_LOG_DIR}/vllm_tp${tp}_dp${dp}.server.outer.log" - log "starting vllm server tp=${tp} dp=${dp}" - bash "${SCRIPT_DIR}/start_vllm_dp.sh" "$tp" "$dp" >> "$outer_log" 2>&1 - if ! engine_is_healthy; then - log "ERROR: vllm health check failed tp=${tp} dp=${dp}" - return 1 - fi - ACTIVE_ENGINE_SERVER_LOG="$( - find "${RUNTIME_BASE}/logs" -maxdepth 1 -type f \ - -name "${EXPERIMENT}_vllm*tp${tp}_dp${dp}_*.log" \ - -printf '%T@ %p\n' 2>/dev/null | sort -nr | head -n 1 | cut -d' ' -f2- - )" - log "vllm server healthy tp=${tp} dp=${dp} log=${ACTIVE_ENGINE_SERVER_LOG:-unknown}" -} - -engine_detect_oom() { - local detail_log="$1" - local tp="$2" - local dp="$3" - local pattern='CUDA out of memory|torch\.OutOfMemoryError|OutOfMemory|out of memory|OOM|RESOURCE_EXHAUSTED|Failed to allocate memory' - local outer_log="${ADAPTIVE_LOG_DIR}/vllm_tp${tp}_dp${dp}.server.outer.log" - local -a logs=("$detail_log" "$outer_log") - if [[ -n "$ACTIVE_ENGINE_SERVER_LOG" ]]; then - logs+=("$ACTIVE_ENGINE_SERVER_LOG") - fi - grep -Eiq "$pattern" "${logs[@]}" 2>/dev/null -} - -engine_run_bench() { - local isl="$1" - local osl="$2" - local concurrency="$3" - local num_prompts="$4" - local output_file="$5" - local warmup_requests - warmup_requests="$(adaptive_warmup_request_count "$concurrency")" - local -a bench_args=( - --backend vllm - --host 127.0.0.1 - --port "$ENGINE_PORT" - --dataset-name "$BENCH_DATASET_NAME" - --random-input-len "$isl" - --random-output-len "$osl" - --random-range-ratio "$RANDOM_RANGE_RATIO" - --num-prompts "$num_prompts" - --max-concurrency "$concurrency" - --request-rate 10000 - --warmup-requests "$warmup_requests" - --output-file "$output_file" - --output-details - --disable-tqdm - ) - if [[ "$BENCH_DATASET_NAME" == "random" ]]; then - bench_args+=(--dataset-path "$DATASET_PATH") - else - bench_args+=(--tokenize-prompt) - fi - - if [[ "$USE_DOCKER_CLIENT" == "1" ]]; then - local -a volume_args=(-v "${MODEL_PATH}:${MODEL_PATH}:ro" -v "${RESULT_BASE}:${RESULT_BASE}") - if [[ "$BENCH_DATASET_NAME" == "random" ]]; then - volume_args+=(-v "${DATASET_PATH}:${DATASET_PATH}:ro") - fi - docker run --rm \ - --network host \ - "${volume_args[@]}" \ - -e PYTHONUNBUFFERED=1 \ - -e HF_HUB_OFFLINE=1 \ - -e TRANSFORMERS_OFFLINE=1 \ - "$DOCKER_CLIENT_IMAGE" \ - python -m sglang.bench_serving "${bench_args[@]}" - else - "$PYTHON" -m sglang.bench_serving "${bench_args[@]}" - fi -} - -export -f engine_run_bench -export ENGINE_PORT MODEL_PATH RESULT_BASE DOCKER_CLIENT_IMAGE USE_DOCKER_CLIENT -export BENCH_DATASET_NAME DATASET_PATH RANDOM_RANGE_RATIO BENCH_WARMUP_MAX_REQUESTS PYTHON - -export SEARCH_START_CONCURRENCY=16 -export SEARCH_ADDEND=16 - -adaptive_main "$@" diff --git a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_bench.sh b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_bench.sh index 1db82ea..3523651 100755 --- a/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_bench.sh +++ b/experiments/h20/dsv4_h20_vllm_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -512,7 +512,7 @@ run_parallel_config() { # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ + "$PYTHON" "${SCRIPT_DIR}/../../../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/h200/dsv4_h200_256k_4k_probe/run_bench.sh b/experiments/h200/dsv4_h200_256k_4k_probe/run_bench.sh index 5b6b8ef..d106448 100644 --- a/experiments/h200/dsv4_h200_256k_4k_probe/run_bench.sh +++ b/experiments/h200/dsv4_h200_256k_4k_probe/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -95,7 +95,7 @@ run_warmup() { num=1 log "warming up ${backend} (input=${input_len}, output=${output_len}, num=${num})" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/warmup.py" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/warmup.py" \ --backend "$backend" \ --port "$port" \ --input-len "$input_len" \ @@ -189,7 +189,7 @@ parse_backend() { local backend="$1" local result_root="${RESULT_BASE}/${RUN_ID}/${backend}" log "parsing ${backend} results in ${result_root}" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ >> "${result_root}/logs/parse.log" 2>&1 || { log "WARNING: parser failed for ${backend}; see ${result_root}/logs/parse.log" } @@ -264,7 +264,7 @@ parse_backend vllm # Generate comparison. log "generating comparison report" -"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/compare.py" \ +"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/compare.py" \ --sglang "${RESULT_BASE}/${RUN_ID}/sglang" \ --vllm "${RESULT_BASE}/${RUN_ID}/vllm" \ --output "${RESULT_BASE}/${RUN_ID}/comparison.md" \ diff --git a/experiments/h200/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh b/experiments/h200/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh index 532b5fb..b972fc0 100644 --- a/experiments/h200/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh +++ b/experiments/h200/dsv4_h200_64k_sglang_vs_vllm/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -95,7 +95,7 @@ run_warmup() { num=1 log "warming up ${backend} (input=${input_len}, output=${output_len}, num=${num})" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/warmup.py" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/warmup.py" \ --backend "$backend" \ --port "$port" \ --input-len "$input_len" \ @@ -189,7 +189,7 @@ parse_backend() { local backend="$1" local result_root="${RESULT_BASE}/${RUN_ID}/${backend}" log "parsing ${backend} results in ${result_root}" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ >> "${result_root}/logs/parse.log" 2>&1 || { log "WARNING: parser failed for ${backend}; see ${result_root}/logs/parse.log" } @@ -264,7 +264,7 @@ parse_backend vllm # Generate comparison. log "generating comparison report" -"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/compare.py" \ +"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/compare.py" \ --sglang "${RESULT_BASE}/${RUN_ID}/sglang" \ --vllm "${RESULT_BASE}/${RUN_ID}/vllm" \ --output "${RESULT_BASE}/${RUN_ID}/comparison.md" \ diff --git a/experiments/h200/dsv4_h200_dspark/run_bench.sh b/experiments/h200/dsv4_h200_dspark/run_bench.sh index c4b677a..d63c90c 100755 --- a/experiments/h200/dsv4_h200_dspark/run_bench.sh +++ b/experiments/h200/dsv4_h200_dspark/run_bench.sh @@ -9,9 +9,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_long_context_matrix/run_bench.sh b/experiments/h200/dsv4_h200_long_context_matrix/run_bench.sh index fa8c89e..0734ec5 100755 --- a/experiments/h200/dsv4_h200_long_context_matrix/run_bench.sh +++ b/experiments/h200/dsv4_h200_long_context_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -107,7 +107,7 @@ run_warmup() { fi log "warming up ${backend} (input=${input_len}, output=${output_len}, num=1)" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/warmup.py" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/warmup.py" \ --backend vllm \ --port "$port" \ --input-len "$input_len" \ @@ -235,7 +235,7 @@ parse_backend() { parser_backend="vllm" fi log "parsing ${backend} results in ${result_root} (parser_backend=${parser_backend})" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend "$parser_backend" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend "$parser_backend" \ >> "${result_root}/logs/parse.log" 2>&1 || { log "WARNING: parser failed for ${backend}; see ${result_root}/logs/parse.log" } diff --git a/experiments/h200/dsv4_h200_max_context_length/run_bench.sh b/experiments/h200/dsv4_h200_max_context_length/run_bench.sh index bc1515d..717b0a1 100755 --- a/experiments/h200/dsv4_h200_max_context_length/run_bench.sh +++ b/experiments/h200/dsv4_h200_max_context_length/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_sglang/run_bench.sh b/experiments/h200/dsv4_h200_sglang/run_bench.sh index 8c494b1..3b4be68 100755 --- a/experiments/h200/dsv4_h200_sglang/run_bench.sh +++ b/experiments/h200/dsv4_h200_sglang/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh index 0979a46..7972a9e 100755 --- a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh index 375fec7..728c895 100755 --- a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh index 53d0fbc..b4f47b9 100755 --- a/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh +++ b/experiments/h200/dsv4_h200_sglang_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -522,7 +522,7 @@ run_parallel_config() { # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ + "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/h200/dsv4_h200_sglang_vs_vllm/run_bench.sh b/experiments/h200/dsv4_h200_sglang_vs_vllm/run_bench.sh index 4299eed..05c943d 100755 --- a/experiments/h200/dsv4_h200_sglang_vs_vllm/run_bench.sh +++ b/experiments/h200/dsv4_h200_sglang_vs_vllm/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -100,7 +100,7 @@ run_warmup() { fi log "warming up ${backend} (phase=${phase}, input=${input_len}, output=${output_len}, num=${num})" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/warmup.py" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/warmup.py" \ --backend "$backend" \ --port "$port" \ --input-len "$input_len" \ @@ -264,7 +264,7 @@ parse_backend() { local backend="$1" local result_root="${RESULT_BASE}/${RUN_ID}/${backend}" log "parsing ${backend} results in ${result_root}" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend "$backend" \ >> "${result_root}/logs/parse.log" 2>&1 || { log "WARNING: parser failed for ${backend}; see ${result_root}/logs/parse.log" } @@ -348,7 +348,7 @@ parse_backend vllm # Generate comparison. log "generating comparison report" -"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/compare.py" \ +"${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/compare.py" \ --sglang "${RESULT_BASE}/${RUN_ID}/sglang" \ --vllm "${RESULT_BASE}/${RUN_ID}/vllm" \ --output "${RESULT_BASE}/${RUN_ID}/comparison.md" \ diff --git a/experiments/h200/dsv4_h200_vllm/run_bench.sh b/experiments/h200/dsv4_h200_vllm/run_bench.sh index 257f57b..b893d05 100755 --- a/experiments/h200/dsv4_h200_vllm/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_vllm_dspark_vs_default/run_bench.sh b/experiments/h200/dsv4_h200_vllm_dspark_vs_default/run_bench.sh index 3cd9019..51174d3 100755 --- a/experiments/h200/dsv4_h200_vllm_dspark_vs_default/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_dspark_vs_default/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_vllm_mtp_vs_default/run_bench.sh b/experiments/h200/dsv4_h200_vllm_mtp_vs_default/run_bench.sh index 7715bb3..e3e5da9 100755 --- a/experiments/h200/dsv4_h200_vllm_mtp_vs_default/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_mtp_vs_default/run_bench.sh @@ -7,9 +7,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -68,7 +68,7 @@ start_server() { run_warmup() { log "warming up mtp (input=4000, output=512, num=2)" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/warmup.py" \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/warmup.py" \ --backend vllm \ --port "$MTP_PORT" \ --input-len 4000 \ @@ -153,7 +153,7 @@ run_benchmark() { parse_mtp() { local result_root="${RESULT_BASE}/${RUN_ID}/mtp" log "parsing mtp results in ${result_root}" - "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ + "${VENV_CLIENT}/bin/python" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ >> "${result_root}/logs/parse.log" 2>&1 || { log "WARNING: parser failed for mtp; see ${result_root}/logs/parse.log" } diff --git a/experiments/h200/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh b/experiments/h200/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh index 217473d..fb15c4a 100755 --- a/experiments/h200/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_tp2_custom_bench/run_bench.sh @@ -15,9 +15,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh b/experiments/h200/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh index ccf7eee..c95a064 100755 --- a/experiments/h200/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_tp4_custom_bench/run_bench.sh @@ -15,9 +15,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh b/experiments/h200/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh index 0e8adfa..89034fc 100755 --- a/experiments/h200/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_tp8_custom_bench/run_bench.sh @@ -14,9 +14,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh index ca23f6a..f76a93d 100755 --- a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh index 1b2b1b2..fa47973 100755 --- a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh index f004020..b2cbe22 100755 --- a/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh +++ b/experiments/h200/dsv4_h200_vllm_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -512,7 +512,7 @@ run_parallel_config() { # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ + "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend vllm \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/p800/dsv4_p800_256k_4k_probe/run_bench.sh b/experiments/p800/dsv4_p800_256k_4k_probe/run_bench.sh index adfe559..27ee8f4 100755 --- a/experiments/p800/dsv4_p800_256k_4k_probe/run_bench.sh +++ b/experiments/p800/dsv4_p800_256k_4k_probe/run_bench.sh @@ -5,13 +5,13 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/server_docker.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/server_docker.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/bench_client_docker.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/bench_client_docker.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_256k_4k_probe/start_server.sh b/experiments/p800/dsv4_p800_256k_4k_probe/start_server.sh index 9e58b8d..0fbdbaf 100755 --- a/experiments/p800/dsv4_p800_256k_4k_probe/start_server.sh +++ b/experiments/p800/dsv4_p800_256k_4k_probe/start_server.sh @@ -5,7 +5,7 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/server_docker.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/server_docker.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_long_context_matrix/run_bench.sh b/experiments/p800/dsv4_p800_long_context_matrix/run_bench.sh index 4b4c7d2..8b435eb 100755 --- a/experiments/p800/dsv4_p800_long_context_matrix/run_bench.sh +++ b/experiments/p800/dsv4_p800_long_context_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -217,7 +217,7 @@ with open('${host_output_file}') as f: parse_backend() { log "parsing results in ${RESULT_ROOT}" - python3 "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$RESULT_ROOT" --backend sglang \ + python3 "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$RESULT_ROOT" --backend sglang \ >> "${LOG_DIR}/parse.log" 2>&1 || { log "WARNING: parse_backend.py failed; see ${LOG_DIR}/parse.log" } diff --git a/experiments/p800/dsv4_p800_long_context_matrix/start_sglang.sh b/experiments/p800/dsv4_p800_long_context_matrix/start_sglang.sh index 4489ace..f4c2636 100755 --- a/experiments/p800/dsv4_p800_long_context_matrix/start_sglang.sh +++ b/experiments/p800/dsv4_p800_long_context_matrix/start_sglang.sh @@ -17,9 +17,9 @@ fi SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_max_context_length/run_bench.sh b/experiments/p800/dsv4_p800_max_context_length/run_bench.sh index 2b5958d..850b61d 100755 --- a/experiments/p800/dsv4_p800_max_context_length/run_bench.sh +++ b/experiments/p800/dsv4_p800_max_context_length/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_max_context_length/start_server.sh b/experiments/p800/dsv4_p800_max_context_length/start_server.sh index 0dbb973..9a6bfb2 100755 --- a/experiments/p800/dsv4_p800_max_context_length/start_server.sh +++ b/experiments/p800/dsv4_p800_max_context_length/start_server.sh @@ -10,9 +10,9 @@ fi SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_sglang/run_bench.sh b/experiments/p800/dsv4_p800_sglang/run_bench.sh index bbd3abe..cf08762 100755 --- a/experiments/p800/dsv4_p800_sglang/run_bench.sh +++ b/experiments/p800/dsv4_p800_sglang/run_bench.sh @@ -5,13 +5,13 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/server_docker.sh" +source "${SCRIPT_DIR}/../../../scripts/common/server_docker.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/bench_client_docker.sh" +source "${SCRIPT_DIR}/../../../scripts/common/bench_client_docker.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_sglang/start_server.sh b/experiments/p800/dsv4_p800_sglang/start_server.sh index 435865a..4309dc7 100755 --- a/experiments/p800/dsv4_p800_sglang/start_server.sh +++ b/experiments/p800/dsv4_p800_sglang/start_server.sh @@ -5,7 +5,7 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/server_docker.sh" +source "${SCRIPT_DIR}/../../../scripts/common/server_docker.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_points.jsonl b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_points.jsonl index 99643ae..cc76a52 100644 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_points.jsonl +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_points.jsonl @@ -1 +1,121 @@ {"status":"COMPLETED","completed":80,"failed":0,"duration_s":41.2688269931823,"request_tps":1.9385091806272121,"input_tps":1985.0334009622652,"output_tps":248.12917512028315,"total_tps":2233.1625760825486,"mean_input_tokens":1024.0,"mean_output_tokens":128.0,"ttft_p50_ms":1839.1637965105474,"ttft_p95_ms":5156.00185089279,"ttft_p99_ms":5158.878551644739,"tpot_p50_ms":47.26517358684809,"tpot_p95_ms":54.75502148450271,"tpot_p99_ms":55.868444815663054,"e2e_p50_ms":7647.177290520631,"e2e_p95_ms":11683.607439068146,"e2e_p99_ms":11687.237907222006,"itl_p50_ms":38.27845409978181,"itl_p95_ms":61.400388937909156,"itl_p99_ms":83.85505519807339,"validation_errors":[],"timestamp":"2026-07-16T07:48:44+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":67.59083261387423,"request_tps":1.1835924622650464,"input_tps":1211.9986813594076,"output_tps":302.9996703398519,"total_tps":1514.9983516992595,"mean_input_tokens":1024.0,"mean_output_tokens":256.0,"ttft_p50_ms":1088.1021856330335,"ttft_p95_ms":1940.051195200067,"ttft_p99_ms":1945.7215382088907,"tpot_p50_ms":48.00720819381668,"tpot_p95_ms":51.906457407692194,"tpot_p99_ms":52.662985878384795,"e2e_p50_ms":13323.23444657959,"e2e_p95_ms":14546.717717882711,"e2e_p99_ms":14597.288672290742,"itl_p50_ms":38.109477958641946,"itl_p95_ms":62.32210020534694,"itl_p99_ms":91.81069683981994,"validation_errors":[],"timestamp":"2026-07-16T07:50:13+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o256_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":85.51100934692658,"request_tps":1.8711040978462112,"input_tps":1916.0105961945203,"output_tps":479.0026490486301,"total_tps":2395.0132452431503,"mean_input_tokens":1024.0,"mean_output_tokens":256.0,"ttft_p50_ms":1161.2384765176103,"ttft_p95_ms":3756.2352412380274,"ttft_p99_ms":3780.965418519918,"tpot_p50_ms":61.98459189412568,"tpot_p95_ms":66.44078092480663,"tpot_p99_ms":66.8051291848259,"e2e_p50_ms":16901.625977130607,"e2e_p95_ms":18837.460162944622,"e2e_p99_ms":18985.462763826363,"itl_p50_ms":44.66060339473188,"itl_p95_ms":68.57059714384377,"itl_p99_ms":449.1709415311925,"validation_errors":[],"timestamp":"2026-07-16T07:52:00+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":256,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":58.086855,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c32_i1024_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c32_i1024_o256_a1.log"} +{"status":"COMPLETED","completed":240,"failed":0,"duration_s":102.57387904892676,"request_tps":2.3397769707580456,"input_tps":2395.9316180562387,"output_tps":598.9829045140597,"total_tps":2994.9145225702982,"mean_input_tokens":1024.0,"mean_output_tokens":256.0,"ttft_p50_ms":4857.287564082071,"ttft_p95_ms":5687.151023850311,"ttft_p99_ms":5947.1146802348085,"tpot_p50_ms":60.30845692353871,"tpot_p95_ms":72.36586667291418,"tpot_p99_ms":78.81875480012964,"e2e_p50_ms":20432.19355738256,"e2e_p95_ms":21489.17214354733,"e2e_p99_ms":22516.328958119266,"itl_p50_ms":51.60140746738762,"itl_p95_ms":73.69143945397809,"itl_p99_ms":89.00051706237718,"validation_errors":[],"timestamp":"2026-07-16T07:54:03+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":256,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":1,"gain_pct":25.047932,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c48_i1024_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c48_i1024_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":123.35641185496934,"request_tps":0.6485272941795385,"input_tps":664.0919492398474,"output_tps":332.0459746199237,"total_tps":996.137923859771,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":1815.1043795514852,"ttft_p95_ms":2051.647847087588,"ttft_p99_ms":2054.4600446219556,"tpot_p50_ms":44.89284182270832,"tpot_p95_ms":47.29539701357892,"tpot_p99_ms":47.49259979092314,"e2e_p50_ms":24692.195138079114,"e2e_p95_ms":25352.763180364855,"e2e_p99_ms":25618.341889507137,"itl_p50_ms":37.906087934970856,"itl_p95_ms":61.653098394162946,"itl_p99_ms":66.99016668368131,"validation_errors":[],"timestamp":"2026-07-16T07:56:28+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":151.8317316111643,"request_tps":1.0537981639421354,"input_tps":1079.0893198767467,"output_tps":539.5446599383733,"total_tps":1618.63397981512,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":1056.6700529307127,"ttft_p95_ms":3730.0430950359428,"ttft_p99_ms":3803.3011273038574,"tpot_p50_ms":55.7243038511538,"tpot_p95_ms":60.350505806735306,"tpot_p99_ms":61.641385694806246,"e2e_p50_ms":29571.674760896713,"e2e_p95_ms":32270.35216266522,"e2e_p99_ms":32279.17558920337,"itl_p50_ms":45.009728986769915,"itl_p95_ms":67.9298322298564,"itl_p99_ms":236.92634512903186,"validation_errors":[],"timestamp":"2026-07-16T07:59:21+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":512,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":62.49095,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c32_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c32_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":240,"failed":0,"duration_s":194.6086190820206,"request_tps":1.2332444530570794,"input_tps":1262.8423199304493,"output_tps":631.4211599652247,"total_tps":1894.263479895674,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":5219.216491095722,"ttft_p95_ms":5779.679291823413,"ttft_p99_ms":6389.283453426324,"tpot_p50_ms":58.08764942142558,"tpot_p95_ms":62.69429164324429,"tpot_p99_ms":91.06430042089896,"e2e_p50_ms":34896.657894947566,"e2e_p95_ms":35910.429730673786,"e2e_p99_ms":49593.8836463075,"itl_p50_ms":51.571647985838354,"itl_p95_ms":73.75171015737578,"itl_p99_ms":77.6099856174551,"validation_errors":[],"timestamp":"2026-07-16T08:02:57+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":512,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":1,"gain_pct":17.028526,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c48_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c48_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":239.2404650580138,"request_tps":0.3343915920770371,"input_tps":342.416990286886,"output_tps":342.416990286886,"total_tps":684.833980573772,"mean_input_tokens":1024.0,"mean_output_tokens":1024.0,"ttft_p50_ms":635.5562334647402,"ttft_p95_ms":1903.8669666741043,"ttft_p99_ms":1906.7863759305328,"tpot_p50_ms":45.73389088913875,"tpot_p95_ms":46.81248970833659,"tpot_p99_ms":47.236905306400615,"e2e_p50_ms":47589.559288928285,"e2e_p95_ms":48763.24818015564,"e2e_p99_ms":49025.48799880082,"itl_p50_ms":38.22040546219796,"itl_p95_ms":62.13192297145724,"itl_p99_ms":69.20204504393043,"validation_errors":[],"timestamp":"2026-07-16T08:07:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o1024_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":283.1469709521625,"request_tps":0.5650775618822773,"input_tps":578.639423367452,"output_tps":578.639423367452,"total_tps":1157.278846734904,"mean_input_tokens":1024.0,"mean_output_tokens":1024.0,"ttft_p50_ms":1039.5908689824864,"ttft_p95_ms":3708.7073334376323,"ttft_p99_ms":3736.5557909896597,"tpot_p50_ms":53.18489079777869,"tpot_p95_ms":56.531733996098275,"tpot_p99_ms":56.90662554211692,"e2e_p50_ms":55562.747771968134,"e2e_p95_ms":58622.54847800359,"e2e_p99_ms":58864.768436276354,"itl_p50_ms":44.60177046712488,"itl_p95_ms":68.54137406917289,"itl_p99_ms":104.98853123513977,"validation_errors":[],"timestamp":"2026-07-16T08:12:22+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":1024,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":68.986773,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c32_i1024_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c32_i1024_o1024_a1.log"} +{"status":"COMPLETED","completed":240,"failed":0,"duration_s":386.91459054895677,"request_tps":0.6202919348672961,"input_tps":635.1789413041112,"output_tps":635.1789413041112,"total_tps":1270.3578826082223,"mean_input_tokens":1024.0,"mean_output_tokens":1024.0,"ttft_p50_ms":5370.0150235090405,"ttft_p95_ms":15133.30938139696,"ttft_p99_ms":32185.267814283718,"tpot_p50_ms":60.0038268968676,"tpot_p95_ms":91.36051824342555,"tpot_p99_ms":103.14346713503731,"e2e_p50_ms":65575.55996044539,"e2e_p95_ms":99691.6358396411,"e2e_p99_ms":112446.33609429697,"itl_p50_ms":51.48644756991416,"itl_p95_ms":73.74423729488626,"itl_p99_ms":83.97575234295799,"validation_errors":[],"timestamp":"2026-07-16T08:19:12+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":1024,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":1,"gain_pct":9.771114,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c48_i1024_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c48_i1024_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":473.66198480315506,"request_tps":0.16889681369140588,"input_tps":172.95033721999962,"output_tps":345.90067443999925,"total_tps":518.8510116599988,"mean_input_tokens":1024.0,"mean_output_tokens":2048.0,"ttft_p50_ms":1107.0592959877104,"ttft_p95_ms":1914.7031036205587,"ttft_p99_ms":1937.422300640028,"tpot_p50_ms":45.60600052494192,"tpot_p95_ms":46.180353910758,"tpot_p99_ms":46.185254772231495,"e2e_p50_ms":94571.73220254481,"e2e_p95_ms":95557.28795382893,"e2e_p99_ms":95606.2740921299,"itl_p50_ms":38.73952745925635,"itl_p95_ms":62.4737826641649,"itl_p99_ms":67.41200149524958,"validation_errors":[],"timestamp":"2026-07-16T08:27:27+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o2048_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":562.1656430249568,"request_tps":0.28461362231077675,"input_tps":291.4443492462354,"output_tps":582.8886984924708,"total_tps":874.3330477387061,"mean_input_tokens":1024.0,"mean_output_tokens":2048.0,"ttft_p50_ms":1720.5767229897901,"ttft_p95_ms":3765.552905423101,"ttft_p99_ms":3778.2479317323305,"tpot_p50_ms":53.25908665974094,"tpot_p95_ms":55.20441076613287,"tpot_p99_ms":55.39557766710268,"e2e_p50_ms":109979.57844298799,"e2e_p95_ms":114323.36549273459,"e2e_p99_ms":114707.30229713488,"itl_p50_ms":46.27890046685934,"itl_p95_ms":69.01592136127874,"itl_p99_ms":77.31038637226447,"validation_errors":[],"timestamp":"2026-07-16T08:37:12+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":2048,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":68.513317,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c32_i1024_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c32_i1024_o2048_a1.log"} +{"status":"COMPLETED","completed":240,"failed":0,"duration_s":785.6881854759995,"request_tps":0.3054646925288802,"input_tps":312.7958451495733,"output_tps":625.5916902991466,"total_tps":938.3875354487201,"mean_input_tokens":1024.0,"mean_output_tokens":2048.0,"ttft_p50_ms":5525.519606540911,"ttft_p95_ms":38321.421813906585,"ttft_p99_ms":61896.90983353647,"tpot_p50_ms":60.786382354451064,"tpot_p95_ms":94.29399440951978,"tpot_p99_ms":106.96621055514711,"e2e_p50_ms":128179.90333400667,"e2e_p95_ms":211186.33220964108,"e2e_p99_ms":235519.09909247872,"itl_p50_ms":52.96039255335927,"itl_p95_ms":74.77014367468655,"itl_p99_ms":93.00403241533787,"validation_errors":[],"timestamp":"2026-07-16T08:50:41+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":2048,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":1,"gain_pct":7.326097,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c48_i1024_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c48_i1024_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":948.3546966928989,"request_tps":0.08435662340153519,"input_tps":86.38118236317203,"output_tps":345.5247294526881,"total_tps":431.9059118158602,"mean_input_tokens":1024.0,"mean_output_tokens":4096.0,"ttft_p50_ms":587.1013974538073,"ttft_p95_ms":2000.1368144294247,"ttft_p99_ms":2003.0350286723112,"tpot_p50_ms":45.94598221112161,"tpot_p95_ms":46.736607726203374,"tpot_p99_ms":46.843049427122,"e2e_p50_ms":188973.83886645548,"e2e_p95_ms":191972.2936287988,"e2e_p99_ms":192374.7143190843,"itl_p50_ms":39.03631155844778,"itl_p95_ms":63.36234454065562,"itl_p99_ms":68.28304521273822,"validation_errors":[],"timestamp":"2026-07-16T09:06:52+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":4096,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o4096_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":840.8592580729164,"request_tps":0.09514077324110604,"input_tps":97.42415179889258,"output_tps":389.6966071955703,"total_tps":487.1207589944629,"mean_input_tokens":1024.0,"mean_output_tokens":4096.0,"ttft_p50_ms":972.442637081258,"ttft_p95_ms":5176.270531245973,"ttft_p99_ms":5179.259422505274,"tpot_p50_ms":40.55560706117929,"tpot_p95_ms":40.93768200460964,"tpot_p99_ms":40.989130662076434,"e2e_p50_ms":167707.99649693072,"e2e_p95_ms":171209.72607890144,"e2e_p99_ms":171213.10957846232,"itl_p50_ms":38.21910894475877,"itl_p95_ms":56.424582388717674,"itl_p99_ms":61.033719554543495,"validation_errors":[],"timestamp":"2026-07-16T09:31:28+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":4096,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i1024_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i1024_o4096_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":750.264118807856,"request_tps":0.0533145581632754,"input_tps":54.59410755919401,"output_tps":218.37643023677603,"total_tps":272.97053779597,"mean_input_tokens":1024.0,"mean_output_tokens":4096.0,"ttft_p50_ms":457.8002068446949,"ttft_p95_ms":1004.7794182202779,"ttft_p99_ms":1006.1946984101087,"tpot_p50_ms":36.49401828900812,"tpot_p95_ms":36.67125693285014,"tpot_p99_ms":36.679555167656694,"e2e_p50_ms":150253.00148909446,"e2e_p95_ms":150511.7817995837,"e2e_p99_ms":150794.05072218738,"itl_p50_ms":34.35514506418258,"itl_p95_ms":53.22418445721268,"itl_p99_ms":59.324696937110275,"validation_errors":[],"timestamp":"2026-07-16T09:44:20+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":4096,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i1024_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i1024_o4096_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":64.04329815506935,"request_tps":1.2491549046442667,"input_tps":5116.538489422916,"output_tps":159.89182779446614,"total_tps":5276.430317217383,"mean_input_tokens":4096.0,"mean_output_tokens":128.0,"ttft_p50_ms":4576.742604491301,"ttft_p95_ms":7486.383566202128,"ttft_p99_ms":8069.572048275731,"tpot_p50_ms":63.25949257419364,"tpot_p95_ms":88.51012927140879,"tpot_p99_ms":93.78976958789158,"e2e_p50_ms":12597.990370937623,"e2e_p95_ms":16119.662079052066,"e2e_p99_ms":16428.539164408576,"itl_p50_ms":38.49607554730028,"itl_p95_ms":59.04390099458394,"itl_p99_ms":859.1562826023437,"validation_errors":[],"timestamp":"2026-07-16T09:45:45+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":42.330964882159606,"request_tps":0.9449347566574843,"input_tps":3870.4527632690556,"output_tps":120.95164885215799,"total_tps":3991.4044121212137,"mean_input_tokens":4096.0,"mean_output_tokens":128.0,"ttft_p50_ms":2789.9403674528003,"ttft_p95_ms":3850.861389178317,"ttft_p99_ms":3857.0726825506426,"tpot_p50_ms":44.318099374200884,"tpot_p95_ms":52.61084187132357,"tpot_p99_ms":56.91049005684083,"e2e_p50_ms":8397.533455048688,"e2e_p95_ms":10301.826596679166,"e2e_p99_ms":10350.095471963286,"itl_p50_ms":34.42738042213023,"itl_p95_ms":54.41078649600968,"itl_p99_ms":108.43641549814492,"validation_errors":[],"timestamp":"2026-07-16T09:46:49+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":95.2746979680378,"request_tps":0.839677287949398,"input_tps":3439.318171440734,"output_tps":214.9573857150459,"total_tps":3654.27555715578,"mean_input_tokens":4096.0,"mean_output_tokens":256.0,"ttft_p50_ms":4683.939669514075,"ttft_p95_ms":8062.7326244954,"ttft_p99_ms":8673.460278843995,"tpot_p50_ms":51.88298721149491,"tpot_p95_ms":64.92892745028561,"tpot_p99_ms":75.71912046924533,"e2e_p50_ms":17954.677841393277,"e2e_p95_ms":22523.24340294581,"e2e_p99_ms":24458.481649178953,"itl_p50_ms":38.46407646778971,"itl_p95_ms":59.771430026739836,"itl_p99_ms":67.52078937599435,"validation_errors":[],"timestamp":"2026-07-16T09:48:46+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":65.8774013610091,"request_tps":0.6071884921628803,"input_tps":2487.0440638991577,"output_tps":155.44025399369735,"total_tps":2642.4843178928554,"mean_input_tokens":4096.0,"mean_output_tokens":256.0,"ttft_p50_ms":1971.202465472743,"ttft_p95_ms":3824.870833219029,"ttft_p99_ms":3841.111220801249,"tpot_p50_ms":43.596625505654394,"tpot_p95_ms":48.1240840446131,"tpot_p99_ms":49.99360711309218,"e2e_p50_ms":13162.176689947955,"e2e_p95_ms":14974.834044638555,"e2e_p99_ms":14991.584688341245,"itl_p50_ms":34.47401651646942,"itl_p95_ms":54.485111916437745,"itl_p99_ms":60.43242933228612,"validation_errors":[],"timestamp":"2026-07-16T09:50:14+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":160.08099704002962,"request_tps":0.4997470123202401,"input_tps":2046.9637624637035,"output_tps":255.87047030796293,"total_tps":2302.8342327716664,"mean_input_tokens":4096.0,"mean_output_tokens":512.0,"ttft_p50_ms":5474.024421186186,"ttft_p95_ms":8615.011906716973,"ttft_p99_ms":8951.299454767715,"tpot_p50_ms":46.494569923772445,"tpot_p95_ms":53.43938896643479,"tpot_p99_ms":76.29711307971617,"e2e_p50_ms":28335.586922592483,"e2e_p95_ms":32704.220330703534,"e2e_p99_ms":44918.695579937164,"itl_p50_ms":38.48353505600244,"itl_p95_ms":59.288766619283706,"itl_p99_ms":65.03905458841473,"validation_errors":[],"timestamp":"2026-07-16T09:53:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":112.8795186879579,"request_tps":0.3543601218798184,"input_tps":1451.4590592197362,"output_tps":181.43238240246703,"total_tps":1632.8914416222033,"mean_input_tokens":4096.0,"mean_output_tokens":512.0,"ttft_p50_ms":1941.342479432933,"ttft_p95_ms":3813.8611613423564,"ttft_p99_ms":3815.337903073523,"tpot_p50_ms":40.17744154405986,"tpot_p95_ms":42.320974605816815,"tpot_p99_ms":43.39261491248045,"e2e_p50_ms":22431.270663044415,"e2e_p95_ms":24335.910513892304,"e2e_p99_ms":24337.53361347597,"itl_p50_ms":34.43968354258686,"itl_p95_ms":54.208886146079735,"itl_p99_ms":59.69965270254761,"validation_errors":[],"timestamp":"2026-07-16T09:55:31+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":282.66458361805417,"request_tps":0.2830209535839788,"input_tps":1159.253825879977,"output_tps":289.81345646999426,"total_tps":1449.0672823499715,"mean_input_tokens":4096.0,"mean_output_tokens":1024.0,"ttft_p50_ms":4658.318318077363,"ttft_p95_ms":10095.44171601301,"ttft_p99_ms":12803.387446557112,"tpot_p50_ms":44.42285834210774,"tpot_p95_ms":49.33753012539997,"tpot_p99_ms":79.11847314414067,"e2e_p50_ms":49335.88792406954,"e2e_p95_ms":58133.938353240934,"e2e_p99_ms":87081.42475777536,"itl_p50_ms":38.47858391236514,"itl_p95_ms":57.68822985701263,"itl_p99_ms":62.88705667946487,"validation_errors":[],"timestamp":"2026-07-16T10:00:35+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":207.04861099808477,"request_tps":0.1931913467430603,"input_tps":791.311756259575,"output_tps":197.82793906489374,"total_tps":989.1396953244687,"mean_input_tokens":4096.0,"mean_output_tokens":1024.0,"ttft_p50_ms":1074.1113065741956,"ttft_p95_ms":3878.6458508693613,"ttft_p99_ms":3916.904898113571,"tpot_p50_ms":39.24930339234399,"tpot_p95_ms":40.05846651063872,"tpot_p99_ms":40.82051611248493,"e2e_p50_ms":41186.332597979344,"e2e_p95_ms":43346.47120544687,"e2e_p99_ms":43701.94057547487,"itl_p50_ms":34.467250341549516,"itl_p95_ms":54.143763531465076,"itl_p99_ms":59.957590696867555,"validation_errors":[],"timestamp":"2026-07-16T10:04:24+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":534.5376915351953,"request_tps":0.14966203743320614,"input_tps":613.0157053264123,"output_tps":306.50785266320617,"total_tps":919.5235579896186,"mean_input_tokens":4096.0,"mean_output_tokens":2048.0,"ttft_p50_ms":5104.571636882611,"ttft_p95_ms":13598.845852899827,"ttft_p99_ms":15530.32725170254,"tpot_p50_ms":42.516595378821116,"tpot_p95_ms":48.41014748285291,"tpot_p99_ms":80.66643662295293,"e2e_p50_ms":90892.7560900338,"e2e_p95_ms":106289.35107835554,"e2e_p99_ms":171094.83309885245,"itl_p50_ms":38.54017308913171,"itl_p95_ms":58.33887221524492,"itl_p99_ms":64.16293417569251,"validation_errors":[],"timestamp":"2026-07-16T10:13:42+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o2048_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":394.49152314290404,"request_tps":0.10139634859913087,"input_tps":415.31944386204003,"output_tps":207.65972193102002,"total_tps":622.97916579306,"mean_input_tokens":4096.0,"mean_output_tokens":2048.0,"ttft_p50_ms":1034.7788120852783,"ttft_p95_ms":3756.3848297810177,"ttft_p99_ms":3805.5001059989445,"tpot_p50_ms":37.88820180411141,"tpot_p95_ms":38.25126371524308,"tpot_p99_ms":38.27709748982381,"e2e_p50_ms":78674.81265543029,"e2e_p95_ms":80982.15972669423,"e2e_p99_ms":80984.67170314165,"itl_p50_ms":34.52226798981428,"itl_p95_ms":53.908159665297724,"itl_p99_ms":59.665183436591185,"validation_errors":[],"timestamp":"2026-07-16T10:20:38+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":2048,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1029.577203406021,"request_tps":0.07770179811222125,"input_tps":318.26656506765823,"output_tps":318.26656506765823,"total_tps":636.5331301353165,"mean_input_tokens":4096.0,"mean_output_tokens":4096.0,"ttft_p50_ms":5483.361787861213,"ttft_p95_ms":22374.68960867263,"ttft_p99_ms":150920.9240530175,"tpot_p50_ms":41.06715219183478,"tpot_p95_ms":46.9445546033481,"tpot_p99_ms":79.17070015191274,"e2e_p50_ms":173329.3942649616,"e2e_p95_ms":302973.24356382014,"e2e_p99_ms":330154.8701896191,"itl_p50_ms":38.36022887844592,"itl_p95_ms":57.20986550441011,"itl_p99_ms":62.143948464654386,"validation_errors":[],"timestamp":"2026-07-16T10:38:11+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":4096,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i4096_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i4096_o4096_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":762.6022247918881,"request_tps":0.052451984402374224,"input_tps":214.84332811212482,"output_tps":214.84332811212482,"total_tps":429.68665622424965,"mean_input_tokens":4096.0,"mean_output_tokens":4096.0,"ttft_p50_ms":1038.2880391553044,"ttft_p95_ms":3893.24660198763,"ttft_p99_ms":3900.2761626965366,"tpot_p50_ms":36.76516193150564,"tpot_p95_ms":37.05582632169512,"tpot_p99_ms":37.09407625586006,"e2e_p50_ms":152243.55608585756,"e2e_p95_ms":154084.0833232156,"e2e_p99_ms":154084.53153002774,"itl_p50_ms":34.463146002963185,"itl_p95_ms":51.49366667028517,"itl_p99_ms":59.961400236934416,"validation_errors":[],"timestamp":"2026-07-16T10:51:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":4096,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i4096_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i4096_o4096_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":214.8227549989242,"request_tps":0.37240002810875705,"input_tps":6101.402060533876,"output_tps":47.6672035979209,"total_tps":6149.069264131796,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":33368.671427015215,"ttft_p95_ms":41899.411271174904,"ttft_p99_ms":42125.512621703085,"tpot_p50_ms":52.554487114163535,"tpot_p95_ms":84.39758658330287,"tpot_p99_ms":111.94620711651932,"e2e_p50_ms":42495.81566103734,"e2e_p95_ms":49210.346029268105,"e2e_p99_ms":52561.92248010076,"itl_p50_ms":34.52568908687681,"itl_p95_ms":52.17893534572764,"itl_p99_ms":66.4443611598108,"validation_errors":[],"timestamp":"2026-07-16T10:55:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":109.44023031182587,"request_tps":0.36549630685195744,"input_tps":5988.291491462471,"output_tps":46.78352727705055,"total_tps":6035.075018739521,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":11529.161532991566,"ttft_p95_ms":21725.308789138217,"ttft_p99_ms":21941.029288116843,"tpot_p50_ms":76.67213628768891,"tpot_p95_ms":101.16250104450535,"tpot_p99_ms":138.8094991766004,"e2e_p50_ms":16583.38130300399,"e2e_p95_ms":32255.35403857939,"e2e_p99_ms":37416.11094208899,"itl_p50_ms":34.63044541422278,"itl_p95_ms":54.926601552870125,"itl_p99_ms":1767.2955490718596,"validation_errors":[],"timestamp":"2026-07-16T10:57:31+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":39.83333700802177,"request_tps":0.12552300097260452,"input_tps":2056.5688479351525,"output_tps":16.06694412449338,"total_tps":2072.6357920596456,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":3952.2010260261595,"ttft_p95_ms":4079.302547173574,"ttft_p99_ms":4102.609633402899,"tpot_p50_ms":31.602383598567936,"tpot_p95_ms":31.640485883140425,"tpot_p99_ms":31.642765031291507,"e2e_p50_ms":7969.095474109054,"e2e_p95_ms":8074.662221223116,"e2e_p99_ms":8093.300537876784,"itl_p50_ms":29.8505169339478,"itl_p95_ms":49.705708399415016,"itl_p99_ms":51.2071222346276,"validation_errors":[],"timestamp":"2026-07-16T10:58:35+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":270.44398706802167,"request_tps":0.2958098675711304,"input_tps":4846.548870285401,"output_tps":75.72732609820939,"total_tps":4922.27619638361,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":38130.55603799876,"ttft_p95_ms":51039.480149233714,"ttft_p99_ms":51170.22580615245,"tpot_p50_ms":44.34177505135463,"tpot_p95_ms":60.19466479852174,"tpot_p99_ms":74.01323430483028,"e2e_p50_ms":50331.59309858456,"e2e_p95_ms":65593.75085806241,"e2e_p99_ms":65883.99838647107,"itl_p50_ms":34.52300652861595,"itl_p95_ms":51.78628868889062,"itl_p99_ms":60.336364856921136,"validation_errors":[],"timestamp":"2026-07-16T11:03:31+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":141.670446844073,"request_tps":0.28234540718308926,"input_tps":4625.947151287734,"output_tps":72.28042423887085,"total_tps":4698.2275755266055,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":11554.360230453312,"ttft_p95_ms":26391.11408374738,"ttft_p99_ms":26587.14834881015,"tpot_p50_ms":56.31336195065695,"tpot_p95_ms":68.25714148759073,"tpot_p99_ms":87.6372731319245,"e2e_p50_ms":21265.938312513754,"e2e_p95_ms":41577.943935687654,"e2e_p99_ms":46795.674697651986,"itl_p50_ms":34.58584542386234,"itl_p95_ms":52.59093521162865,"itl_p99_ms":61.120713129639775,"validation_errors":[],"timestamp":"2026-07-16T11:06:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":60.537773119984195,"request_tps":0.08259306119652168,"input_tps":1353.2047146438113,"output_tps":21.14382366630955,"total_tps":1374.3485383101208,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":3945.7851310726255,"ttft_p95_ms":4093.810023879632,"ttft_p99_ms":4111.218916652724,"tpot_p50_ms":31.821994341033346,"tpot_p95_ms":32.01389185737307,"tpot_p99_ms":32.03676525156433,"e2e_p50_ms":12080.163977108896,"e2e_p95_ms":12253.15336487256,"e2e_p99_ms":12279.754239274189,"itl_p50_ms":29.860365903005004,"itl_p95_ms":48.98134677205235,"itl_p99_ms":55.7887158729136,"validation_errors":[],"timestamp":"2026-07-16T11:07:42+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":256,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":390.9635538929142,"request_tps":0.20462265396204216,"input_tps":3352.5375625140987,"output_tps":104.76679882856558,"total_tps":3457.304361342664,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":52149.89124494605,"ttft_p95_ms":69626.60835890565,"ttft_p99_ms":69753.87210340472,"tpot_p50_ms":40.293584739718284,"tpot_p95_ms":48.141768489123194,"tpot_p99_ms":54.960999020335215,"e2e_p50_ms":73592.7937864326,"e2e_p95_ms":93283.15660590306,"e2e_p99_ms":93629.28111733867,"itl_p50_ms":34.522912465035915,"itl_p95_ms":51.489772961940616,"itl_p99_ms":59.926832308992736,"validation_errors":[],"timestamp":"2026-07-16T11:14:38+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":206.38675206596963,"request_tps":0.19381088950522546,"input_tps":3175.397613653614,"output_tps":99.23117542667543,"total_tps":3274.628789080289,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":11589.022769010626,"ttft_p95_ms":35687.3680736986,"ttft_p99_ms":35902.96996307094,"tpot_p50_ms":46.31837186974942,"tpot_p95_ms":52.744000344951814,"tpot_p99_ms":61.98560190418053,"e2e_p50_ms":30626.526765991002,"e2e_p95_ms":60265.1604500832,"e2e_p99_ms":65432.796396943726,"itl_p50_ms":34.61834159679711,"itl_p95_ms":51.886969921179116,"itl_p99_ms":60.28145674150437,"validation_errors":[],"timestamp":"2026-07-16T11:18:29+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":100.54988616588525,"request_tps":0.04972656052291394,"input_tps":814.719967607422,"output_tps":25.459998987731936,"total_tps":840.1799665951539,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":3932.124942075461,"ttft_p95_ms":4127.591159660369,"ttft_p99_ms":4166.129624564201,"tpot_p50_ms":31.635086914185443,"tpot_p95_ms":31.776042928674975,"tpot_p99_ms":31.788027078066378,"e2e_p50_ms":20141.837269999087,"e2e_p95_ms":20233.498825039715,"e2e_p99_ms":20244.73103923723,"itl_p50_ms":29.790908098220825,"itl_p95_ms":50.02487420570105,"itl_p99_ms":54.69925904646516,"validation_errors":[],"timestamp":"2026-07-16T11:20:34+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":512,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":632.0467019469943,"request_tps":0.1265729253132138,"input_tps":2073.7708083316948,"output_tps":129.61067552073092,"total_tps":2203.3814838524254,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":80126.8644919619,"ttft_p95_ms":106781.9698160165,"ttft_p99_ms":107180.66528041381,"tpot_p50_ms":38.308278328050854,"tpot_p95_ms":42.30265493578899,"tpot_p99_ms":45.4902937474118,"e2e_p50_ms":120186.66901800316,"e2e_p95_ms":148021.29719536752,"e2e_p99_ms":148871.62709633118,"itl_p50_ms":34.52094295062125,"itl_p95_ms":51.694627082906685,"itl_p99_ms":59.966907631605864,"validation_errors":[],"timestamp":"2026-07-16T11:31:32+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":335.39274771884084,"request_tps":0.11926316317826863,"input_tps":1954.0076655127532,"output_tps":122.12547909454707,"total_tps":2076.1331446073004,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":11495.135091943666,"ttft_p95_ms":54293.259906279854,"ttft_p99_ms":54475.75780400075,"tpot_p50_ms":41.142632146611746,"tpot_p95_ms":44.4825864581911,"tpot_p99_ms":48.84856583131568,"e2e_p50_ms":49175.88768794667,"e2e_p95_ms":97491.41289745457,"e2e_p99_ms":102364.25640797475,"itl_p50_ms":34.5357779879123,"itl_p95_ms":52.090133994352065,"itl_p99_ms":60.2705589751713,"validation_errors":[],"timestamp":"2026-07-16T11:37:32+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":181.57242772518657,"request_tps":0.02753722061571814,"input_tps":451.169822567926,"output_tps":28.198113910495376,"total_tps":479.3679364784214,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":3940.7668448984623,"ttft_p95_ms":4193.481011828408,"ttft_p99_ms":4228.460800005123,"tpot_p50_ms":31.63807294540879,"tpot_p95_ms":31.730153837458815,"tpot_p99_ms":31.734481836830287,"e2e_p50_ms":36320.72346494533,"e2e_p95_ms":36463.93566392362,"e2e_p99_ms":36480.87247610092,"itl_p50_ms":29.840212780982256,"itl_p95_ms":46.52133162599058,"itl_p99_ms":54.09107577055689,"validation_errors":[],"timestamp":"2026-07-16T11:40:58+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":1024,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1115.2481476869434,"request_tps":0.07173291447820136,"input_tps":1175.2720708108511,"output_tps":146.9090088513564,"total_tps":1322.1810796622076,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":136037.42311848328,"ttft_p95_ms":181299.1330652847,"ttft_p99_ms":181444.37291203765,"tpot_p50_ms":37.30771111527574,"tpot_p95_ms":39.328867612731,"tpot_p99_ms":40.83525040460833,"e2e_p50_ms":213226.2396500446,"e2e_p95_ms":259703.93927409317,"e2e_p99_ms":260337.14725418712,"itl_p50_ms":34.49960285797715,"itl_p95_ms":51.94193584611639,"itl_p99_ms":60.15488663455472,"validation_errors":[],"timestamp":"2026-07-16T12:00:00+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":584.8935852469876,"request_tps":0.06838850862607578,"input_tps":1120.4773253296255,"output_tps":140.0596656662032,"total_tps":1260.5369909958285,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":11495.196483563632,"ttft_p95_ms":91582.56554905092,"ttft_p99_ms":91693.05469256826,"tpot_p50_ms":38.41638142817654,"tpot_p95_ms":41.38107354754197,"tpot_p99_ms":42.587399352946996,"e2e_p50_ms":89565.8936999971,"e2e_p95_ms":172048.6941325711,"e2e_p99_ms":176627.42560487473,"itl_p50_ms":34.59935949649662,"itl_p95_ms":51.98211170500145,"itl_p99_ms":60.23820979986343,"validation_errors":[],"timestamp":"2026-07-16T12:10:09+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":2048,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":342.39226463483647,"request_tps":0.014603133646528293,"input_tps":239.25774166471956,"output_tps":29.907217708089945,"total_tps":269.1649593728095,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":3971.6174241621047,"ttft_p95_ms":4094.661511434242,"ttft_p99_ms":4115.855502327904,"tpot_p50_ms":31.558119324293,"tpot_p95_ms":31.620720006947725,"tpot_p99_ms":31.632591595734013,"e2e_p50_ms":68487.32287203893,"e2e_p95_ms":68711.56300739385,"e2e_p99_ms":68739.65807267465,"itl_p50_ms":29.78591388091445,"itl_p95_ms":45.330130099318914,"itl_p99_ms":55.03684663679451,"validation_errors":[],"timestamp":"2026-07-16T12:16:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":2048,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":765.7493072859943,"request_tps":0.1044728336530135,"input_tps":6846.731626283893,"output_tps":13.372522707585729,"total_tps":6860.104148991479,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":128567.64044612646,"ttft_p95_ms":151961.6587853641,"ttft_p99_ms":156960.82069837945,"tpot_p50_ms":161.96727417372753,"tpot_p95_ms":305.1854889790388,"tpot_p99_ms":437.706828691014,"e2e_p50_ms":146586.26923209522,"e2e_p95_ms":179067.64789508888,"e2e_p99_ms":186050.76994073341,"itl_p50_ms":36.512586986646056,"itl_p95_ms":53.31773438956598,"itl_p99_ms":61.62503688596189,"validation_errors":[],"timestamp":"2026-07-16T12:29:43+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":386.69948341907,"request_tps":0.10343949685769714,"input_tps":6779.01086606604,"output_tps":13.240255597785234,"total_tps":6792.251121663825,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":52697.72171450313,"ttft_p95_ms":77315.9501316608,"ttft_p99_ms":77702.88953800453,"tpot_p50_ms":161.86251543282611,"tpot_p95_ms":301.6361779644242,"tpot_p99_ms":390.316338291553,"e2e_p50_ms":73942.72513350006,"e2e_p95_ms":98705.0266831764,"e2e_p99_ms":108562.47204806423,"itl_p50_ms":36.458888091146946,"itl_p95_ms":54.5758139109239,"itl_p99_ms":62.127575105987496,"validation_errors":[],"timestamp":"2026-07-16T12:36:49+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":109.37776633584872,"request_tps":0.045713129528055126,"input_tps":2995.8556567506207,"output_tps":5.851280579591056,"total_tps":3001.706937330212,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":17729.769844096154,"ttft_p95_ms":17954.093989543617,"ttft_p99_ms":17964.415912255645,"tpot_p50_ms":32.0961656301891,"tpot_p95_ms":32.58936951105578,"tpot_p99_ms":32.640932479609305,"e2e_p50_ms":21835.877286968753,"e2e_p95_ms":22044.52510732226,"e2e_p99_ms":22048.520961580798,"itl_p50_ms":30.422291019931436,"itl_p95_ms":47.328294953331344,"itl_p99_ms":56.53701915405691,"validation_errors":[],"timestamp":"2026-07-16T12:39:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":832.5039105331525,"request_tps":0.09609564470245716,"input_tps":6297.724171220232,"output_tps":24.600485043829032,"total_tps":6322.324656264061,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":136335.5990794953,"ttft_p95_ms":162983.08086185716,"ttft_p99_ms":166663.45600951457,"tpot_p50_ms":100.12674616266263,"tpot_p95_ms":175.98028106304506,"tpot_p99_ms":236.80305566614055,"e2e_p50_ms":156982.1368450066,"e2e_p95_ms":189405.96394817112,"e2e_p99_ms":206104.3898640247,"itl_p50_ms":36.563907051458955,"itl_p95_ms":53.42142162844535,"itl_p99_ms":62.24962119013072,"validation_errors":[],"timestamp":"2026-07-16T12:53:50+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":419.1958255460486,"request_tps":0.09542079754228375,"input_tps":6253.497387731108,"output_tps":24.42772417082464,"total_tps":6277.925111901933,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":52760.39652095642,"ttft_p95_ms":81073.16757803783,"ttft_p99_ms":81568.98809597129,"tpot_p50_ms":99.82632672934629,"tpot_p95_ms":169.0839746892087,"tpot_p99_ms":210.72348307306856,"e2e_p50_ms":80142.28945260402,"e2e_p95_ms":107118.48287502071,"e2e_p99_ms":117810.70945826358,"itl_p50_ms":36.5109599661082,"itl_p95_ms":53.218919737264514,"itl_p99_ms":61.9426407897845,"validation_errors":[],"timestamp":"2026-07-16T13:01:28+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":129.80575013905764,"request_tps":0.038519094837044014,"input_tps":2524.3873992405165,"output_tps":9.860888278283268,"total_tps":2534.2482875187998,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":17821.7026530765,"ttft_p95_ms":17941.041450900957,"ttft_p99_ms":17956.14996048622,"tpot_p50_ms":32.05825664723913,"tpot_p95_ms":32.285422013122954,"tpot_p99_ms":32.28693708328201,"e2e_p50_ms":26052.55355197005,"e2e_p95_ms":26127.57891532965,"e2e_p99_ms":26133.34180940874,"itl_p50_ms":30.41153005324304,"itl_p95_ms":46.34950009640306,"itl_p99_ms":56.67572288773954,"validation_errors":[],"timestamp":"2026-07-16T13:04:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":256,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":973.7963511580601,"request_tps":0.08215270051573127,"input_tps":5383.959380998965,"output_tps":42.06218266405441,"total_tps":5426.021563663019,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":153779.00782134384,"ttft_p95_ms":181627.71489518927,"ttft_p99_ms":183322.4648270011,"tpot_p50_ms":69.34998192962937,"tpot_p95_ms":106.79924657778838,"tpot_p99_ms":136.58104994009412,"e2e_p50_ms":189005.7152614463,"e2e_p95_ms":234631.01633958286,"e2e_p99_ms":236243.1676932983,"itl_p50_ms":36.63114202208817,"itl_p95_ms":53.81391057744622,"itl_p99_ms":62.73157489486038,"validation_errors":[],"timestamp":"2026-07-16T13:21:10+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":487.3736931588501,"request_tps":0.08207254630578259,"input_tps":5378.706394695768,"output_tps":42.021143708560686,"total_tps":5420.727538404329,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":52526.89733938314,"ttft_p95_ms":84162.39211051725,"ttft_p99_ms":89592.60615824256,"tpot_p50_ms":77.38403851756873,"tpot_p95_ms":114.1166405163389,"tpot_p99_ms":167.8516751349313,"e2e_p50_ms":79288.78492198419,"e2e_p95_ms":137056.76206527278,"e2e_p99_ms":142416.35865559801,"itl_p50_ms":36.64115187712014,"itl_p95_ms":55.09973680600524,"itl_p99_ms":62.820788719691336,"validation_errors":[],"timestamp":"2026-07-16T13:29:57+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":170.8164885980077,"request_tps":0.029271178918604212,"input_tps":1918.3159816096456,"output_tps":14.986843606325357,"total_tps":1933.302825215971,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":17657.30693913065,"ttft_p95_ms":17947.263365006074,"ttft_p99_ms":17974.353000214323,"tpot_p50_ms":32.16144313886325,"tpot_p95_ms":32.28488007000849,"tpot_p99_ms":32.28917933327971,"e2e_p50_ms":34091.804383089766,"e2e_p95_ms":34388.406378310174,"e2e_p99_ms":34406.406796555966,"itl_p50_ms":30.36797302775085,"itl_p95_ms":46.908468334004276,"itl_p99_ms":55.469688819721355,"validation_errors":[],"timestamp":"2026-07-16T13:33:25+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":512,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1253.2045440499205,"request_tps":0.0638363468915201,"input_tps":4183.578829882661,"output_tps":65.36841921691658,"total_tps":4248.947249099578,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":202414.35351443943,"ttft_p95_ms":223751.87583953375,"ttft_p99_ms":226827.1145781991,"tpot_p50_ms":53.94717318803356,"tpot_p95_ms":71.33589557350184,"tpot_p99_ms":86.74920972760991,"e2e_p50_ms":247910.9200029634,"e2e_p95_ms":278792.46321712853,"e2e_p99_ms":292009.2727964232,"itl_p50_ms":36.60361119545996,"itl_p95_ms":54.5408145524561,"itl_p99_ms":62.74137625936419,"validation_errors":[],"timestamp":"2026-07-16T13:54:59+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i65536_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":626.4431783319451,"request_tps":0.06385255899267604,"input_tps":4184.641306144017,"output_tps":65.38502040850027,"total_tps":4250.026326552517,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":51137.51745200716,"ttft_p95_ms":96054.72948674572,"ttft_p99_ms":112415.82493769703,"tpot_p50_ms":70.99086488608475,"tpot_p95_ms":93.18777094192289,"tpot_p99_ms":94.39113497935946,"e2e_p50_ms":112995.05897297058,"e2e_p95_ms":168512.40640780888,"e2e_p99_ms":178801.59651178867,"itl_p50_ms":36.47732094395906,"itl_p95_ms":61.783290666062385,"itl_p99_ms":68.58176962938157,"validation_errors":[],"timestamp":"2026-07-16T14:06:05+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i65536_o1024_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":256.18154512299225,"request_tps":0.019517409021791596,"input_tps":1279.092917652134,"output_tps":19.985826838314594,"total_tps":1299.0787444904488,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":17658.26961491257,"ttft_p95_ms":18013.91982375644,"ttft_p99_ms":18047.10275827907,"tpot_p50_ms":32.7376385434701,"tpot_p95_ms":32.79802809666955,"tpot_p99_ms":32.798152351014515,"e2e_p50_ms":51210.01679683104,"e2e_p95_ms":51457.70765142515,"e2e_p99_ms":51481.52730548754,"itl_p50_ms":30.361955054104328,"itl_p95_ms":53.69571931660175,"itl_p99_ms":60.88819689117371,"validation_errors":[],"timestamp":"2026-07-16T14:11:00+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":1024,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i65536_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1677.9406901570037,"request_tps":0.047677489716584955,"input_tps":6249.183932132223,"output_tps":6.102718683722874,"total_tps":6255.286650815946,"mean_input_tokens":131072.0,"mean_output_tokens":128.0,"ttft_p50_ms":300625.9035965195,"ttft_p95_ms":338951.3012407697,"ttft_p99_ms":340203.0271626217,"tpot_p50_ms":41.211593476322115,"tpot_p95_ms":345.65387451581245,"tpot_p99_ms":346.94827124932925,"e2e_p50_ms":343158.3466755692,"e2e_p95_ms":344526.5441220952,"e2e_p99_ms":345839.6717084944,"itl_p50_ms":33.60768407583237,"itl_p95_ms":58.138947119005024,"itl_p99_ms":181.5207198727876,"validation_errors":[],"timestamp":"2026-07-16T14:40:04+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i131072_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i131072_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":818.1570529809687,"request_tps":0.04889036873086816,"input_tps":6408.158410292352,"output_tps":6.257967197551125,"total_tps":6414.4163774899025,"mean_input_tokens":131072.0,"mean_output_tokens":128.0,"ttft_p50_ms":129584.32166394778,"ttft_p95_ms":167181.3310468453,"ttft_p99_ms":168040.67650724668,"tpot_p50_ms":36.84465442857463,"tpot_p95_ms":343.3015784548456,"tpot_p99_ms":345.92269529082523,"e2e_p50_ms":171526.36588714086,"e2e_p95_ms":173146.40125581063,"e2e_p99_ms":173291.67424835498,"itl_p50_ms":33.540375996381044,"itl_p95_ms":58.00506006926298,"itl_p99_ms":276.8040000228211,"validation_errors":[],"timestamp":"2026-07-16T14:54:46+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c8_i131072_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c8_i131072_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":226.54693996487185,"request_tps":0.022070481290876385,"input_tps":2892.8221237577495,"output_tps":2.8250216052321773,"total_tps":2895.647145362982,"mean_input_tokens":131072.0,"mean_output_tokens":128.0,"ttft_p50_ms":40983.879873063415,"ttft_p95_ms":41427.96418489888,"ttft_p99_ms":41483.68903929368,"tpot_p50_ms":33.28039894415814,"tpot_p95_ms":33.693099089744166,"tpot_p99_ms":33.698522558318466,"e2e_p50_ms":45259.45955491625,"e2e_p95_ms":45639.66128444299,"e2e_p99_ms":45701.79939834401,"itl_p50_ms":30.720365000888705,"itl_p95_ms":55.317965033464134,"itl_p99_ms":62.53650498110801,"validation_errors":[],"timestamp":"2026-07-16T14:59:34+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c1_i131072_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c1_i131072_o128_a1.log"} +{"timestamp":"2026-07-16T16:03:27+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":2,"status":"FAILED","completed":0,"failed":80,"error_type":"benchmark rc=124","validation_errors":[],"gain_pct":null,"plateau_streak":0,"raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i131072_o256_a2.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i131072_o256_a2.log"} +{"timestamp":"2026-07-16T17:11:50+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":2,"status":"FAILED","completed":0,"failed":80,"error_type":"benchmark rc=124","validation_errors":[],"gain_pct":null,"plateau_streak":0,"raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/raw_outputs/sglang_adaptive_c16_i131072_o512_a2.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/logs/sglang_c16_i131072_o512_a2.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":40.91812721895985,"request_tps":1.9551236930250107,"input_tps":2002.046661657611,"output_tps":250.25583270720136,"total_tps":2252.3024943648124,"mean_input_tokens":1024.0,"mean_output_tokens":128.0,"ttft_p50_ms":2898.4853614820167,"ttft_p95_ms":5916.232369153295,"ttft_p99_ms":5917.164536933415,"tpot_p50_ms":36.42805760238671,"tpot_p95_ms":45.71566509713573,"tpot_p99_ms":52.58390468757908,"e2e_p50_ms":7513.254255056381,"e2e_p95_ms":10505.285932659172,"e2e_p99_ms":10506.672543857712,"itl_p50_ms":33.98504899814725,"itl_p95_ms":55.32407406717539,"itl_p99_ms":61.596665140241385,"validation_errors":[],"timestamp":"2026-07-16T17:22:33+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":23.175778079079464,"request_tps":1.7259398956752865,"input_tps":1767.3624531714934,"output_tps":220.92030664643667,"total_tps":1988.28275981793,"mean_input_tokens":1024.0,"mean_output_tokens":128.0,"ttft_p50_ms":586.4921135362238,"ttft_p95_ms":622.7808163734153,"ttft_p99_ms":623.125078517478,"tpot_p50_ms":31.85279993340373,"tpot_p95_ms":32.22969000018752,"tpot_p99_ms":32.44051498659762,"e2e_p50_ms":4641.481174970977,"e2e_p95_ms":4670.240912691224,"e2e_p99_ms":4670.475004990585,"itl_p50_ms":29.695762554183602,"itl_p95_ms":49.66246752301232,"itl_p99_ms":57.65656981850043,"validation_errors":[],"timestamp":"2026-07-16T17:23:16+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i1024_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i1024_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":49.60790970898233,"request_tps":1.6126460572378174,"input_tps":1651.349562611525,"output_tps":412.83739065288125,"total_tps":2064.1869532644064,"mean_input_tokens":1024.0,"mean_output_tokens":256.0,"ttft_p50_ms":935.7815920375288,"ttft_p95_ms":987.9027201328427,"ttft_p99_ms":988.8454519142397,"tpot_p50_ms":35.24840932302907,"tpot_p95_ms":35.34214854765428,"tpot_p99_ms":35.575809728372484,"e2e_p50_ms":9909.394073998556,"e2e_p95_ms":9992.725722829346,"e2e_p99_ms":9993.638517349027,"itl_p50_ms":33.14175654668361,"itl_p95_ms":51.57231901539489,"itl_p99_ms":59.3762440350838,"validation_errors":[],"timestamp":"2026-07-16T17:24:25+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o256_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":72.94000257505104,"request_tps":2.1935836900385253,"input_tps":2246.22969859945,"output_tps":561.5574246498625,"total_tps":2807.7871232493126,"mean_input_tokens":1024.0,"mean_output_tokens":256.0,"ttft_p50_ms":2986.9613859336823,"ttft_p95_ms":5638.966712448744,"ttft_p99_ms":5717.785984254442,"tpot_p50_ms":43.25931125649196,"tpot_p95_ms":52.76475823822194,"tpot_p99_ms":60.14857301456559,"e2e_p50_ms":14733.08369051665,"e2e_p95_ms":16383.586864836978,"e2e_p99_ms":16391.066697821952,"itl_p50_ms":40.04870797507465,"itl_p95_ms":61.18088683579117,"itl_p99_ms":67.89261806756254,"validation_errors":[],"timestamp":"2026-07-16T17:25:59+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":256,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":36.023877,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c32_i1024_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c32_i1024_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":94.76783368107863,"request_tps":0.8441682888861142,"input_tps":864.4283278193809,"output_tps":432.21416390969046,"total_tps":1296.6424917290713,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":947.5187105126679,"ttft_p95_ms":1040.429971262347,"ttft_p99_ms":1041.2525061913766,"tpot_p50_ms":35.295242745540456,"tpot_p95_ms":35.35704588877044,"tpot_p99_ms":35.459834945952224,"e2e_p50_ms":18922.931105131283,"e2e_p95_ms":19052.50374833122,"e2e_p99_ms":19055.464482815005,"itl_p50_ms":33.14921457786113,"itl_p95_ms":51.78932250710204,"itl_p99_ms":59.60216530133038,"validation_errors":[],"timestamp":"2026-07-16T17:27:54+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":116.47414470394142,"request_tps":1.3736954274847377,"input_tps":1406.6641177443714,"output_tps":703.3320588721857,"total_tps":2109.996176616557,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":1728.7794339936227,"ttft_p95_ms":1780.919153173454,"ttft_p99_ms":1782.5818243529648,"tpot_p50_ms":42.263241610289306,"tpot_p95_ms":42.444828835041704,"tpot_p99_ms":42.508655317581514,"e2e_p50_ms":23282.43359492626,"e2e_p95_ms":23419.968224375043,"e2e_p99_ms":23422.798421205953,"itl_p50_ms":40.06265802308917,"itl_p95_ms":58.03677138173953,"itl_p99_ms":67.34221158549191,"validation_errors":[],"timestamp":"2026-07-16T17:30:11+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":512,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":62.727675,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c32_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c32_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":240,"failed":0,"duration_s":538.8399277571589,"request_tps":0.44540129199215867,"input_tps":456.0909229999705,"output_tps":228.04546149998524,"total_tps":684.1363844999557,"mean_input_tokens":1024.0,"mean_output_tokens":512.0,"ttft_p50_ms":2837.4407634837553,"ttft_p95_ms":8041.054585680831,"ttft_p99_ms":8290.32544139307,"tpot_p50_ms":202.26081837298543,"tpot_p95_ms":209.53028491557413,"tpot_p99_ms":212.22897237452446,"e2e_p50_ms":106162.95209643431,"e2e_p95_ms":111388.48089557141,"e2e_p99_ms":111393.2417512266,"itl_p50_ms":197.14024697896093,"itl_p95_ms":226.85785598587245,"itl_p99_ms":233.4809394343756,"validation_errors":[],"timestamp":"2026-07-16T17:39:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":512,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":1,"gain_pct":-67.576416,"plateau_streak":1,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c48_i1024_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c48_i1024_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":195.86190408095717,"request_tps":0.40845104807585736,"input_tps":418.25387322967794,"output_tps":418.25387322967794,"total_tps":836.5077464593559,"mean_input_tokens":1024.0,"mean_output_tokens":1024.0,"ttft_p50_ms":2806.6968551138416,"ttft_p95_ms":2903.0821086023934,"ttft_p99_ms":2903.6138558760285,"tpot_p50_ms":35.50938803068511,"tpot_p95_ms":36.587063905092016,"tpot_p99_ms":36.846195016498285,"e2e_p50_ms":39140.523318434134,"e2e_p95_ms":39359.496829425916,"e2e_p99_ms":39362.11585983634,"itl_p50_ms":33.30066951457411,"itl_p95_ms":52.0355909015052,"itl_p99_ms":61.44081162754446,"validation_errors":[],"timestamp":"2026-07-16T17:43:08+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o1024_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":235.65979528496973,"request_tps":0.6789448314954245,"input_tps":695.2395074513147,"output_tps":695.2395074513147,"total_tps":1390.4790149026294,"mean_input_tokens":1024.0,"mean_output_tokens":1024.0,"ttft_p50_ms":2981.6380555275828,"ttft_p95_ms":5691.65476138005,"ttft_p99_ms":5695.868596371729,"tpot_p50_ms":42.80786131479516,"tpot_p95_ms":45.161864508836935,"tpot_p99_ms":45.66859811666182,"e2e_p50_ms":47101.06354195159,"e2e_p95_ms":49224.63555838913,"e2e_p99_ms":49227.704358948395,"itl_p50_ms":40.25210742838681,"itl_p95_ms":59.54401966882868,"itl_p99_ms":67.25824875058606,"validation_errors":[],"timestamp":"2026-07-16T17:47:25+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":1024,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":66.224284,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c32_i1024_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c32_i1024_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":375.86731938296,"request_tps":0.21284106352031737,"input_tps":217.949249044805,"output_tps":435.89849808961,"total_tps":653.8477471344149,"mean_input_tokens":1024.0,"mean_output_tokens":2048.0,"ttft_p50_ms":1012.9853350808844,"ttft_p95_ms":1060.0732530583628,"ttft_p99_ms":1060.6082381820306,"tpot_p50_ms":36.26880324916558,"tpot_p95_ms":36.36031736202677,"tpot_p99_ms":36.50945045857321,"e2e_p50_ms":75173.97456045728,"e2e_p95_ms":75301.6543291742,"e2e_p99_ms":75354.01330835419,"itl_p50_ms":34.3891941010952,"itl_p95_ms":56.740176025778055,"itl_p99_ms":61.57854507677257,"validation_errors":[],"timestamp":"2026-07-16T17:54:03+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o2048_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":455.1479200378526,"request_tps":0.35153406827981004,"input_tps":359.9708859185255,"output_tps":719.941771837051,"total_tps":1079.9126577555764,"mean_input_tokens":1024.0,"mean_output_tokens":2048.0,"ttft_p50_ms":2997.489857953042,"ttft_p95_ms":4408.118210406974,"ttft_p99_ms":4409.418380276766,"tpot_p50_ms":43.03901399925722,"tpot_p95_ms":43.72752447672071,"tpot_p99_ms":43.95574388355626,"e2e_p50_ms":91594.98853096738,"e2e_p95_ms":92511.16100725485,"e2e_p99_ms":92513.411904776,"itl_p50_ms":40.91528104618192,"itl_p95_ms":61.7140918620862,"itl_p99_ms":67.57471586111933,"validation_errors":[],"timestamp":"2026-07-16T18:01:59+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":2048,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":65.16271,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c32_i1024_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c32_i1024_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":751.3376580928452,"request_tps":0.1064767606658605,"input_tps":109.03220292184115,"output_tps":436.1288116873646,"total_tps":545.1610146092057,"mean_input_tokens":1024.0,"mean_output_tokens":4096.0,"ttft_p50_ms":1016.063955030404,"ttft_p95_ms":1555.9732719440944,"ttft_p99_ms":1557.605766323395,"tpot_p50_ms":36.36836581573491,"tpot_p95_ms":36.73897300841474,"tpot_p99_ms":36.74271235461387,"e2e_p50_ms":149854.72987149842,"e2e_p95_ms":151522.32534443028,"e2e_p99_ms":151525.3242665343,"itl_p50_ms":34.41172291059047,"itl_p95_ms":56.844778708182275,"itl_p99_ms":61.78853524383158,"validation_errors":[],"timestamp":"2026-07-16T18:14:52+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":4096,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i1024_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i1024_o4096_a1.log"} +{"status":"COMPLETED","completed":160,"failed":0,"duration_s":898.5537388259545,"request_tps":0.17806391881364284,"input_tps":182.33745286517026,"output_tps":729.349811460681,"total_tps":911.6872643258514,"mean_input_tokens":1024.0,"mean_output_tokens":4096.0,"ttft_p50_ms":3016.2806854350492,"ttft_p95_ms":3879.865962965414,"ttft_p99_ms":3881.218241781462,"tpot_p50_ms":43.30877847119516,"tpot_p95_ms":43.5328147706794,"tpot_p99_ms":43.585850752513146,"e2e_p50_ms":179637.5885854941,"e2e_p95_ms":181239.3637237721,"e2e_p99_ms":181245.79159579705,"itl_p50_ms":40.978090022690594,"itl_p95_ms":61.92717199446633,"itl_p99_ms":67.58996407035738,"validation_errors":[],"timestamp":"2026-07-16T18:30:14+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":4096,"concurrency":32,"num_prompts":160,"warmup_requests":32,"attempt":1,"gain_pct":67.23266,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c32_i1024_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c32_i1024_o4096_a1.log"} +{"timestamp":"2026-07-16T19:35:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":4096,"concurrency":48,"num_prompts":240,"warmup_requests":48,"attempt":2,"status":"FAILED","completed":0,"failed":240,"error_type":"benchmark rc=124","validation_errors":[],"gain_pct":null,"plateau_streak":0,"raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c48_i1024_o4096_a2.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c48_i1024_o4096_a2.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":81.84067250299267,"request_tps":0.9775090740740753,"input_tps":4003.8771674074123,"output_tps":125.12116148148164,"total_tps":4128.998328888893,"mean_input_tokens":4096.0,"mean_output_tokens":128.0,"ttft_p50_ms":7960.9383844072,"ttft_p95_ms":11711.344682960771,"ttft_p99_ms":11753.230824810453,"tpot_p50_ms":66.47893063738708,"tpot_p95_ms":110.99464661427078,"tpot_p99_ms":112.9407165098686,"e2e_p50_ms":16375.667929998599,"e2e_p95_ms":16515.728015080094,"e2e_p99_ms":16517.64271946624,"itl_p50_ms":34.67924799770117,"itl_p95_ms":57.22848706645891,"itl_p99_ms":62.60195288574323,"validation_errors":[],"timestamp":"2026-07-16T19:42:19+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":49.58700227085501,"request_tps":0.8066629997415711,"input_tps":3304.091646941475,"output_tps":103.2528639669211,"total_tps":3407.3445109083964,"mean_input_tokens":4096.0,"mean_output_tokens":128.0,"ttft_p50_ms":5025.09073051624,"ttft_p95_ms":5915.067831950728,"ttft_p99_ms":5934.279049527831,"tpot_p50_ms":39.28249386619542,"tpot_p95_ms":62.01645719083613,"tpot_p99_ms":62.2953052225332,"e2e_p50_ms":10006.435959017836,"e2e_p95_ms":10091.394168324769,"e2e_p99_ms":10092.632040372118,"itl_p50_ms":30.577882542274892,"itl_p95_ms":49.62789301062003,"itl_p99_ms":57.752098962664604,"validation_errors":[],"timestamp":"2026-07-16T19:43:30+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":18.347577747888863,"request_tps":0.2725155368574643,"input_tps":1116.2236389681739,"output_tps":34.88198871775543,"total_tps":1151.1056276859294,"mean_input_tokens":4096.0,"mean_output_tokens":128.0,"ttft_p50_ms":250.82844705320895,"ttft_p95_ms":256.7141311708838,"ttft_p99_ms":257.63806560076773,"tpot_p50_ms":26.909995739766227,"tpot_p95_ms":27.076458471907873,"tpot_p99_ms":27.091512166432977,"e2e_p50_ms":3668.39790600352,"e2e_p95_ms":3695.424357103184,"e2e_p99_ms":3698.260110737756,"itl_p50_ms":24.74083099514246,"itl_p95_ms":47.6031189551577,"itl_p99_ms":49.66991530265658,"validation_errors":[],"timestamp":"2026-07-16T19:44:08+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":78.74828192405403,"request_tps":1.0158951794929716,"input_tps":4161.106655203212,"output_tps":260.0691659502007,"total_tps":4421.1758211534125,"mean_input_tokens":4096.0,"mean_output_tokens":256.0,"ttft_p50_ms":3091.937036952004,"ttft_p95_ms":11642.126731504686,"ttft_p99_ms":11714.910287000239,"tpot_p50_ms":37.18031867425524,"tpot_p95_ms":68.26280832984575,"tpot_p99_ms":73.90357631190186,"e2e_p50_ms":15660.35121458117,"e2e_p95_ms":21110.056800360326,"e2e_p99_ms":21111.51674753055,"itl_p50_ms":34.62257899809629,"itl_p95_ms":57.151046535000205,"itl_p99_ms":62.4327381211333,"validation_errors":[],"timestamp":"2026-07-16T19:45:48+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":70.42232585605234,"request_tps":0.5680016885804441,"input_tps":2326.534916425499,"output_tps":145.4084322765937,"total_tps":2471.943348702093,"mean_input_tokens":4096.0,"mean_output_tokens":256.0,"ttft_p50_ms":5035.245958482847,"ttft_p95_ms":5984.106912557036,"ttft_p99_ms":6042.94591952581,"tpot_p50_ms":35.78011738455982,"tpot_p95_ms":46.93394760007733,"tpot_p99_ms":47.50900067536927,"e2e_p50_ms":14144.402876030654,"e2e_p95_ms":14412.95803807443,"e2e_p99_ms":14413.140350417234,"itl_p50_ms":30.597370583564043,"itl_p95_ms":50.928640994243324,"itl_p99_ms":56.54409846989438,"validation_errors":[],"timestamp":"2026-07-16T19:47:19+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o256_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":35.27845326298848,"request_tps":0.1417295696817192,"input_tps":580.5243174163219,"output_tps":36.28276983852012,"total_tps":616.807087254842,"mean_input_tokens":4096.0,"mean_output_tokens":256.0,"ttft_p50_ms":268.0197979789227,"ttft_p95_ms":291.9198756571859,"ttft_p99_ms":292.9051029961556,"tpot_p50_ms":26.539895318302456,"tpot_p95_ms":26.827862831928275,"tpot_p99_ms":26.862892086781997,"e2e_p50_ms":7060.824715998024,"e2e_p95_ms":7112.919608317316,"e2e_p99_ms":7118.816237747669,"itl_p50_ms":24.650061037391424,"itl_p95_ms":44.851708575151854,"itl_p99_ms":49.219665322452784,"validation_errors":[],"timestamp":"2026-07-16T19:48:14+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":256,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":126.69021714199334,"request_tps":0.63146154300404,"input_tps":2586.466480144548,"output_tps":323.3083100180685,"total_tps":2909.7747901626167,"mean_input_tokens":4096.0,"mean_output_tokens":512.0,"ttft_p50_ms":3073.0733450036496,"ttft_p95_ms":11019.555248494726,"ttft_p99_ms":11065.65095537575,"tpot_p50_ms":38.32130294193836,"tpot_p95_ms":53.03848213837399,"tpot_p99_ms":58.19536544753461,"e2e_p50_ms":25832.93817844242,"e2e_p95_ms":30626.584287162405,"e2e_p99_ms":30627.329029503744,"itl_p50_ms":34.66650005429983,"itl_p95_ms":56.83219095226377,"itl_p99_ms":62.253236598335214,"validation_errors":[],"timestamp":"2026-07-16T19:50:41+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":112.55219594296068,"request_tps":0.3553906671023215,"input_tps":1455.6801724511088,"output_tps":181.9600215563886,"total_tps":1637.6401940074975,"mean_input_tokens":4096.0,"mean_output_tokens":512.0,"ttft_p50_ms":5065.312996157445,"ttft_p95_ms":6016.085800586734,"ttft_p99_ms":6074.749361765571,"tpot_p50_ms":34.33472787484767,"tpot_p95_ms":40.08658976554269,"tpot_p99_ms":40.18997862394681,"e2e_p50_ms":22591.4841449121,"e2e_p95_ms":22795.752924506087,"e2e_p99_ms":22796.309465037193,"itl_p50_ms":30.596577562391758,"itl_p95_ms":50.246595905628055,"itl_p99_ms":56.772089656442404,"validation_errors":[],"timestamp":"2026-07-16T19:52:54+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o512_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":69.34047695202753,"request_tps":0.07210795511918958,"input_tps":295.3541841682005,"output_tps":36.91927302102506,"total_tps":332.27345718922555,"mean_input_tokens":4096.0,"mean_output_tokens":512.0,"ttft_p50_ms":248.92402300611138,"ttft_p95_ms":266.0276424139738,"ttft_p99_ms":268.3548469096422,"tpot_p50_ms":26.661601453862808,"tpot_p95_ms":26.700651652677212,"tpot_p99_ms":26.707691252833037,"e2e_p50_ms":13878.469962859526,"e2e_p95_ms":13891.861443966627,"e2e_p99_ms":13893.72499614954,"itl_p50_ms":24.624896002933383,"itl_p95_ms":45.65126097295369,"itl_p99_ms":49.571911790408194,"validation_errors":[],"timestamp":"2026-07-16T19:54:24+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":512,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":219.81785867293365,"request_tps":0.363937673139796,"input_tps":1490.6887091806045,"output_tps":372.67217729515113,"total_tps":1863.3608864757555,"mean_input_tokens":4096.0,"mean_output_tokens":1024.0,"ttft_p50_ms":3086.4613954909146,"ttft_p95_ms":11646.606297139078,"ttft_p99_ms":11760.603103360627,"tpot_p50_ms":37.052429024333115,"tpot_p95_ms":44.42799982865398,"tpot_p99_ms":45.902305760968,"e2e_p50_ms":43634.44979744963,"e2e_p95_ms":49578.70842582779,"e2e_p99_ms":49580.79158608103,"itl_p50_ms":34.58560153376311,"itl_p95_ms":56.85673162806779,"itl_p99_ms":61.24365669908002,"validation_errors":[],"timestamp":"2026-07-16T19:58:24+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":196.3215377200395,"request_tps":0.2037473853584074,"input_tps":834.5492904280367,"output_tps":208.63732260700917,"total_tps":1043.1866130350459,"mean_input_tokens":4096.0,"mean_output_tokens":1024.0,"ttft_p50_ms":5105.625024996698,"ttft_p95_ms":5958.576378412545,"ttft_p99_ms":6064.859193183947,"tpot_p50_ms":33.55810541205767,"tpot_p95_ms":36.34212366745689,"tpot_p99_ms":36.55665332609869,"e2e_p50_ms":39269.71339690499,"e2e_p95_ms":39593.57306190068,"e2e_p99_ms":39594.54825331457,"itl_p50_ms":30.5747389793396,"itl_p95_ms":50.719153042882674,"itl_p99_ms":56.99894693214446,"validation_errors":[],"timestamp":"2026-07-16T20:02:01+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o1024_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":137.4567740750499,"request_tps":0.0363750716081119,"input_tps":148.99229330682635,"output_tps":37.24807332670659,"total_tps":186.24036663353294,"mean_input_tokens":4096.0,"mean_output_tokens":1024.0,"ttft_p50_ms":250.88916602544487,"ttft_p95_ms":253.6237696185708,"ttft_p99_ms":253.63623090088367,"tpot_p50_ms":26.61167698223162,"tpot_p95_ms":26.682842084191165,"tpot_p99_ms":26.683082640242063,"e2e_p50_ms":27477.38489904441,"e2e_p95_ms":27547.797128511593,"e2e_p99_ms":27548.662202609703,"itl_p50_ms":24.606115883216262,"itl_p95_ms":47.627381072379656,"itl_p99_ms":50.5407242104411,"validation_errors":[],"timestamp":"2026-07-16T20:04:39+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":1024,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":407.2602603479754,"request_tps":0.19643458443906506,"input_tps":804.5960578624105,"output_tps":402.29802893120524,"total_tps":1206.8940867936158,"mean_input_tokens":4096.0,"mean_output_tokens":2048.0,"ttft_p50_ms":3240.20006461069,"ttft_p95_ms":11741.664571419824,"ttft_p99_ms":11840.223265609238,"tpot_p50_ms":36.97641234489819,"tpot_p95_ms":40.52301196099482,"tpot_p99_ms":41.266952687252264,"e2e_p50_ms":80944.1796640167,"e2e_p95_ms":87096.1231123074,"e2e_p99_ms":87098.59982100548,"itl_p50_ms":34.466654527932405,"itl_p95_ms":56.807562068570405,"itl_p99_ms":61.39076543040574,"validation_errors":[],"timestamp":"2026-07-16T20:11:47+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o2048_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":363.24991058488376,"request_tps":0.11011702641741698,"input_tps":451.03934020573996,"output_tps":225.51967010286998,"total_tps":676.5590103086099,"mean_input_tokens":4096.0,"mean_output_tokens":2048.0,"ttft_p50_ms":5055.334709468298,"ttft_p95_ms":5906.876079260837,"ttft_p99_ms":6023.611045922153,"tpot_p50_ms":33.129859284787166,"tpot_p95_ms":34.53019272802082,"tpot_p99_ms":34.57053666370664,"e2e_p50_ms":72728.55528548826,"e2e_p95_ms":73087.71414748626,"e2e_p99_ms":73088.92423222307,"itl_p50_ms":30.50815698225051,"itl_p95_ms":50.1406642026268,"itl_p99_ms":56.95488589582965,"validation_errors":[],"timestamp":"2026-07-16T20:18:11+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":2048,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o2048_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":273.5867548820097,"request_tps":0.018275738539156837,"input_tps":74.8574250563864,"output_tps":37.4287125281932,"total_tps":112.28613758457962,"mean_input_tokens":4096.0,"mean_output_tokens":2048.0,"ttft_p50_ms":261.7046369705349,"ttft_p95_ms":349.8987853061408,"ttft_p99_ms":362.0478419493884,"tpot_p50_ms":26.592822536383906,"tpot_p95_ms":26.61543222684295,"tpot_p99_ms":26.61759176356131,"e2e_p50_ms":54736.32405605167,"e2e_p95_ms":54748.984178341925,"e2e_p99_ms":54749.01293158531,"itl_p50_ms":24.60544602945447,"itl_p95_ms":47.4006032804027,"itl_p99_ms":49.956067339517176,"validation_errors":[],"timestamp":"2026-07-16T20:23:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":2048,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":781.9880918110721,"request_tps":0.10230334814270799,"input_tps":419.0345139925319,"output_tps":419.0345139925319,"total_tps":838.0690279850638,"mean_input_tokens":4096.0,"mean_output_tokens":4096.0,"ttft_p50_ms":3097.252030042,"ttft_p95_ms":11633.983958070166,"ttft_p99_ms":11746.600034988951,"tpot_p50_ms":36.86781218838315,"tpot_p95_ms":38.49309932329363,"tpot_p99_ms":38.90570051973158,"e2e_p50_ms":155683.48012503702,"e2e_p95_ms":162603.52390279295,"e2e_p99_ms":162604.7951752739,"itl_p50_ms":34.45248049683869,"itl_p95_ms":56.79855167400092,"itl_p99_ms":61.650155324023224,"validation_errors":[],"timestamp":"2026-07-16T20:36:29+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":4096,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i4096_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i4096_o4096_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":697.7969978589099,"request_tps":0.057323261812152054,"input_tps":234.79608038257481,"output_tps":234.79608038257481,"total_tps":469.59216076514963,"mean_input_tokens":4096.0,"mean_output_tokens":4096.0,"ttft_p50_ms":5059.23117196653,"ttft_p95_ms":6020.107874658424,"ttft_p99_ms":6025.99322974449,"tpot_p50_ms":32.921226551871115,"tpot_p95_ms":33.65806957458343,"tpot_p99_ms":33.68422295676342,"e2e_p50_ms":139956.67370257434,"e2e_p95_ms":140090.03641876625,"e2e_p99_ms":140090.8339079097,"itl_p50_ms":30.493864906020463,"itl_p95_ms":50.366084929555655,"itl_p99_ms":57.08304369123653,"validation_errors":[],"timestamp":"2026-07-16T20:48:28+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":4096,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i4096_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i4096_o4096_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":548.1068854129408,"request_tps":0.009122308318081111,"input_tps":37.36497487086023,"output_tps":37.36497487086023,"total_tps":74.72994974172046,"mean_input_tokens":4096.0,"mean_output_tokens":4096.0,"ttft_p50_ms":270.29989496804774,"ttft_p95_ms":280.77137977816164,"ttft_p99_ms":282.4734351504594,"tpot_p50_ms":26.703159420515647,"tpot_p95_ms":26.721241031938273,"tpot_p99_ms":26.721443862750927,"e2e_p50_ms":109619.73772197962,"e2e_p95_ms":109701.13868489861,"e2e_p99_ms":109702.0101480931,"itl_p50_ms":24.71006289124489,"itl_p95_ms":45.58566838968543,"itl_p99_ms":49.87936850637197,"validation_errors":[],"timestamp":"2026-07-16T20:57:57+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":4096,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i4096_o4096_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i4096_o4096_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":256.7601454681717,"request_tps":0.31157483516037693,"input_tps":5104.842099267616,"output_tps":39.88157890052825,"total_tps":5144.723678168144,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":25187.957516987808,"ttft_p95_ms":47894.87515146611,"ttft_p99_ms":49477.59126278339,"tpot_p50_ms":198.17720644458336,"tpot_p95_ms":368.67252661503744,"tpot_p99_ms":391.46561684295915,"e2e_p50_ms":54196.20531098917,"e2e_p95_ms":54292.84199364483,"e2e_p99_ms":54294.34734713286,"itl_p50_ms":35.08518589660525,"itl_p95_ms":58.37633586488664,"itl_p99_ms":64.08695708261803,"validation_errors":[],"timestamp":"2026-07-16T21:02:35+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":141.1061867091339,"request_tps":0.283474459432832,"input_tps":4644.44554334752,"output_tps":36.2847308074025,"total_tps":4680.730274154922,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":14016.848506988026,"ttft_p95_ms":24515.615026152227,"ttft_p99_ms":24870.744992317632,"tpot_p50_ms":105.65934030661785,"tpot_p95_ms":191.19213953821398,"tpot_p99_ms":193.04246251707838,"e2e_p50_ms":28676.901251543313,"e2e_p95_ms":29117.543387238402,"e2e_p99_ms":29119.045873775613,"itl_p50_ms":30.759848072193563,"itl_p95_ms":50.73059933492914,"itl_p99_ms":57.838678041007384,"validation_errors":[],"timestamp":"2026-07-16T21:05:20+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":30.202961774077266,"request_tps":0.16554667841520834,"input_tps":2712.3167791547735,"output_tps":21.189974837146668,"total_tps":2733.5067539919205,"mean_input_tokens":16384.0,"mean_output_tokens":128.0,"ttft_p50_ms":3189.695225097239,"ttft_p95_ms":3287.1102603152394,"ttft_p99_ms":3298.7277361750603,"tpot_p50_ms":27.00854029953773,"tpot_p95_ms":27.13199242340826,"tpot_p99_ms":27.15187601517798,"e2e_p50_ms":6534.987838938832,"e2e_p95_ms":6732.873298088089,"e2e_p99_ms":6747.015990102664,"itl_p50_ms":24.80782102793455,"itl_p95_ms":45.76855925843107,"itl_p99_ms":49.268019353039556,"validation_errors":[],"timestamp":"2026-07-16T21:06:13+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i16384_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i16384_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":278.9136179748457,"request_tps":0.2868271566690406,"input_tps":4699.376134865561,"output_tps":73.42775210727439,"total_tps":4772.803886972835,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":24543.6039295746,"ttft_p95_ms":47646.79518254706,"ttft_p99_ms":49249.08797713695,"tpot_p50_ms":123.02242240768072,"tpot_p95_ms":202.41332247026955,"tpot_p99_ms":213.29141370800562,"e2e_p50_ms":58673.51271153893,"e2e_p95_ms":59028.76297895564,"e2e_p99_ms":59031.15820220439,"itl_p50_ms":35.10860598180443,"itl_p95_ms":59.34609065297991,"itl_p99_ms":63.67699921829625,"validation_errors":[],"timestamp":"2026-07-16T21:11:13+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":162.15329257398844,"request_tps":0.24668015903376442,"input_tps":4041.607725609196,"output_tps":63.15012071264369,"total_tps":4104.75784632184,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":14056.083535542712,"ttft_p95_ms":24572.28690140182,"ttft_p99_ms":24781.60693206126,"tpot_p50_ms":69.22504512739239,"tpot_p95_ms":111.92492204695465,"tpot_p99_ms":112.1908097408584,"e2e_p50_ms":32945.15154813416,"e2e_p95_ms":33157.10190224927,"e2e_p99_ms":33157.680223954376,"itl_p50_ms":30.78671346884221,"itl_p95_ms":52.047876664437354,"itl_p99_ms":57.51584505662322,"validation_errors":[],"timestamp":"2026-07-16T21:14:19+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":47.27609982597642,"request_tps":0.10576168546908539,"input_tps":1732.799454725495,"output_tps":27.07499148008586,"total_tps":1759.8744462055809,"mean_input_tokens":16384.0,"mean_output_tokens":256.0,"ttft_p50_ms":3095.8918230608106,"ttft_p95_ms":3272.758443793282,"ttft_p99_ms":3304.284312753007,"tpot_p50_ms":27.063766945445657,"tpot_p95_ms":27.16058336417465,"tpot_p99_ms":27.17001092872199,"e2e_p50_ms":9997.152394149452,"e2e_p95_ms":10198.707201657817,"e2e_p99_ms":10232.637099577114,"itl_p50_ms":24.803407955914736,"itl_p95_ms":48.14038993790746,"itl_p99_ms":50.25792038068175,"validation_errors":[],"timestamp":"2026-07-16T21:15:29+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":256,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i16384_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i16384_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":326.68043872504495,"request_tps":0.2448876348771317,"input_tps":4012.2390098269257,"output_tps":125.38246905709143,"total_tps":4137.621478884017,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":24534.058870980516,"ttft_p95_ms":47585.56273026624,"ttft_p99_ms":49259.400079401676,"tpot_p50_ms":80.12083880425615,"tpot_p95_ms":119.69057329970916,"tpot_p99_ms":125.067007240863,"e2e_p50_ms":68319.87011944875,"e2e_p95_ms":68537.59080437012,"e2e_p99_ms":68539.8418288771,"itl_p50_ms":35.064953030087054,"itl_p95_ms":58.41207688208669,"itl_p99_ms":63.36872420506552,"validation_errors":[],"timestamp":"2026-07-16T21:21:17+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":204.11131394607946,"request_tps":0.19597149823143506,"input_tps":3210.797027023832,"output_tps":100.33740709449475,"total_tps":3311.134434118327,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":14109.92513189558,"ttft_p95_ms":24601.094605890103,"ttft_p99_ms":24752.130593049806,"tpot_p50_ms":50.97825450885589,"tpot_p95_ms":72.14095964029964,"tpot_p99_ms":72.50778853802152,"e2e_p50_ms":41347.00059145689,"e2e_p95_ms":41603.95504550543,"e2e_p99_ms":41605.48491582973,"itl_p50_ms":30.709165963344276,"itl_p95_ms":51.023998111486435,"itl_p99_ms":57.31804177630693,"validation_errors":[],"timestamp":"2026-07-16T21:25:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":81.74180358601734,"request_tps":0.06116821235463043,"input_tps":1002.179991218265,"output_tps":31.31812472557078,"total_tps":1033.4981159438357,"mean_input_tokens":16384.0,"mean_output_tokens":512.0,"ttft_p50_ms":3152.1379488985986,"ttft_p95_ms":3280.934850871563,"ttft_p99_ms":3303.358238041401,"tpot_p50_ms":26.90626537553062,"tpot_p95_ms":26.921899768354372,"tpot_p99_ms":26.923089013626313,"e2e_p50_ms":16891.912296880037,"e2e_p95_ms":17034.689639275894,"e2e_p99_ms":17060.389525359496,"itl_p50_ms":24.79128399863839,"itl_p95_ms":48.06917109526691,"itl_p99_ms":50.64208318479359,"validation_errors":[],"timestamp":"2026-07-16T21:26:50+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":512,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i16384_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i16384_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":423.16662176605314,"request_tps":0.1890508274639578,"input_tps":3097.4087571694845,"output_tps":193.58804732309278,"total_tps":3290.996804492577,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":24554.481841041707,"ttft_p95_ms":47767.74098525056,"ttft_p99_ms":49284.376522910316,"tpot_p50_ms":58.8259498166778,"tpot_p95_ms":78.8021439746958,"tpot_p99_ms":81.38845075320822,"e2e_p50_ms":87506.93495490123,"e2e_p95_ms":88091.98981691152,"e2e_p99_ms":88095.21754639223,"itl_p50_ms":35.09103413671255,"itl_p95_ms":58.61164742382244,"itl_p99_ms":64.02320196386427,"validation_errors":[],"timestamp":"2026-07-16T21:34:14+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":288.24989658407867,"request_tps":0.13876847996832684,"input_tps":2273.582775801067,"output_tps":142.09892348756668,"total_tps":2415.6816992886334,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":14033.50293962285,"ttft_p95_ms":24491.35324291419,"ttft_p99_ms":24759.231015304103,"tpot_p50_ms":41.92089678206183,"tpot_p95_ms":52.53722619232089,"tpot_p99_ms":52.69310522824981,"e2e_p50_ms":58040.07501259912,"e2e_p95_ms":58448.35202746326,"e2e_p99_ms":58450.28124490753,"itl_p50_ms":30.73286300059408,"itl_p95_ms":50.72684104088693,"itl_p99_ms":57.92215333785862,"validation_errors":[],"timestamp":"2026-07-16T21:39:26+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":150.84414941398427,"request_tps":0.03314679435314225,"input_tps":543.0770786818827,"output_tps":33.942317417617666,"total_tps":577.0193960995003,"mean_input_tokens":16384.0,"mean_output_tokens":1024.0,"ttft_p50_ms":3190.0415259879082,"ttft_p95_ms":3315.6291431281716,"ttft_p99_ms":3326.89755034633,"tpot_p50_ms":26.902348835946206,"tpot_p95_ms":26.929367462743006,"tpot_p99_ms":26.932126218243663,"e2e_p50_ms":30694.867955055088,"e2e_p95_ms":30857.89103670977,"e2e_p99_ms":30877.166467448696,"itl_p50_ms":24.886570870876312,"itl_p95_ms":45.76608098577708,"itl_p99_ms":50.4940598970279,"validation_errors":[],"timestamp":"2026-07-16T21:42:20+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":1024,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i16384_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i16384_o1024_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":614.5522862081416,"request_tps":0.13017606767620898,"input_tps":2132.804692807008,"output_tps":266.600586600876,"total_tps":2399.4052794078843,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":24607.956594438292,"ttft_p95_ms":47697.99113486661,"ttft_p99_ms":49292.4090649304,"tpot_p50_ms":48.10298325354711,"tpot_p95_ms":58.110758199371965,"tpot_p99_ms":59.46979306338756,"e2e_p50_ms":125611.4937895909,"e2e_p95_ms":126450.29024835676,"e2e_p99_ms":126453.36987355025,"itl_p50_ms":35.09154496714473,"itl_p95_ms":58.45766137354075,"itl_p99_ms":64.08793116686866,"validation_errors":[],"timestamp":"2026-07-16T21:52:57+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":2048,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":457.1198655979242,"request_tps":0.08750440094673856,"input_tps":1433.6721051113645,"output_tps":179.20901313892057,"total_tps":1612.881118250285,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":14047.149811987765,"ttft_p95_ms":24552.844235277735,"ttft_p99_ms":24749.485357583035,"tpot_p50_ms":37.538202461908504,"tpot_p95_ms":42.66441069203499,"tpot_p99_ms":42.9089602763204,"e2e_p50_ms":91750.90344448108,"e2e_p95_ms":92103.39087992907,"e2e_p99_ms":92105.43445243035,"itl_p50_ms":30.728504993021488,"itl_p95_ms":51.14136257907376,"itl_p99_ms":58.35309494053946,"validation_errors":[],"timestamp":"2026-07-16T22:00:58+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":2048,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":288.6096170139499,"request_tps":0.017324440022933558,"input_tps":283.8436253357434,"output_tps":35.480453166967926,"total_tps":319.32407850271136,"mean_input_tokens":16384.0,"mean_output_tokens":2048.0,"ttft_p50_ms":3130.5205351673067,"ttft_p95_ms":3326.790696242824,"ttft_p99_ms":3348.429289693013,"tpot_p50_ms":26.910882942859878,"tpot_p95_ms":26.939971389955993,"tpot_p99_ms":26.940729667464,"e2e_p50_ms":58267.43282494135,"e2e_p95_ms":58423.72469645925,"e2e_p99_ms":58437.077996963635,"itl_p50_ms":24.81332910247147,"itl_p95_ms":48.45747402869165,"itl_p99_ms":51.16548259742558,"validation_errors":[],"timestamp":"2026-07-16T22:06:10+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":2048,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i16384_o2048_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i16384_o2048_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1164.4257685828488,"request_tps":0.06870339197093095,"input_tps":4502.545496206931,"output_tps":8.794034172279162,"total_tps":4511.33953037921,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":120625.26360049378,"ttft_p95_ms":230919.13024774985,"ttft_p99_ms":233213.97572936953,"tpot_p50_ms":847.999912252151,"tpot_p95_ms":1666.0538613161907,"tpot_p99_ms":1748.680016982211,"e2e_p50_ms":236925.28252699412,"e2e_p95_ms":241650.54931361228,"e2e_p99_ms":241654.8121647397,"itl_p50_ms":36.83545487001538,"itl_p95_ms":60.686023067682996,"itl_p99_ms":88.91421182081103,"validation_errors":[],"timestamp":"2026-07-16T22:26:09+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":587.0072881928645,"request_tps":0.06814225445674156,"input_tps":4465.770788077015,"output_tps":8.72220857046292,"total_tps":4474.492996647478,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":60264.91823955439,"ttft_p95_ms":116942.56678273669,"ttft_p99_ms":116968.76564948354,"tpot_p50_ms":419.5593314018012,"tpot_p95_ms":818.3646683727837,"tpot_p99_ms":829.4111081236225,"e2e_p50_ms":119654.99441663269,"e2e_p95_ms":121336.96261633886,"e2e_p99_ms":121339.05124852201,"itl_p50_ms":31.66217007674277,"itl_p95_ms":56.3370821531862,"itl_p99_ms":61.347583890892565,"validation_errors":[],"timestamp":"2026-07-16T22:36:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":128,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":76.5485741731245,"request_tps":0.0653180030328436,"input_tps":4280.680646760438,"output_tps":8.360704388203981,"total_tps":4289.041351148642,"mean_input_tokens":65536.0,"mean_output_tokens":128.0,"ttft_p50_ms":14596.343518933281,"ttft_p95_ms":14810.734046529979,"ttft_p99_ms":14839.107076432556,"tpot_p50_ms":27.468161299956712,"tpot_p95_ms":27.539072871046862,"tpot_p99_ms":27.552164999592144,"e2e_p50_ms":18084.800004027784,"e2e_p95_ms":18259.232654888183,"e2e_p99_ms":18286.724362093955,"itl_p50_ms":25.447878055274487,"itl_p95_ms":46.60883932374417,"itl_p99_ms":51.75249738618731,"validation_errors":[],"timestamp":"2026-07-16T22:38:22+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":128,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i65536_o128_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i65536_o128_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1179.4044030888472,"request_tps":0.06783084732470125,"input_tps":4445.362410271621,"output_tps":17.36469691512352,"total_tps":4462.727107186744,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":118520.80365340225,"ttft_p95_ms":219668.39720717393,"ttft_p99_ms":232060.81151018385,"tpot_p50_ms":456.4013590040964,"tpot_p95_ms":834.1273255841166,"tpot_p99_ms":887.1042354636113,"e2e_p50_ms":241645.95780789386,"e2e_p95_ms":242439.91744236555,"e2e_p99_ms":242443.70790595654,"itl_p50_ms":36.84396401513368,"itl_p95_ms":61.602629092521966,"itl_p99_ms":67.18409784138203,"validation_errors":[],"timestamp":"2026-07-16T22:58:38+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":608.7442546100356,"request_tps":0.06570903905388674,"input_tps":4306.307583435521,"output_tps":16.821513997795005,"total_tps":4323.129097433316,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":60408.123059547506,"ttft_p95_ms":116816.32042041747,"ttft_p99_ms":117158.49739520112,"tpot_p50_ms":225.85036875497477,"tpot_p95_ms":424.8813510864662,"tpot_p99_ms":430.2625958223681,"e2e_p50_ms":123890.77430358157,"e2e_p95_ms":125879.0796181187,"e2e_p99_ms":125880.81930729328,"itl_p50_ms":31.708508962765336,"itl_p95_ms":55.004988797009,"itl_p99_ms":60.499521191231906,"validation_errors":[],"timestamp":"2026-07-16T23:09:22+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":256,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":94.3989737580996,"request_tps":0.052966677506607864,"input_tps":3471.224177073053,"output_tps":13.559469441691613,"total_tps":3484.7836465147443,"mean_input_tokens":65536.0,"mean_output_tokens":256.0,"ttft_p50_ms":14599.749388871714,"ttft_p95_ms":14722.214041789994,"ttft_p99_ms":14730.883328346536,"tpot_p50_ms":27.881732368476545,"tpot_p95_ms":27.903486282619486,"tpot_p99_ms":27.905507876263826,"e2e_p50_ms":21712.56085904315,"e2e_p95_ms":21806.284560635686,"e2e_p99_ms":21809.799600169063,"itl_p50_ms":25.44583403505385,"itl_p95_ms":49.89882283844054,"itl_p99_ms":52.76037988718599,"validation_errors":[],"timestamp":"2026-07-16T23:11:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":256,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i65536_o256_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i65536_o256_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1229.7762659930158,"request_tps":0.0650524832949202,"input_tps":4263.27954521589,"output_tps":33.30687144699914,"total_tps":4296.586416662889,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":117602.89832751732,"ttft_p95_ms":220174.25819333873,"ttft_p99_ms":232176.57839990687,"tpot_p50_ms":250.0348996674941,"tpot_p95_ms":436.4520301115598,"tpot_p99_ms":462.66546465371283,"e2e_p50_ms":251397.16331893578,"e2e_p95_ms":252849.9990378623,"e2e_p99_ms":252851.82005017297,"itl_p50_ms":36.86228208243847,"itl_p95_ms":60.973369050771,"itl_p99_ms":67.14509757934138,"validation_errors":[],"timestamp":"2026-07-16T23:32:37+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":512,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":652.6434917969164,"request_tps":0.061289203834498415,"input_tps":4016.649262497688,"output_tps":31.38007236326319,"total_tps":4048.029334860951,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":60481.61956341937,"ttft_p95_ms":116843.18738668226,"ttft_p99_ms":117324.80349385645,"tpot_p50_ms":131.188325970543,"tpot_p95_ms":229.34224289243244,"tpot_p99_ms":231.58556326105642,"e2e_p50_ms":132822.63546145987,"e2e_p95_ms":134675.67310340237,"e2e_p99_ms":134677.4176377803,"itl_p50_ms":31.73411381430924,"itl_p95_ms":53.03707402199507,"itl_p99_ms":61.14733610767871,"validation_errors":[],"timestamp":"2026-07-16T23:44:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":512,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":130.4824561579153,"request_tps":0.03831932772593422,"input_tps":2511.295461846825,"output_tps":19.619495795678322,"total_tps":2530.9149576425034,"mean_input_tokens":65536.0,"mean_output_tokens":512.0,"ttft_p50_ms":14583.937218878418,"ttft_p95_ms":14795.338988443837,"ttft_p99_ms":14826.081829732284,"tpot_p50_ms":27.982182919691134,"tpot_p95_ms":28.00536987685221,"tpot_p99_ms":28.00938046322279,"e2e_p50_ms":28872.23194888793,"e2e_p95_ms":29090.035974793136,"e2e_p99_ms":29125.418794117868,"itl_p50_ms":25.67037590779364,"itl_p95_ms":50.41482762899249,"itl_p99_ms":53.483981639146805,"validation_errors":[],"timestamp":"2026-07-16T23:46:51+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":512,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i65536_o512_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i65536_o512_a1.log"} +{"status":"COMPLETED","completed":80,"failed":0,"duration_s":1330.2602253910154,"request_tps":0.06013860932847547,"input_tps":3941.2439009509685,"output_tps":61.58193595235888,"total_tps":4002.8258369033274,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":118513.18043493666,"ttft_p95_ms":230933.78256635042,"ttft_p99_ms":232188.58536266722,"tpot_p50_ms":137.7084055620283,"tpot_p95_ms":237.46356306838157,"tpot_p99_ms":250.69030466432596,"e2e_p50_ms":271416.1142479861,"e2e_p95_ms":272710.0276091136,"e2e_p99_ms":272711.17356425617,"itl_p50_ms":36.84138902463019,"itl_p95_ms":61.22844358906149,"itl_p99_ms":67.15500121004875,"validation_errors":[],"timestamp":"2026-07-17T00:09:38+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":1024,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i65536_o1024_a1.log"} +{"status":"COMPLETED","completed":40,"failed":0,"duration_s":739.3616181469988,"request_tps":0.054100725569510504,"input_tps":3545.5451509234404,"output_tps":55.399142983178756,"total_tps":3600.9442939066194,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":60294.08791498281,"ttft_p95_ms":116841.35833900655,"ttft_p99_ms":116967.29428508785,"tpot_p50_ms":82.37472311334132,"tpot_p95_ms":131.37778040708542,"tpot_p99_ms":132.5520263926535,"e2e_p50_ms":150254.07721009105,"e2e_p95_ms":151760.6172351283,"e2e_p99_ms":151763.06750814896,"itl_p50_ms":31.72238008119166,"itl_p95_ms":53.61441783607006,"itl_p99_ms":61.79662807378918,"validation_errors":[],"timestamp":"2026-07-17T00:22:33+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":1024,"concurrency":8,"num_prompts":40,"warmup_requests":8,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c8_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c8_i65536_o1024_a1.log"} +{"status":"COMPLETED","completed":5,"failed":0,"duration_s":201.46440248796716,"request_tps":0.024818280243323058,"input_tps":1626.49081402642,"output_tps":25.41391896916281,"total_tps":1651.9047329955827,"mean_input_tokens":65536.0,"mean_output_tokens":1024.0,"ttft_p50_ms":14628.885317826644,"ttft_p95_ms":14868.979735858738,"ttft_p99_ms":14913.150340802968,"tpot_p50_ms":27.836642022547792,"tpot_p95_ms":27.844008100505842,"tpot_p99_ms":27.84499085022142,"e2e_p50_ms":43054.00717793964,"e2e_p95_ms":43330.51605997607,"e2e_p99_ms":43373.7068128027,"itl_p50_ms":25.475302943959832,"itl_p95_ms":50.770317669957876,"itl_p99_ms":54.30494689382613,"validation_errors":[],"timestamp":"2026-07-17T00:26:29+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":1024,"concurrency":1,"num_prompts":5,"warmup_requests":1,"attempt":1,"gain_pct":null,"plateau_streak":0,"error_type":"","raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c1_i65536_o1024_a1.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c1_i65536_o1024_a1.log"} +{"timestamp":"2026-07-17T01:02:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":131072,"osl":128,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":2,"status":"FAILED","completed":0,"failed":80,"error_type":"benchmark rc=124","validation_errors":[],"gain_pct":null,"plateau_streak":0,"raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i131072_o128_a2.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i131072_o128_a2.log"} +{"timestamp":"2026-07-17T02:10:35+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":131072,"osl":256,"concurrency":16,"num_prompts":80,"warmup_requests":16,"attempt":2,"status":"FAILED","completed":0,"failed":80,"error_type":"benchmark rc=124","validation_errors":[],"gain_pct":null,"plateau_streak":0,"raw_file":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/raw_outputs/sglang_adaptive_c16_i131072_o256_a2.jsonl","detail_log":"/data1/yy/sskj/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/logs/sglang_c16_i131072_o256_a2.log"} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_shapes.jsonl b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_shapes.jsonl index ee81796..2c1cbef 100644 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_shapes.jsonl +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_shapes.jsonl @@ -1 +1,47 @@ {"timestamp":"2026-07-16T07:48:44+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":1,"search_cap":512,"max_successful_concurrency":16,"saturation_concurrency":null,"stop_probe_concurrency":16,"best_tps_concurrency":16,"best_total_tps":2233.1625760825486,"last_total_tps":2233.1625760825486} +{"timestamp":"2026-07-16T07:54:03+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":48,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":48,"best_total_tps":2994.9145225702982,"last_total_tps":2994.9145225702982} +{"timestamp":"2026-07-16T08:02:57+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":48,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":48,"best_total_tps":1894.263479895674,"last_total_tps":1894.263479895674} +{"timestamp":"2026-07-16T08:19:12+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":48,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":48,"best_total_tps":1270.3578826082223,"last_total_tps":1270.3578826082223} +{"timestamp":"2026-07-16T08:50:41+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":48,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":48,"best_total_tps":938.3875354487201,"last_total_tps":938.3875354487201} +{"timestamp":"2026-07-16T09:44:20+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":1024,"osl":4096,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":487.1207589944629,"last_total_tps":272.97053779597} +{"timestamp":"2026-07-16T09:46:49+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":5276.430317217383,"last_total_tps":3991.4044121212137} +{"timestamp":"2026-07-16T09:50:14+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":3654.27555715578,"last_total_tps":2642.4843178928554} +{"timestamp":"2026-07-16T09:55:31+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":2302.8342327716664,"last_total_tps":1632.8914416222033} +{"timestamp":"2026-07-16T10:04:24+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":1449.0672823499715,"last_total_tps":989.1396953244687} +{"timestamp":"2026-07-16T10:20:38+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":919.5235579896186,"last_total_tps":622.97916579306} +{"timestamp":"2026-07-16T10:51:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":4096,"osl":4096,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":636.5331301353165,"last_total_tps":429.68665622424965} +{"timestamp":"2026-07-16T10:58:35+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":6149.069264131796,"last_total_tps":2072.6357920596456} +{"timestamp":"2026-07-16T11:07:42+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4922.27619638361,"last_total_tps":1374.3485383101208} +{"timestamp":"2026-07-16T11:20:34+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":3457.304361342664,"last_total_tps":840.1799665951539} +{"timestamp":"2026-07-16T11:40:58+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":2203.3814838524254,"last_total_tps":479.3679364784214} +{"timestamp":"2026-07-16T12:16:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":16384,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":1322.1810796622076,"last_total_tps":269.1649593728095} +{"timestamp":"2026-07-16T12:39:17+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":6860.104148991479,"last_total_tps":3001.706937330212} +{"timestamp":"2026-07-16T13:04:16+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":6322.324656264061,"last_total_tps":2534.2482875187998} +{"timestamp":"2026-07-16T13:33:26+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":5426.021563663019,"last_total_tps":1933.302825215971} +{"timestamp":"2026-07-16T14:11:00+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":65536,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":8,"best_total_tps":4250.026326552517,"last_total_tps":1299.0787444904488} +{"timestamp":"2026-07-16T14:59:34+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":8,"best_total_tps":6414.4163774899025,"last_total_tps":2895.647145362982} +{"timestamp":"2026-07-16T16:03:27+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":256,"status":"FAILED","stop_reason":"BENCHMARK_FAILED","tested_points":1,"search_cap":512,"max_successful_concurrency":0,"saturation_concurrency":null,"stop_probe_concurrency":16,"best_tps_concurrency":null,"best_total_tps":null,"last_total_tps":null} +{"timestamp":"2026-07-16T17:11:50+00:00","engine":"sglang","tp":4,"dp":2,"mark":"Y","isl":131072,"osl":512,"status":"FAILED","stop_reason":"BENCHMARK_FAILED","tested_points":1,"search_cap":512,"max_successful_concurrency":0,"saturation_concurrency":null,"stop_probe_concurrency":16,"best_tps_concurrency":null,"best_total_tps":null,"last_total_tps":null} +{"timestamp":"2026-07-16T17:23:16+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":2,"search_cap":512,"max_successful_concurrency":8,"saturation_concurrency":null,"stop_probe_concurrency":8,"best_tps_concurrency":16,"best_total_tps":2252.3024943648124,"last_total_tps":1988.28275981793} +{"timestamp":"2026-07-16T17:25:59+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":2,"search_cap":512,"max_successful_concurrency":32,"saturation_concurrency":null,"stop_probe_concurrency":32,"best_tps_concurrency":32,"best_total_tps":2807.7871232493126,"last_total_tps":2807.7871232493126} +{"timestamp":"2026-07-16T17:39:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":48,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":32,"best_total_tps":2109.996176616557,"last_total_tps":684.1363844999557} +{"timestamp":"2026-07-16T17:47:25+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":2,"search_cap":512,"max_successful_concurrency":32,"saturation_concurrency":null,"stop_probe_concurrency":32,"best_tps_concurrency":32,"best_total_tps":1390.4790149026294,"last_total_tps":1390.4790149026294} +{"timestamp":"2026-07-16T18:01:59+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":2,"search_cap":512,"max_successful_concurrency":32,"saturation_concurrency":null,"stop_probe_concurrency":32,"best_tps_concurrency":32,"best_total_tps":1079.9126577555764,"last_total_tps":1079.9126577555764} +{"timestamp":"2026-07-16T19:35:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":1024,"osl":4096,"status":"FAILED","stop_reason":"BENCHMARK_FAILED","tested_points":3,"search_cap":512,"max_successful_concurrency":32,"saturation_concurrency":null,"stop_probe_concurrency":48,"best_tps_concurrency":32,"best_total_tps":911.6872643258514,"last_total_tps":911.6872643258514} +{"timestamp":"2026-07-16T19:44:08+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4128.998328888893,"last_total_tps":1151.1056276859294} +{"timestamp":"2026-07-16T19:48:14+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4421.1758211534125,"last_total_tps":616.807087254842} +{"timestamp":"2026-07-16T19:54:24+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":2909.7747901626167,"last_total_tps":332.27345718922555} +{"timestamp":"2026-07-16T20:04:39+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":1863.3608864757555,"last_total_tps":186.24036663353294} +{"timestamp":"2026-07-16T20:23:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":1206.8940867936158,"last_total_tps":112.28613758457962} +{"timestamp":"2026-07-16T20:57:57+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":4096,"osl":4096,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":838.0690279850638,"last_total_tps":74.72994974172046} +{"timestamp":"2026-07-16T21:06:13+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":5144.723678168144,"last_total_tps":2733.5067539919205} +{"timestamp":"2026-07-16T21:15:29+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4772.803886972835,"last_total_tps":1759.8744462055809} +{"timestamp":"2026-07-16T21:26:50+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4137.621478884017,"last_total_tps":1033.4981159438357} +{"timestamp":"2026-07-16T21:42:20+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":3290.996804492577,"last_total_tps":577.0193960995003} +{"timestamp":"2026-07-16T22:06:10+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":16384,"osl":2048,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_BACKOFF_FOUND","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":2399.4052794078843,"last_total_tps":319.32407850271136} +{"timestamp":"2026-07-16T22:38:22+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":128,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4511.33953037921,"last_total_tps":4289.041351148642} +{"timestamp":"2026-07-16T23:11:31+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":256,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4462.727107186744,"last_total_tps":3484.7836465147443} +{"timestamp":"2026-07-16T23:46:51+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":512,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4296.586416662889,"last_total_tps":2530.9149576425034} +{"timestamp":"2026-07-17T00:26:29+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":65536,"osl":1024,"status":"TTFT_BOUNDARY","stop_reason":"TTFT_SLO_EXCEEDED","tested_points":3,"search_cap":512,"max_successful_concurrency":1,"saturation_concurrency":null,"stop_probe_concurrency":1,"best_tps_concurrency":16,"best_total_tps":4002.8258369033274,"last_total_tps":1651.9047329955827} +{"timestamp":"2026-07-17T01:02:05+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":131072,"osl":128,"status":"FAILED","stop_reason":"BENCHMARK_FAILED","tested_points":1,"search_cap":512,"max_successful_concurrency":0,"saturation_concurrency":null,"stop_probe_concurrency":16,"best_tps_concurrency":null,"best_total_tps":null,"last_total_tps":null} +{"timestamp":"2026-07-17T02:10:35+00:00","engine":"sglang","tp":8,"dp":1,"mark":"Y","isl":131072,"osl":256,"status":"FAILED","stop_reason":"BENCHMARK_FAILED","tested_points":1,"search_cap":512,"max_successful_concurrency":0,"saturation_concurrency":null,"stop_probe_concurrency":16,"best_tps_concurrency":null,"best_total_tps":null,"last_total_tps":null} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.jsonl b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.jsonl new file mode 100644 index 0000000..06b0b14 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.jsonl @@ -0,0 +1,47 @@ +{"timestamp": "2026-07-16T07:48:44+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 1, "search_cap": 512, "max_successful_concurrency": 16, "saturation_concurrency": null, "stop_probe_concurrency": 16, "best_tps_concurrency": 16, "best_total_tps": 2233.1625760825486, "last_total_tps": 2233.1625760825486} +{"timestamp": "2026-07-16T07:54:03+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 48, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 48, "best_total_tps": 2994.9145225702982, "last_total_tps": 2994.9145225702982} +{"timestamp": "2026-07-16T08:02:57+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 48, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 48, "best_total_tps": 1894.263479895674, "last_total_tps": 1894.263479895674} +{"timestamp": "2026-07-16T08:19:12+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 48, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 48, "best_total_tps": 1270.3578826082223, "last_total_tps": 1270.3578826082223} +{"timestamp": "2026-07-16T08:50:41+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 48, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 48, "best_total_tps": 938.3875354487201, "last_total_tps": 938.3875354487201} +{"timestamp": "2026-07-16T09:44:20+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 1024, "osl": 4096, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 487.1207589944629, "last_total_tps": 272.97053779597} +{"timestamp": "2026-07-16T09:46:49+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 5276.430317217383, "last_total_tps": 3991.4044121212137} +{"timestamp": "2026-07-16T09:50:14+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 3654.27555715578, "last_total_tps": 2642.4843178928554} +{"timestamp": "2026-07-16T09:55:31+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 2302.8342327716664, "last_total_tps": 1632.8914416222033} +{"timestamp": "2026-07-16T10:04:24+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 1449.0672823499715, "last_total_tps": 989.1396953244687} +{"timestamp": "2026-07-16T10:20:38+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 919.5235579896186, "last_total_tps": 622.97916579306} +{"timestamp": "2026-07-16T10:51:17+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 4096, "osl": 4096, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 636.5331301353165, "last_total_tps": 429.68665622424965} +{"timestamp": "2026-07-16T10:58:35+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 16384, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 6149.069264131796, "last_total_tps": 2072.6357920596456} +{"timestamp": "2026-07-16T11:07:42+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 16384, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4922.27619638361, "last_total_tps": 1374.3485383101208} +{"timestamp": "2026-07-16T11:20:34+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 16384, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 3457.304361342664, "last_total_tps": 840.1799665951539} +{"timestamp": "2026-07-16T11:40:58+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 16384, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 2203.3814838524254, "last_total_tps": 479.3679364784214} +{"timestamp": "2026-07-16T12:16:16+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 16384, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 1322.1810796622076, "last_total_tps": 269.1649593728095} +{"timestamp": "2026-07-16T12:39:17+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 65536, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 6860.104148991479, "last_total_tps": 3001.706937330212} +{"timestamp": "2026-07-16T13:04:16+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 65536, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 6322.324656264061, "last_total_tps": 2534.2482875187998} +{"timestamp": "2026-07-16T13:33:26+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 65536, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 5426.021563663019, "last_total_tps": 1933.302825215971} +{"timestamp": "2026-07-16T14:11:00+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 65536, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 8, "best_total_tps": 4250.026326552517, "last_total_tps": 1299.0787444904488} +{"timestamp": "2026-07-16T14:59:34+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 131072, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 8, "best_total_tps": 6414.4163774899025, "last_total_tps": 2895.647145362982} +{"timestamp": "2026-07-16T16:03:27+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 131072, "osl": 256, "status": "FAILED", "stop_reason": "BENCHMARK_FAILED", "tested_points": 1, "search_cap": 512, "max_successful_concurrency": 0, "saturation_concurrency": null, "stop_probe_concurrency": 16, "best_tps_concurrency": null, "best_total_tps": null, "last_total_tps": null} +{"timestamp": "2026-07-16T17:11:50+00:00", "engine": "sglang", "tp": 4, "dp": 2, "mark": "Y", "isl": 131072, "osl": 512, "status": "FAILED", "stop_reason": "BENCHMARK_FAILED", "tested_points": 1, "search_cap": 512, "max_successful_concurrency": 0, "saturation_concurrency": null, "stop_probe_concurrency": 16, "best_tps_concurrency": null, "best_total_tps": null, "last_total_tps": null} +{"timestamp": "2026-07-16T17:23:16+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 8, "saturation_concurrency": null, "stop_probe_concurrency": 8, "best_tps_concurrency": 16, "best_total_tps": 2252.3024943648124, "last_total_tps": 1988.28275981793} +{"timestamp": "2026-07-16T17:25:59+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 32, "saturation_concurrency": null, "stop_probe_concurrency": 32, "best_tps_concurrency": 32, "best_total_tps": 2807.7871232493126, "last_total_tps": 2807.7871232493126} +{"timestamp": "2026-07-16T17:39:31+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 48, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 32, "best_total_tps": 2109.996176616557, "last_total_tps": 684.1363844999557} +{"timestamp": "2026-07-16T17:47:25+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 32, "saturation_concurrency": null, "stop_probe_concurrency": 32, "best_tps_concurrency": 32, "best_total_tps": 1390.4790149026294, "last_total_tps": 1390.4790149026294} +{"timestamp": "2026-07-16T18:01:59+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 2, "search_cap": 512, "max_successful_concurrency": 32, "saturation_concurrency": null, "stop_probe_concurrency": 32, "best_tps_concurrency": 32, "best_total_tps": 1079.9126577555764, "last_total_tps": 1079.9126577555764} +{"timestamp": "2026-07-16T19:35:31+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 1024, "osl": 4096, "status": "FAILED", "stop_reason": "BENCHMARK_FAILED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 32, "saturation_concurrency": null, "stop_probe_concurrency": 48, "best_tps_concurrency": 32, "best_total_tps": 911.6872643258514, "last_total_tps": 911.6872643258514} +{"timestamp": "2026-07-16T19:44:08+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4128.998328888893, "last_total_tps": 1151.1056276859294} +{"timestamp": "2026-07-16T19:48:14+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4421.1758211534125, "last_total_tps": 616.807087254842} +{"timestamp": "2026-07-16T19:54:24+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 2909.7747901626167, "last_total_tps": 332.27345718922555} +{"timestamp": "2026-07-16T20:04:39+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 1863.3608864757555, "last_total_tps": 186.24036663353294} +{"timestamp": "2026-07-16T20:23:05+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 1206.8940867936158, "last_total_tps": 112.28613758457962} +{"timestamp": "2026-07-16T20:57:57+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 4096, "osl": 4096, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 838.0690279850638, "last_total_tps": 74.72994974172046} +{"timestamp": "2026-07-16T21:06:13+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 16384, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 5144.723678168144, "last_total_tps": 2733.5067539919205} +{"timestamp": "2026-07-16T21:15:29+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 16384, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4772.803886972835, "last_total_tps": 1759.8744462055809} +{"timestamp": "2026-07-16T21:26:50+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 16384, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4137.621478884017, "last_total_tps": 1033.4981159438357} +{"timestamp": "2026-07-16T21:42:20+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 16384, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 3290.996804492577, "last_total_tps": 577.0193960995003} +{"timestamp": "2026-07-16T22:06:10+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 16384, "osl": 2048, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_BACKOFF_FOUND", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 2399.4052794078843, "last_total_tps": 319.32407850271136} +{"timestamp": "2026-07-16T22:38:22+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 65536, "osl": 128, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4511.33953037921, "last_total_tps": 4289.041351148642} +{"timestamp": "2026-07-16T23:11:31+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 65536, "osl": 256, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4462.727107186744, "last_total_tps": 3484.7836465147443} +{"timestamp": "2026-07-16T23:46:51+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 65536, "osl": 512, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4296.586416662889, "last_total_tps": 2530.9149576425034} +{"timestamp": "2026-07-17T00:26:29+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 65536, "osl": 1024, "status": "TTFT_BOUNDARY", "stop_reason": "TTFT_SLO_EXCEEDED", "tested_points": 3, "search_cap": 512, "max_successful_concurrency": 1, "saturation_concurrency": null, "stop_probe_concurrency": 1, "best_tps_concurrency": 16, "best_total_tps": 4002.8258369033274, "last_total_tps": 1651.9047329955827} +{"timestamp": "2026-07-17T01:02:05+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 131072, "osl": 128, "status": "FAILED", "stop_reason": "BENCHMARK_FAILED", "tested_points": 1, "search_cap": 512, "max_successful_concurrency": 0, "saturation_concurrency": null, "stop_probe_concurrency": 16, "best_tps_concurrency": null, "best_total_tps": null, "last_total_tps": null} +{"timestamp": "2026-07-17T02:10:35+00:00", "engine": "sglang", "tp": 8, "dp": 1, "mark": "Y", "isl": 131072, "osl": 256, "status": "FAILED", "stop_reason": "BENCHMARK_FAILED", "tested_points": 1, "search_cap": 512, "max_successful_concurrency": 0, "saturation_concurrency": null, "stop_probe_concurrency": 16, "best_tps_concurrency": null, "best_total_tps": null, "last_total_tps": null} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.md b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.md new file mode 100644 index 0000000..a7e754e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/adaptive_summary.md @@ -0,0 +1,53 @@ +# Adaptive concurrency search summary + +| Engine | TP | DP | ISL | OSL | Stop | Saturation C | Best TPS C | Best Total TPS | Max successful C | +|---|---:|---:|---:|---:|---|---:|---:|---:|---:| +| sglang | 4 | 2 | 1024 | 128 | TTFT_SLO_EXCEEDED | None | 16 | 2233.1625760825486 | 16 | +| sglang | 4 | 2 | 1024 | 256 | TTFT_SLO_EXCEEDED | None | 48 | 2994.9145225702982 | 48 | +| sglang | 4 | 2 | 1024 | 512 | TTFT_SLO_EXCEEDED | None | 48 | 1894.263479895674 | 48 | +| sglang | 4 | 2 | 1024 | 1024 | TTFT_SLO_EXCEEDED | None | 48 | 1270.3578826082223 | 48 | +| sglang | 4 | 2 | 1024 | 2048 | TTFT_SLO_EXCEEDED | None | 48 | 938.3875354487201 | 48 | +| sglang | 4 | 2 | 1024 | 4096 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 487.1207589944629 | 8 | +| sglang | 4 | 2 | 4096 | 128 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 5276.430317217383 | 8 | +| sglang | 4 | 2 | 4096 | 256 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 3654.27555715578 | 8 | +| sglang | 4 | 2 | 4096 | 512 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 2302.8342327716664 | 8 | +| sglang | 4 | 2 | 4096 | 1024 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 1449.0672823499715 | 8 | +| sglang | 4 | 2 | 4096 | 2048 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 919.5235579896186 | 8 | +| sglang | 4 | 2 | 4096 | 4096 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 636.5331301353165 | 8 | +| sglang | 4 | 2 | 16384 | 128 | TTFT_SLO_EXCEEDED | None | 16 | 6149.069264131796 | 1 | +| sglang | 4 | 2 | 16384 | 256 | TTFT_SLO_EXCEEDED | None | 16 | 4922.27619638361 | 1 | +| sglang | 4 | 2 | 16384 | 512 | TTFT_SLO_EXCEEDED | None | 16 | 3457.304361342664 | 1 | +| sglang | 4 | 2 | 16384 | 1024 | TTFT_SLO_EXCEEDED | None | 16 | 2203.3814838524254 | 1 | +| sglang | 4 | 2 | 16384 | 2048 | TTFT_SLO_EXCEEDED | None | 16 | 1322.1810796622076 | 1 | +| sglang | 4 | 2 | 65536 | 128 | TTFT_SLO_EXCEEDED | None | 16 | 6860.104148991479 | 1 | +| sglang | 4 | 2 | 65536 | 256 | TTFT_SLO_EXCEEDED | None | 16 | 6322.324656264061 | 1 | +| sglang | 4 | 2 | 65536 | 512 | TTFT_SLO_EXCEEDED | None | 16 | 5426.021563663019 | 1 | +| sglang | 4 | 2 | 65536 | 1024 | TTFT_SLO_EXCEEDED | None | 8 | 4250.026326552517 | 1 | +| sglang | 4 | 2 | 131072 | 128 | TTFT_SLO_EXCEEDED | None | 8 | 6414.4163774899025 | 1 | +| sglang | 4 | 2 | 131072 | 256 | BENCHMARK_FAILED | None | None | None | 0 | +| sglang | 4 | 2 | 131072 | 512 | BENCHMARK_FAILED | None | None | None | 0 | +| sglang | 8 | 1 | 1024 | 128 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 2252.3024943648124 | 8 | +| sglang | 8 | 1 | 1024 | 256 | TTFT_SLO_EXCEEDED | None | 32 | 2807.7871232493126 | 32 | +| sglang | 8 | 1 | 1024 | 512 | TTFT_SLO_EXCEEDED | None | 32 | 2109.996176616557 | 48 | +| sglang | 8 | 1 | 1024 | 1024 | TTFT_SLO_EXCEEDED | None | 32 | 1390.4790149026294 | 32 | +| sglang | 8 | 1 | 1024 | 2048 | TTFT_SLO_EXCEEDED | None | 32 | 1079.9126577555764 | 32 | +| sglang | 8 | 1 | 1024 | 4096 | BENCHMARK_FAILED | None | 32 | 911.6872643258514 | 32 | +| sglang | 8 | 1 | 4096 | 128 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 4128.998328888893 | 1 | +| sglang | 8 | 1 | 4096 | 256 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 4421.1758211534125 | 1 | +| sglang | 8 | 1 | 4096 | 512 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 2909.7747901626167 | 1 | +| sglang | 8 | 1 | 4096 | 1024 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 1863.3608864757555 | 1 | +| sglang | 8 | 1 | 4096 | 2048 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 1206.8940867936158 | 1 | +| sglang | 8 | 1 | 4096 | 4096 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 838.0690279850638 | 1 | +| sglang | 8 | 1 | 16384 | 128 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 5144.723678168144 | 1 | +| sglang | 8 | 1 | 16384 | 256 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 4772.803886972835 | 1 | +| sglang | 8 | 1 | 16384 | 512 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 4137.621478884017 | 1 | +| sglang | 8 | 1 | 16384 | 1024 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 3290.996804492577 | 1 | +| sglang | 8 | 1 | 16384 | 2048 | TTFT_SLO_BACKOFF_FOUND | None | 16 | 2399.4052794078843 | 1 | +| sglang | 8 | 1 | 65536 | 128 | TTFT_SLO_EXCEEDED | None | 16 | 4511.33953037921 | 1 | +| sglang | 8 | 1 | 65536 | 256 | TTFT_SLO_EXCEEDED | None | 16 | 4462.727107186744 | 1 | +| sglang | 8 | 1 | 65536 | 512 | TTFT_SLO_EXCEEDED | None | 16 | 4296.586416662889 | 1 | +| sglang | 8 | 1 | 65536 | 1024 | TTFT_SLO_EXCEEDED | None | 16 | 4002.8258369033274 | 1 | +| sglang | 8 | 1 | 131072 | 128 | BENCHMARK_FAILED | None | None | None | 0 | +| sglang | 8 | 1 | 131072 | 256 | BENCHMARK_FAILED | None | None | None | 0 | + +`Saturation C` is the first point in the final low-gain streak. `Best TPS C` is the tested point with the highest observed Total TPS. diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/run_manifest.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/run_manifest.json index 0e9d63c..767b099 100644 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/run_manifest.json +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/run_manifest.json @@ -12,11 +12,16 @@ "start_concurrency": 16, "max_concurrency": 512, "multiplier": 2, + "initial_backoff_concurrencies": [ + 8, + 1 + ], "num_prompts_multiplier": 5, "min_tps_gain_pct": 2.0, "plateau_patience": 2, "warmup_max_requests": 0, "ttft_slo_ms": 4000.0, - "enable_ttft_slo_stop": 1 + "enable_ttft_slo_stop": 1, + "ttft_group_skip_ms": 8000.0 } } diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o1024_a1.json new file mode 100644 index 0000000..b0bd6f6 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 239.2404650580138, + "request_tps": 0.3343915920770371, + "input_tps": 342.416990286886, + "output_tps": 342.416990286886, + "total_tps": 684.833980573772, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 635.5562334647402, + "ttft_p95_ms": 1903.8669666741043, + "ttft_p99_ms": 1906.7863759305328, + "tpot_p50_ms": 45.73389088913875, + "tpot_p95_ms": 46.81248970833659, + "tpot_p99_ms": 47.236905306400615, + "e2e_p50_ms": 47589.559288928285, + "e2e_p95_ms": 48763.24818015564, + "e2e_p99_ms": 49025.48799880082, + "itl_p50_ms": 38.22040546219796, + "itl_p95_ms": 62.13192297145724, + "itl_p99_ms": 69.20204504393043, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o2048_a1.json new file mode 100644 index 0000000..ca2b03f --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 473.66198480315506, + "request_tps": 0.16889681369140588, + "input_tps": 172.95033721999962, + "output_tps": 345.90067443999925, + "total_tps": 518.8510116599988, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 1107.0592959877104, + "ttft_p95_ms": 1914.7031036205587, + "ttft_p99_ms": 1937.422300640028, + "tpot_p50_ms": 45.60600052494192, + "tpot_p95_ms": 46.180353910758, + "tpot_p99_ms": 46.185254772231495, + "e2e_p50_ms": 94571.73220254481, + "e2e_p95_ms": 95557.28795382893, + "e2e_p99_ms": 95606.2740921299, + "itl_p50_ms": 38.73952745925635, + "itl_p95_ms": 62.4737826641649, + "itl_p99_ms": 67.41200149524958, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o256_a1.json new file mode 100644 index 0000000..3ee71a1 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 67.59083261387423, + "request_tps": 1.1835924622650464, + "input_tps": 1211.9986813594076, + "output_tps": 302.9996703398519, + "total_tps": 1514.9983516992595, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 1088.1021856330335, + "ttft_p95_ms": 1940.051195200067, + "ttft_p99_ms": 1945.7215382088907, + "tpot_p50_ms": 48.00720819381668, + "tpot_p95_ms": 51.906457407692194, + "tpot_p99_ms": 52.662985878384795, + "e2e_p50_ms": 13323.23444657959, + "e2e_p95_ms": 14546.717717882711, + "e2e_p99_ms": 14597.288672290742, + "itl_p50_ms": 38.109477958641946, + "itl_p95_ms": 62.32210020534694, + "itl_p99_ms": 91.81069683981994, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o4096_a1.json new file mode 100644 index 0000000..2ad5b3e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 840.8592580729164, + "request_tps": 0.09514077324110604, + "input_tps": 97.42415179889258, + "output_tps": 389.6966071955703, + "total_tps": 487.1207589944629, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 972.442637081258, + "ttft_p95_ms": 5176.270531245973, + "ttft_p99_ms": 5179.259422505274, + "tpot_p50_ms": 40.55560706117929, + "tpot_p95_ms": 40.93768200460964, + "tpot_p99_ms": 40.989130662076434, + "e2e_p50_ms": 167707.99649693072, + "e2e_p95_ms": 171209.72607890144, + "e2e_p99_ms": 171213.10957846232, + "itl_p50_ms": 38.21910894475877, + "itl_p95_ms": 56.424582388717674, + "itl_p99_ms": 61.033719554543495, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o512_a1.json new file mode 100644 index 0000000..6824bd0 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 123.35641185496934, + "request_tps": 0.6485272941795385, + "input_tps": 664.0919492398474, + "output_tps": 332.0459746199237, + "total_tps": 996.137923859771, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 1815.1043795514852, + "ttft_p95_ms": 2051.647847087588, + "ttft_p99_ms": 2054.4600446219556, + "tpot_p50_ms": 44.89284182270832, + "tpot_p95_ms": 47.29539701357892, + "tpot_p99_ms": 47.49259979092314, + "e2e_p50_ms": 24692.195138079114, + "e2e_p95_ms": 25352.763180364855, + "e2e_p99_ms": 25618.341889507137, + "itl_p50_ms": 37.906087934970856, + "itl_p95_ms": 61.653098394162946, + "itl_p99_ms": 66.99016668368131, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i131072_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i131072_o128_a1.json new file mode 100644 index 0000000..f021967 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i131072_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1677.9406901570037, + "request_tps": 0.047677489716584955, + "input_tps": 6249.183932132223, + "output_tps": 6.102718683722874, + "total_tps": 6255.286650815946, + "mean_input_tokens": 131072.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 300625.9035965195, + "ttft_p95_ms": 338951.3012407697, + "ttft_p99_ms": 340203.0271626217, + "tpot_p50_ms": 41.211593476322115, + "tpot_p95_ms": 345.65387451581245, + "tpot_p99_ms": 346.94827124932925, + "e2e_p50_ms": 343158.3466755692, + "e2e_p95_ms": 344526.5441220952, + "e2e_p99_ms": 345839.6717084944, + "itl_p50_ms": 33.60768407583237, + "itl_p95_ms": 58.138947119005024, + "itl_p99_ms": 181.5207198727876, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o1024_a1.json new file mode 100644 index 0000000..a6a3fc5 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 632.0467019469943, + "request_tps": 0.1265729253132138, + "input_tps": 2073.7708083316948, + "output_tps": 129.61067552073092, + "total_tps": 2203.3814838524254, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 80126.8644919619, + "ttft_p95_ms": 106781.9698160165, + "ttft_p99_ms": 107180.66528041381, + "tpot_p50_ms": 38.308278328050854, + "tpot_p95_ms": 42.30265493578899, + "tpot_p99_ms": 45.4902937474118, + "e2e_p50_ms": 120186.66901800316, + "e2e_p95_ms": 148021.29719536752, + "e2e_p99_ms": 148871.62709633118, + "itl_p50_ms": 34.52094295062125, + "itl_p95_ms": 51.694627082906685, + "itl_p99_ms": 59.966907631605864, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o128_a1.json new file mode 100644 index 0000000..ec28e7c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 214.8227549989242, + "request_tps": 0.37240002810875705, + "input_tps": 6101.402060533876, + "output_tps": 47.6672035979209, + "total_tps": 6149.069264131796, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 33368.671427015215, + "ttft_p95_ms": 41899.411271174904, + "ttft_p99_ms": 42125.512621703085, + "tpot_p50_ms": 52.554487114163535, + "tpot_p95_ms": 84.39758658330287, + "tpot_p99_ms": 111.94620711651932, + "e2e_p50_ms": 42495.81566103734, + "e2e_p95_ms": 49210.346029268105, + "e2e_p99_ms": 52561.92248010076, + "itl_p50_ms": 34.52568908687681, + "itl_p95_ms": 52.17893534572764, + "itl_p99_ms": 66.4443611598108, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o2048_a1.json new file mode 100644 index 0000000..40c3b2d --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1115.2481476869434, + "request_tps": 0.07173291447820136, + "input_tps": 1175.2720708108511, + "output_tps": 146.9090088513564, + "total_tps": 1322.1810796622076, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 136037.42311848328, + "ttft_p95_ms": 181299.1330652847, + "ttft_p99_ms": 181444.37291203765, + "tpot_p50_ms": 37.30771111527574, + "tpot_p95_ms": 39.328867612731, + "tpot_p99_ms": 40.83525040460833, + "e2e_p50_ms": 213226.2396500446, + "e2e_p95_ms": 259703.93927409317, + "e2e_p99_ms": 260337.14725418712, + "itl_p50_ms": 34.49960285797715, + "itl_p95_ms": 51.94193584611639, + "itl_p99_ms": 60.15488663455472, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o256_a1.json new file mode 100644 index 0000000..776927c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 270.44398706802167, + "request_tps": 0.2958098675711304, + "input_tps": 4846.548870285401, + "output_tps": 75.72732609820939, + "total_tps": 4922.27619638361, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 38130.55603799876, + "ttft_p95_ms": 51039.480149233714, + "ttft_p99_ms": 51170.22580615245, + "tpot_p50_ms": 44.34177505135463, + "tpot_p95_ms": 60.19466479852174, + "tpot_p99_ms": 74.01323430483028, + "e2e_p50_ms": 50331.59309858456, + "e2e_p95_ms": 65593.75085806241, + "e2e_p99_ms": 65883.99838647107, + "itl_p50_ms": 34.52300652861595, + "itl_p95_ms": 51.78628868889062, + "itl_p99_ms": 60.336364856921136, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o512_a1.json new file mode 100644 index 0000000..3358e2a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 390.9635538929142, + "request_tps": 0.20462265396204216, + "input_tps": 3352.5375625140987, + "output_tps": 104.76679882856558, + "total_tps": 3457.304361342664, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 52149.89124494605, + "ttft_p95_ms": 69626.60835890565, + "ttft_p99_ms": 69753.87210340472, + "tpot_p50_ms": 40.293584739718284, + "tpot_p95_ms": 48.141768489123194, + "tpot_p99_ms": 54.960999020335215, + "e2e_p50_ms": 73592.7937864326, + "e2e_p95_ms": 93283.15660590306, + "e2e_p99_ms": 93629.28111733867, + "itl_p50_ms": 34.522912465035915, + "itl_p95_ms": 51.489772961940616, + "itl_p99_ms": 59.926832308992736, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o1024_a1.json new file mode 100644 index 0000000..d5ed5cd --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 282.66458361805417, + "request_tps": 0.2830209535839788, + "input_tps": 1159.253825879977, + "output_tps": 289.81345646999426, + "total_tps": 1449.0672823499715, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 4658.318318077363, + "ttft_p95_ms": 10095.44171601301, + "ttft_p99_ms": 12803.387446557112, + "tpot_p50_ms": 44.42285834210774, + "tpot_p95_ms": 49.33753012539997, + "tpot_p99_ms": 79.11847314414067, + "e2e_p50_ms": 49335.88792406954, + "e2e_p95_ms": 58133.938353240934, + "e2e_p99_ms": 87081.42475777536, + "itl_p50_ms": 38.47858391236514, + "itl_p95_ms": 57.68822985701263, + "itl_p99_ms": 62.88705667946487, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o128_a1.json new file mode 100644 index 0000000..c69ff84 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 64.04329815506935, + "request_tps": 1.2491549046442667, + "input_tps": 5116.538489422916, + "output_tps": 159.89182779446614, + "total_tps": 5276.430317217383, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 4576.742604491301, + "ttft_p95_ms": 7486.383566202128, + "ttft_p99_ms": 8069.572048275731, + "tpot_p50_ms": 63.25949257419364, + "tpot_p95_ms": 88.51012927140879, + "tpot_p99_ms": 93.78976958789158, + "e2e_p50_ms": 12597.990370937623, + "e2e_p95_ms": 16119.662079052066, + "e2e_p99_ms": 16428.539164408576, + "itl_p50_ms": 38.49607554730028, + "itl_p95_ms": 59.04390099458394, + "itl_p99_ms": 859.1562826023437, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o2048_a1.json new file mode 100644 index 0000000..33a289b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 534.5376915351953, + "request_tps": 0.14966203743320614, + "input_tps": 613.0157053264123, + "output_tps": 306.50785266320617, + "total_tps": 919.5235579896186, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 5104.571636882611, + "ttft_p95_ms": 13598.845852899827, + "ttft_p99_ms": 15530.32725170254, + "tpot_p50_ms": 42.516595378821116, + "tpot_p95_ms": 48.41014748285291, + "tpot_p99_ms": 80.66643662295293, + "e2e_p50_ms": 90892.7560900338, + "e2e_p95_ms": 106289.35107835554, + "e2e_p99_ms": 171094.83309885245, + "itl_p50_ms": 38.54017308913171, + "itl_p95_ms": 58.33887221524492, + "itl_p99_ms": 64.16293417569251, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o256_a1.json new file mode 100644 index 0000000..350b715 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 95.2746979680378, + "request_tps": 0.839677287949398, + "input_tps": 3439.318171440734, + "output_tps": 214.9573857150459, + "total_tps": 3654.27555715578, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 4683.939669514075, + "ttft_p95_ms": 8062.7326244954, + "ttft_p99_ms": 8673.460278843995, + "tpot_p50_ms": 51.88298721149491, + "tpot_p95_ms": 64.92892745028561, + "tpot_p99_ms": 75.71912046924533, + "e2e_p50_ms": 17954.677841393277, + "e2e_p95_ms": 22523.24340294581, + "e2e_p99_ms": 24458.481649178953, + "itl_p50_ms": 38.46407646778971, + "itl_p95_ms": 59.771430026739836, + "itl_p99_ms": 67.52078937599435, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o4096_a1.json new file mode 100644 index 0000000..488091b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1029.577203406021, + "request_tps": 0.07770179811222125, + "input_tps": 318.26656506765823, + "output_tps": 318.26656506765823, + "total_tps": 636.5331301353165, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 5483.361787861213, + "ttft_p95_ms": 22374.68960867263, + "ttft_p99_ms": 150920.9240530175, + "tpot_p50_ms": 41.06715219183478, + "tpot_p95_ms": 46.9445546033481, + "tpot_p99_ms": 79.17070015191274, + "e2e_p50_ms": 173329.3942649616, + "e2e_p95_ms": 302973.24356382014, + "e2e_p99_ms": 330154.8701896191, + "itl_p50_ms": 38.36022887844592, + "itl_p95_ms": 57.20986550441011, + "itl_p99_ms": 62.143948464654386, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o512_a1.json new file mode 100644 index 0000000..a36feb2 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i4096_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 160.08099704002962, + "request_tps": 0.4997470123202401, + "input_tps": 2046.9637624637035, + "output_tps": 255.87047030796293, + "total_tps": 2302.8342327716664, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 5474.024421186186, + "ttft_p95_ms": 8615.011906716973, + "ttft_p99_ms": 8951.299454767715, + "tpot_p50_ms": 46.494569923772445, + "tpot_p95_ms": 53.43938896643479, + "tpot_p99_ms": 76.29711307971617, + "e2e_p50_ms": 28335.586922592483, + "e2e_p95_ms": 32704.220330703534, + "e2e_p99_ms": 44918.695579937164, + "itl_p50_ms": 38.48353505600244, + "itl_p95_ms": 59.288766619283706, + "itl_p99_ms": 65.03905458841473, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o1024_a1.json new file mode 100644 index 0000000..f7ff19e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1253.2045440499205, + "request_tps": 0.0638363468915201, + "input_tps": 4183.578829882661, + "output_tps": 65.36841921691658, + "total_tps": 4248.947249099578, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 202414.35351443943, + "ttft_p95_ms": 223751.87583953375, + "ttft_p99_ms": 226827.1145781991, + "tpot_p50_ms": 53.94717318803356, + "tpot_p95_ms": 71.33589557350184, + "tpot_p99_ms": 86.74920972760991, + "e2e_p50_ms": 247910.9200029634, + "e2e_p95_ms": 278792.46321712853, + "e2e_p99_ms": 292009.2727964232, + "itl_p50_ms": 36.60361119545996, + "itl_p95_ms": 54.5408145524561, + "itl_p99_ms": 62.74137625936419, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o128_a1.json new file mode 100644 index 0000000..c919b65 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 765.7493072859943, + "request_tps": 0.1044728336530135, + "input_tps": 6846.731626283893, + "output_tps": 13.372522707585729, + "total_tps": 6860.104148991479, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 128567.64044612646, + "ttft_p95_ms": 151961.6587853641, + "ttft_p99_ms": 156960.82069837945, + "tpot_p50_ms": 161.96727417372753, + "tpot_p95_ms": 305.1854889790388, + "tpot_p99_ms": 437.706828691014, + "e2e_p50_ms": 146586.26923209522, + "e2e_p95_ms": 179067.64789508888, + "e2e_p99_ms": 186050.76994073341, + "itl_p50_ms": 36.512586986646056, + "itl_p95_ms": 53.31773438956598, + "itl_p99_ms": 61.62503688596189, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o256_a1.json new file mode 100644 index 0000000..6cebf4a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 832.5039105331525, + "request_tps": 0.09609564470245716, + "input_tps": 6297.724171220232, + "output_tps": 24.600485043829032, + "total_tps": 6322.324656264061, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 136335.5990794953, + "ttft_p95_ms": 162983.08086185716, + "ttft_p99_ms": 166663.45600951457, + "tpot_p50_ms": 100.12674616266263, + "tpot_p95_ms": 175.98028106304506, + "tpot_p99_ms": 236.80305566614055, + "e2e_p50_ms": 156982.1368450066, + "e2e_p95_ms": 189405.96394817112, + "e2e_p99_ms": 206104.3898640247, + "itl_p50_ms": 36.563907051458955, + "itl_p95_ms": 53.42142162844535, + "itl_p99_ms": 62.24962119013072, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o512_a1.json new file mode 100644 index 0000000..64630a7 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c16_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 973.7963511580601, + "request_tps": 0.08215270051573127, + "input_tps": 5383.959380998965, + "output_tps": 42.06218266405441, + "total_tps": 5426.021563663019, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 153779.00782134384, + "ttft_p95_ms": 181627.71489518927, + "ttft_p99_ms": 183322.4648270011, + "tpot_p50_ms": 69.34998192962937, + "tpot_p95_ms": 106.79924657778838, + "tpot_p99_ms": 136.58104994009412, + "e2e_p50_ms": 189005.7152614463, + "e2e_p95_ms": 234631.01633958286, + "e2e_p99_ms": 236243.1676932983, + "itl_p50_ms": 36.63114202208817, + "itl_p95_ms": 53.81391057744622, + "itl_p99_ms": 62.73157489486038, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i131072_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i131072_o128_a1.json new file mode 100644 index 0000000..1e8db1b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i131072_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 226.54693996487185, + "request_tps": 0.022070481290876385, + "input_tps": 2892.8221237577495, + "output_tps": 2.8250216052321773, + "total_tps": 2895.647145362982, + "mean_input_tokens": 131072.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 40983.879873063415, + "ttft_p95_ms": 41427.96418489888, + "ttft_p99_ms": 41483.68903929368, + "tpot_p50_ms": 33.28039894415814, + "tpot_p95_ms": 33.693099089744166, + "tpot_p99_ms": 33.698522558318466, + "e2e_p50_ms": 45259.45955491625, + "e2e_p95_ms": 45639.66128444299, + "e2e_p99_ms": 45701.79939834401, + "itl_p50_ms": 30.720365000888705, + "itl_p95_ms": 55.317965033464134, + "itl_p99_ms": 62.53650498110801, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o1024_a1.json new file mode 100644 index 0000000..4aa8bea --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 181.57242772518657, + "request_tps": 0.02753722061571814, + "input_tps": 451.169822567926, + "output_tps": 28.198113910495376, + "total_tps": 479.3679364784214, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 3940.7668448984623, + "ttft_p95_ms": 4193.481011828408, + "ttft_p99_ms": 4228.460800005123, + "tpot_p50_ms": 31.63807294540879, + "tpot_p95_ms": 31.730153837458815, + "tpot_p99_ms": 31.734481836830287, + "e2e_p50_ms": 36320.72346494533, + "e2e_p95_ms": 36463.93566392362, + "e2e_p99_ms": 36480.87247610092, + "itl_p50_ms": 29.840212780982256, + "itl_p95_ms": 46.52133162599058, + "itl_p99_ms": 54.09107577055689, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o128_a1.json new file mode 100644 index 0000000..ebbe95c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 39.83333700802177, + "request_tps": 0.12552300097260452, + "input_tps": 2056.5688479351525, + "output_tps": 16.06694412449338, + "total_tps": 2072.6357920596456, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 3952.2010260261595, + "ttft_p95_ms": 4079.302547173574, + "ttft_p99_ms": 4102.609633402899, + "tpot_p50_ms": 31.602383598567936, + "tpot_p95_ms": 31.640485883140425, + "tpot_p99_ms": 31.642765031291507, + "e2e_p50_ms": 7969.095474109054, + "e2e_p95_ms": 8074.662221223116, + "e2e_p99_ms": 8093.300537876784, + "itl_p50_ms": 29.8505169339478, + "itl_p95_ms": 49.705708399415016, + "itl_p99_ms": 51.2071222346276, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o2048_a1.json new file mode 100644 index 0000000..a323701 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 342.39226463483647, + "request_tps": 0.014603133646528293, + "input_tps": 239.25774166471956, + "output_tps": 29.907217708089945, + "total_tps": 269.1649593728095, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 3971.6174241621047, + "ttft_p95_ms": 4094.661511434242, + "ttft_p99_ms": 4115.855502327904, + "tpot_p50_ms": 31.558119324293, + "tpot_p95_ms": 31.620720006947725, + "tpot_p99_ms": 31.632591595734013, + "e2e_p50_ms": 68487.32287203893, + "e2e_p95_ms": 68711.56300739385, + "e2e_p99_ms": 68739.65807267465, + "itl_p50_ms": 29.78591388091445, + "itl_p95_ms": 45.330130099318914, + "itl_p99_ms": 55.03684663679451, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o256_a1.json new file mode 100644 index 0000000..be6b00e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 60.537773119984195, + "request_tps": 0.08259306119652168, + "input_tps": 1353.2047146438113, + "output_tps": 21.14382366630955, + "total_tps": 1374.3485383101208, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 3945.7851310726255, + "ttft_p95_ms": 4093.810023879632, + "ttft_p99_ms": 4111.218916652724, + "tpot_p50_ms": 31.821994341033346, + "tpot_p95_ms": 32.01389185737307, + "tpot_p99_ms": 32.03676525156433, + "e2e_p50_ms": 12080.163977108896, + "e2e_p95_ms": 12253.15336487256, + "e2e_p99_ms": 12279.754239274189, + "itl_p50_ms": 29.860365903005004, + "itl_p95_ms": 48.98134677205235, + "itl_p99_ms": 55.7887158729136, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o512_a1.json new file mode 100644 index 0000000..9c1662c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 100.54988616588525, + "request_tps": 0.04972656052291394, + "input_tps": 814.719967607422, + "output_tps": 25.459998987731936, + "total_tps": 840.1799665951539, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 3932.124942075461, + "ttft_p95_ms": 4127.591159660369, + "ttft_p99_ms": 4166.129624564201, + "tpot_p50_ms": 31.635086914185443, + "tpot_p95_ms": 31.776042928674975, + "tpot_p99_ms": 31.788027078066378, + "e2e_p50_ms": 20141.837269999087, + "e2e_p95_ms": 20233.498825039715, + "e2e_p99_ms": 20244.73103923723, + "itl_p50_ms": 29.790908098220825, + "itl_p95_ms": 50.02487420570105, + "itl_p99_ms": 54.69925904646516, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o1024_a1.json new file mode 100644 index 0000000..18a612e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 256.18154512299225, + "request_tps": 0.019517409021791596, + "input_tps": 1279.092917652134, + "output_tps": 19.985826838314594, + "total_tps": 1299.0787444904488, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 17658.26961491257, + "ttft_p95_ms": 18013.91982375644, + "ttft_p99_ms": 18047.10275827907, + "tpot_p50_ms": 32.7376385434701, + "tpot_p95_ms": 32.79802809666955, + "tpot_p99_ms": 32.798152351014515, + "e2e_p50_ms": 51210.01679683104, + "e2e_p95_ms": 51457.70765142515, + "e2e_p99_ms": 51481.52730548754, + "itl_p50_ms": 30.361955054104328, + "itl_p95_ms": 53.69571931660175, + "itl_p99_ms": 60.88819689117371, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o128_a1.json new file mode 100644 index 0000000..b4118f8 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 109.37776633584872, + "request_tps": 0.045713129528055126, + "input_tps": 2995.8556567506207, + "output_tps": 5.851280579591056, + "total_tps": 3001.706937330212, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 17729.769844096154, + "ttft_p95_ms": 17954.093989543617, + "ttft_p99_ms": 17964.415912255645, + "tpot_p50_ms": 32.0961656301891, + "tpot_p95_ms": 32.58936951105578, + "tpot_p99_ms": 32.640932479609305, + "e2e_p50_ms": 21835.877286968753, + "e2e_p95_ms": 22044.52510732226, + "e2e_p99_ms": 22048.520961580798, + "itl_p50_ms": 30.422291019931436, + "itl_p95_ms": 47.328294953331344, + "itl_p99_ms": 56.53701915405691, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o256_a1.json new file mode 100644 index 0000000..9b95d7a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 129.80575013905764, + "request_tps": 0.038519094837044014, + "input_tps": 2524.3873992405165, + "output_tps": 9.860888278283268, + "total_tps": 2534.2482875187998, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 17821.7026530765, + "ttft_p95_ms": 17941.041450900957, + "ttft_p99_ms": 17956.14996048622, + "tpot_p50_ms": 32.05825664723913, + "tpot_p95_ms": 32.285422013122954, + "tpot_p99_ms": 32.28693708328201, + "e2e_p50_ms": 26052.55355197005, + "e2e_p95_ms": 26127.57891532965, + "e2e_p99_ms": 26133.34180940874, + "itl_p50_ms": 30.41153005324304, + "itl_p95_ms": 46.34950009640306, + "itl_p99_ms": 56.67572288773954, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o512_a1.json new file mode 100644 index 0000000..279a883 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c1_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 170.8164885980077, + "request_tps": 0.029271178918604212, + "input_tps": 1918.3159816096456, + "output_tps": 14.986843606325357, + "total_tps": 1933.302825215971, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 17657.30693913065, + "ttft_p95_ms": 17947.263365006074, + "ttft_p99_ms": 17974.353000214323, + "tpot_p50_ms": 32.16144313886325, + "tpot_p95_ms": 32.28488007000849, + "tpot_p99_ms": 32.28917933327971, + "e2e_p50_ms": 34091.804383089766, + "e2e_p95_ms": 34388.406378310174, + "e2e_p99_ms": 34406.406796555966, + "itl_p50_ms": 30.36797302775085, + "itl_p95_ms": 46.908468334004276, + "itl_p99_ms": 55.469688819721355, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o1024_a1.json new file mode 100644 index 0000000..248f358 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 283.1469709521625, + "request_tps": 0.5650775618822773, + "input_tps": 578.639423367452, + "output_tps": 578.639423367452, + "total_tps": 1157.278846734904, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 1039.5908689824864, + "ttft_p95_ms": 3708.7073334376323, + "ttft_p99_ms": 3736.5557909896597, + "tpot_p50_ms": 53.18489079777869, + "tpot_p95_ms": 56.531733996098275, + "tpot_p99_ms": 56.90662554211692, + "e2e_p50_ms": 55562.747771968134, + "e2e_p95_ms": 58622.54847800359, + "e2e_p99_ms": 58864.768436276354, + "itl_p50_ms": 44.60177046712488, + "itl_p95_ms": 68.54137406917289, + "itl_p99_ms": 104.98853123513977, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o2048_a1.json new file mode 100644 index 0000000..6b099ab --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 562.1656430249568, + "request_tps": 0.28461362231077675, + "input_tps": 291.4443492462354, + "output_tps": 582.8886984924708, + "total_tps": 874.3330477387061, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 1720.5767229897901, + "ttft_p95_ms": 3765.552905423101, + "ttft_p99_ms": 3778.2479317323305, + "tpot_p50_ms": 53.25908665974094, + "tpot_p95_ms": 55.20441076613287, + "tpot_p99_ms": 55.39557766710268, + "e2e_p50_ms": 109979.57844298799, + "e2e_p95_ms": 114323.36549273459, + "e2e_p99_ms": 114707.30229713488, + "itl_p50_ms": 46.27890046685934, + "itl_p95_ms": 69.01592136127874, + "itl_p99_ms": 77.31038637226447, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o256_a1.json new file mode 100644 index 0000000..42543c1 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 85.51100934692658, + "request_tps": 1.8711040978462112, + "input_tps": 1916.0105961945203, + "output_tps": 479.0026490486301, + "total_tps": 2395.0132452431503, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 1161.2384765176103, + "ttft_p95_ms": 3756.2352412380274, + "ttft_p99_ms": 3780.965418519918, + "tpot_p50_ms": 61.98459189412568, + "tpot_p95_ms": 66.44078092480663, + "tpot_p99_ms": 66.8051291848259, + "e2e_p50_ms": 16901.625977130607, + "e2e_p95_ms": 18837.460162944622, + "e2e_p99_ms": 18985.462763826363, + "itl_p50_ms": 44.66060339473188, + "itl_p95_ms": 68.57059714384377, + "itl_p99_ms": 449.1709415311925, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o512_a1.json new file mode 100644 index 0000000..bb5d004 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c32_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 151.8317316111643, + "request_tps": 1.0537981639421354, + "input_tps": 1079.0893198767467, + "output_tps": 539.5446599383733, + "total_tps": 1618.63397981512, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 1056.6700529307127, + "ttft_p95_ms": 3730.0430950359428, + "ttft_p99_ms": 3803.3011273038574, + "tpot_p50_ms": 55.7243038511538, + "tpot_p95_ms": 60.350505806735306, + "tpot_p99_ms": 61.641385694806246, + "e2e_p50_ms": 29571.674760896713, + "e2e_p95_ms": 32270.35216266522, + "e2e_p99_ms": 32279.17558920337, + "itl_p50_ms": 45.009728986769915, + "itl_p95_ms": 67.9298322298564, + "itl_p99_ms": 236.92634512903186, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o1024_a1.json new file mode 100644 index 0000000..d979b73 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 240, + "failed": 0, + "duration_s": 386.91459054895677, + "request_tps": 0.6202919348672961, + "input_tps": 635.1789413041112, + "output_tps": 635.1789413041112, + "total_tps": 1270.3578826082223, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 5370.0150235090405, + "ttft_p95_ms": 15133.30938139696, + "ttft_p99_ms": 32185.267814283718, + "tpot_p50_ms": 60.0038268968676, + "tpot_p95_ms": 91.36051824342555, + "tpot_p99_ms": 103.14346713503731, + "e2e_p50_ms": 65575.55996044539, + "e2e_p95_ms": 99691.6358396411, + "e2e_p99_ms": 112446.33609429697, + "itl_p50_ms": 51.48644756991416, + "itl_p95_ms": 73.74423729488626, + "itl_p99_ms": 83.97575234295799, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o2048_a1.json new file mode 100644 index 0000000..99e22ea --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 240, + "failed": 0, + "duration_s": 785.6881854759995, + "request_tps": 0.3054646925288802, + "input_tps": 312.7958451495733, + "output_tps": 625.5916902991466, + "total_tps": 938.3875354487201, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 5525.519606540911, + "ttft_p95_ms": 38321.421813906585, + "ttft_p99_ms": 61896.90983353647, + "tpot_p50_ms": 60.786382354451064, + "tpot_p95_ms": 94.29399440951978, + "tpot_p99_ms": 106.96621055514711, + "e2e_p50_ms": 128179.90333400667, + "e2e_p95_ms": 211186.33220964108, + "e2e_p99_ms": 235519.09909247872, + "itl_p50_ms": 52.96039255335927, + "itl_p95_ms": 74.77014367468655, + "itl_p99_ms": 93.00403241533787, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o256_a1.json new file mode 100644 index 0000000..ec3349e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 240, + "failed": 0, + "duration_s": 102.57387904892676, + "request_tps": 2.3397769707580456, + "input_tps": 2395.9316180562387, + "output_tps": 598.9829045140597, + "total_tps": 2994.9145225702982, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 4857.287564082071, + "ttft_p95_ms": 5687.151023850311, + "ttft_p99_ms": 5947.1146802348085, + "tpot_p50_ms": 60.30845692353871, + "tpot_p95_ms": 72.36586667291418, + "tpot_p99_ms": 78.81875480012964, + "e2e_p50_ms": 20432.19355738256, + "e2e_p95_ms": 21489.17214354733, + "e2e_p99_ms": 22516.328958119266, + "itl_p50_ms": 51.60140746738762, + "itl_p95_ms": 73.69143945397809, + "itl_p99_ms": 89.00051706237718, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o512_a1.json new file mode 100644 index 0000000..9350fa6 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c48_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 240, + "failed": 0, + "duration_s": 194.6086190820206, + "request_tps": 1.2332444530570794, + "input_tps": 1262.8423199304493, + "output_tps": 631.4211599652247, + "total_tps": 1894.263479895674, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 5219.216491095722, + "ttft_p95_ms": 5779.679291823413, + "ttft_p99_ms": 6389.283453426324, + "tpot_p50_ms": 58.08764942142558, + "tpot_p95_ms": 62.69429164324429, + "tpot_p99_ms": 91.06430042089896, + "e2e_p50_ms": 34896.657894947566, + "e2e_p95_ms": 35910.429730673786, + "e2e_p99_ms": 49593.8836463075, + "itl_p50_ms": 51.571647985838354, + "itl_p95_ms": 73.75171015737578, + "itl_p99_ms": 77.6099856174551, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i1024_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i1024_o4096_a1.json new file mode 100644 index 0000000..af5a11a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i1024_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 750.264118807856, + "request_tps": 0.0533145581632754, + "input_tps": 54.59410755919401, + "output_tps": 218.37643023677603, + "total_tps": 272.97053779597, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 457.8002068446949, + "ttft_p95_ms": 1004.7794182202779, + "ttft_p99_ms": 1006.1946984101087, + "tpot_p50_ms": 36.49401828900812, + "tpot_p95_ms": 36.67125693285014, + "tpot_p99_ms": 36.679555167656694, + "e2e_p50_ms": 150253.00148909446, + "e2e_p95_ms": 150511.7817995837, + "e2e_p99_ms": 150794.05072218738, + "itl_p50_ms": 34.35514506418258, + "itl_p95_ms": 53.22418445721268, + "itl_p99_ms": 59.324696937110275, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i131072_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i131072_o128_a1.json new file mode 100644 index 0000000..2723e11 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i131072_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 818.1570529809687, + "request_tps": 0.04889036873086816, + "input_tps": 6408.158410292352, + "output_tps": 6.257967197551125, + "total_tps": 6414.4163774899025, + "mean_input_tokens": 131072.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 129584.32166394778, + "ttft_p95_ms": 167181.3310468453, + "ttft_p99_ms": 168040.67650724668, + "tpot_p50_ms": 36.84465442857463, + "tpot_p95_ms": 343.3015784548456, + "tpot_p99_ms": 345.92269529082523, + "e2e_p50_ms": 171526.36588714086, + "e2e_p95_ms": 173146.40125581063, + "e2e_p99_ms": 173291.67424835498, + "itl_p50_ms": 33.540375996381044, + "itl_p95_ms": 58.00506006926298, + "itl_p99_ms": 276.8040000228211, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o1024_a1.json new file mode 100644 index 0000000..e8fdb52 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 335.39274771884084, + "request_tps": 0.11926316317826863, + "input_tps": 1954.0076655127532, + "output_tps": 122.12547909454707, + "total_tps": 2076.1331446073004, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 11495.135091943666, + "ttft_p95_ms": 54293.259906279854, + "ttft_p99_ms": 54475.75780400075, + "tpot_p50_ms": 41.142632146611746, + "tpot_p95_ms": 44.4825864581911, + "tpot_p99_ms": 48.84856583131568, + "e2e_p50_ms": 49175.88768794667, + "e2e_p95_ms": 97491.41289745457, + "e2e_p99_ms": 102364.25640797475, + "itl_p50_ms": 34.5357779879123, + "itl_p95_ms": 52.090133994352065, + "itl_p99_ms": 60.2705589751713, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o128_a1.json new file mode 100644 index 0000000..36c0cfa --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 109.44023031182587, + "request_tps": 0.36549630685195744, + "input_tps": 5988.291491462471, + "output_tps": 46.78352727705055, + "total_tps": 6035.075018739521, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 11529.161532991566, + "ttft_p95_ms": 21725.308789138217, + "ttft_p99_ms": 21941.029288116843, + "tpot_p50_ms": 76.67213628768891, + "tpot_p95_ms": 101.16250104450535, + "tpot_p99_ms": 138.8094991766004, + "e2e_p50_ms": 16583.38130300399, + "e2e_p95_ms": 32255.35403857939, + "e2e_p99_ms": 37416.11094208899, + "itl_p50_ms": 34.63044541422278, + "itl_p95_ms": 54.926601552870125, + "itl_p99_ms": 1767.2955490718596, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o2048_a1.json new file mode 100644 index 0000000..587dcdb --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 584.8935852469876, + "request_tps": 0.06838850862607578, + "input_tps": 1120.4773253296255, + "output_tps": 140.0596656662032, + "total_tps": 1260.5369909958285, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 11495.196483563632, + "ttft_p95_ms": 91582.56554905092, + "ttft_p99_ms": 91693.05469256826, + "tpot_p50_ms": 38.41638142817654, + "tpot_p95_ms": 41.38107354754197, + "tpot_p99_ms": 42.587399352946996, + "e2e_p50_ms": 89565.8936999971, + "e2e_p95_ms": 172048.6941325711, + "e2e_p99_ms": 176627.42560487473, + "itl_p50_ms": 34.59935949649662, + "itl_p95_ms": 51.98211170500145, + "itl_p99_ms": 60.23820979986343, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o256_a1.json new file mode 100644 index 0000000..250cade --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 141.670446844073, + "request_tps": 0.28234540718308926, + "input_tps": 4625.947151287734, + "output_tps": 72.28042423887085, + "total_tps": 4698.2275755266055, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 11554.360230453312, + "ttft_p95_ms": 26391.11408374738, + "ttft_p99_ms": 26587.14834881015, + "tpot_p50_ms": 56.31336195065695, + "tpot_p95_ms": 68.25714148759073, + "tpot_p99_ms": 87.6372731319245, + "e2e_p50_ms": 21265.938312513754, + "e2e_p95_ms": 41577.943935687654, + "e2e_p99_ms": 46795.674697651986, + "itl_p50_ms": 34.58584542386234, + "itl_p95_ms": 52.59093521162865, + "itl_p99_ms": 61.120713129639775, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o512_a1.json new file mode 100644 index 0000000..b2bdbc3 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 206.38675206596963, + "request_tps": 0.19381088950522546, + "input_tps": 3175.397613653614, + "output_tps": 99.23117542667543, + "total_tps": 3274.628789080289, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 11589.022769010626, + "ttft_p95_ms": 35687.3680736986, + "ttft_p99_ms": 35902.96996307094, + "tpot_p50_ms": 46.31837186974942, + "tpot_p95_ms": 52.744000344951814, + "tpot_p99_ms": 61.98560190418053, + "e2e_p50_ms": 30626.526765991002, + "e2e_p95_ms": 60265.1604500832, + "e2e_p99_ms": 65432.796396943726, + "itl_p50_ms": 34.61834159679711, + "itl_p95_ms": 51.886969921179116, + "itl_p99_ms": 60.28145674150437, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o1024_a1.json new file mode 100644 index 0000000..83466fc --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 207.04861099808477, + "request_tps": 0.1931913467430603, + "input_tps": 791.311756259575, + "output_tps": 197.82793906489374, + "total_tps": 989.1396953244687, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 1074.1113065741956, + "ttft_p95_ms": 3878.6458508693613, + "ttft_p99_ms": 3916.904898113571, + "tpot_p50_ms": 39.24930339234399, + "tpot_p95_ms": 40.05846651063872, + "tpot_p99_ms": 40.82051611248493, + "e2e_p50_ms": 41186.332597979344, + "e2e_p95_ms": 43346.47120544687, + "e2e_p99_ms": 43701.94057547487, + "itl_p50_ms": 34.467250341549516, + "itl_p95_ms": 54.143763531465076, + "itl_p99_ms": 59.957590696867555, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o128_a1.json new file mode 100644 index 0000000..198e7c9 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 42.330964882159606, + "request_tps": 0.9449347566574843, + "input_tps": 3870.4527632690556, + "output_tps": 120.95164885215799, + "total_tps": 3991.4044121212137, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 2789.9403674528003, + "ttft_p95_ms": 3850.861389178317, + "ttft_p99_ms": 3857.0726825506426, + "tpot_p50_ms": 44.318099374200884, + "tpot_p95_ms": 52.61084187132357, + "tpot_p99_ms": 56.91049005684083, + "e2e_p50_ms": 8397.533455048688, + "e2e_p95_ms": 10301.826596679166, + "e2e_p99_ms": 10350.095471963286, + "itl_p50_ms": 34.42738042213023, + "itl_p95_ms": 54.41078649600968, + "itl_p99_ms": 108.43641549814492, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o2048_a1.json new file mode 100644 index 0000000..42f0ffd --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 394.49152314290404, + "request_tps": 0.10139634859913087, + "input_tps": 415.31944386204003, + "output_tps": 207.65972193102002, + "total_tps": 622.97916579306, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 1034.7788120852783, + "ttft_p95_ms": 3756.3848297810177, + "ttft_p99_ms": 3805.5001059989445, + "tpot_p50_ms": 37.88820180411141, + "tpot_p95_ms": 38.25126371524308, + "tpot_p99_ms": 38.27709748982381, + "e2e_p50_ms": 78674.81265543029, + "e2e_p95_ms": 80982.15972669423, + "e2e_p99_ms": 80984.67170314165, + "itl_p50_ms": 34.52226798981428, + "itl_p95_ms": 53.908159665297724, + "itl_p99_ms": 59.665183436591185, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o256_a1.json new file mode 100644 index 0000000..4d7b8a8 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 65.8774013610091, + "request_tps": 0.6071884921628803, + "input_tps": 2487.0440638991577, + "output_tps": 155.44025399369735, + "total_tps": 2642.4843178928554, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 1971.202465472743, + "ttft_p95_ms": 3824.870833219029, + "ttft_p99_ms": 3841.111220801249, + "tpot_p50_ms": 43.596625505654394, + "tpot_p95_ms": 48.1240840446131, + "tpot_p99_ms": 49.99360711309218, + "e2e_p50_ms": 13162.176689947955, + "e2e_p95_ms": 14974.834044638555, + "e2e_p99_ms": 14991.584688341245, + "itl_p50_ms": 34.47401651646942, + "itl_p95_ms": 54.485111916437745, + "itl_p99_ms": 60.43242933228612, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o4096_a1.json new file mode 100644 index 0000000..fb82ac9 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 762.6022247918881, + "request_tps": 0.052451984402374224, + "input_tps": 214.84332811212482, + "output_tps": 214.84332811212482, + "total_tps": 429.68665622424965, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 1038.2880391553044, + "ttft_p95_ms": 3893.24660198763, + "ttft_p99_ms": 3900.2761626965366, + "tpot_p50_ms": 36.76516193150564, + "tpot_p95_ms": 37.05582632169512, + "tpot_p99_ms": 37.09407625586006, + "e2e_p50_ms": 152243.55608585756, + "e2e_p95_ms": 154084.0833232156, + "e2e_p99_ms": 154084.53153002774, + "itl_p50_ms": 34.463146002963185, + "itl_p95_ms": 51.49366667028517, + "itl_p99_ms": 59.961400236934416, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o512_a1.json new file mode 100644 index 0000000..b22f964 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i4096_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 112.8795186879579, + "request_tps": 0.3543601218798184, + "input_tps": 1451.4590592197362, + "output_tps": 181.43238240246703, + "total_tps": 1632.8914416222033, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 1941.342479432933, + "ttft_p95_ms": 3813.8611613423564, + "ttft_p99_ms": 3815.337903073523, + "tpot_p50_ms": 40.17744154405986, + "tpot_p95_ms": 42.320974605816815, + "tpot_p99_ms": 43.39261491248045, + "e2e_p50_ms": 22431.270663044415, + "e2e_p95_ms": 24335.910513892304, + "e2e_p99_ms": 24337.53361347597, + "itl_p50_ms": 34.43968354258686, + "itl_p95_ms": 54.208886146079735, + "itl_p99_ms": 59.69965270254761, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o1024_a1.json new file mode 100644 index 0000000..213f20b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 626.4431783319451, + "request_tps": 0.06385255899267604, + "input_tps": 4184.641306144017, + "output_tps": 65.38502040850027, + "total_tps": 4250.026326552517, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 51137.51745200716, + "ttft_p95_ms": 96054.72948674572, + "ttft_p99_ms": 112415.82493769703, + "tpot_p50_ms": 70.99086488608475, + "tpot_p95_ms": 93.18777094192289, + "tpot_p99_ms": 94.39113497935946, + "e2e_p50_ms": 112995.05897297058, + "e2e_p95_ms": 168512.40640780888, + "e2e_p99_ms": 178801.59651178867, + "itl_p50_ms": 36.47732094395906, + "itl_p95_ms": 61.783290666062385, + "itl_p99_ms": 68.58176962938157, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o128_a1.json new file mode 100644 index 0000000..f05750d --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 386.69948341907, + "request_tps": 0.10343949685769714, + "input_tps": 6779.01086606604, + "output_tps": 13.240255597785234, + "total_tps": 6792.251121663825, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 52697.72171450313, + "ttft_p95_ms": 77315.9501316608, + "ttft_p99_ms": 77702.88953800453, + "tpot_p50_ms": 161.86251543282611, + "tpot_p95_ms": 301.6361779644242, + "tpot_p99_ms": 390.316338291553, + "e2e_p50_ms": 73942.72513350006, + "e2e_p95_ms": 98705.0266831764, + "e2e_p99_ms": 108562.47204806423, + "itl_p50_ms": 36.458888091146946, + "itl_p95_ms": 54.5758139109239, + "itl_p99_ms": 62.127575105987496, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o256_a1.json new file mode 100644 index 0000000..cc1b59b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 419.1958255460486, + "request_tps": 0.09542079754228375, + "input_tps": 6253.497387731108, + "output_tps": 24.42772417082464, + "total_tps": 6277.925111901933, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 52760.39652095642, + "ttft_p95_ms": 81073.16757803783, + "ttft_p99_ms": 81568.98809597129, + "tpot_p50_ms": 99.82632672934629, + "tpot_p95_ms": 169.0839746892087, + "tpot_p99_ms": 210.72348307306856, + "e2e_p50_ms": 80142.28945260402, + "e2e_p95_ms": 107118.48287502071, + "e2e_p99_ms": 117810.70945826358, + "itl_p50_ms": 36.5109599661082, + "itl_p95_ms": 53.218919737264514, + "itl_p99_ms": 61.9426407897845, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o512_a1.json new file mode 100644 index 0000000..b79cd28 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp4_dp2/metrics/sglang_c8_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 487.3736931588501, + "request_tps": 0.08207254630578259, + "input_tps": 5378.706394695768, + "output_tps": 42.021143708560686, + "total_tps": 5420.727538404329, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 52526.89733938314, + "ttft_p95_ms": 84162.39211051725, + "ttft_p99_ms": 89592.60615824256, + "tpot_p50_ms": 77.38403851756873, + "tpot_p95_ms": 114.1166405163389, + "tpot_p99_ms": 167.8516751349313, + "e2e_p50_ms": 79288.78492198419, + "e2e_p95_ms": 137056.76206527278, + "e2e_p99_ms": 142416.35865559801, + "itl_p50_ms": 36.64115187712014, + "itl_p95_ms": 55.09973680600524, + "itl_p99_ms": 62.820788719691336, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o1024_a1.json new file mode 100644 index 0000000..bfc4d18 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 195.86190408095717, + "request_tps": 0.40845104807585736, + "input_tps": 418.25387322967794, + "output_tps": 418.25387322967794, + "total_tps": 836.5077464593559, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 2806.6968551138416, + "ttft_p95_ms": 2903.0821086023934, + "ttft_p99_ms": 2903.6138558760285, + "tpot_p50_ms": 35.50938803068511, + "tpot_p95_ms": 36.587063905092016, + "tpot_p99_ms": 36.846195016498285, + "e2e_p50_ms": 39140.523318434134, + "e2e_p95_ms": 39359.496829425916, + "e2e_p99_ms": 39362.11585983634, + "itl_p50_ms": 33.30066951457411, + "itl_p95_ms": 52.0355909015052, + "itl_p99_ms": 61.44081162754446, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o128_a1.json new file mode 100644 index 0000000..e427a70 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 40.91812721895985, + "request_tps": 1.9551236930250107, + "input_tps": 2002.046661657611, + "output_tps": 250.25583270720136, + "total_tps": 2252.3024943648124, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 2898.4853614820167, + "ttft_p95_ms": 5916.232369153295, + "ttft_p99_ms": 5917.164536933415, + "tpot_p50_ms": 36.42805760238671, + "tpot_p95_ms": 45.71566509713573, + "tpot_p99_ms": 52.58390468757908, + "e2e_p50_ms": 7513.254255056381, + "e2e_p95_ms": 10505.285932659172, + "e2e_p99_ms": 10506.672543857712, + "itl_p50_ms": 33.98504899814725, + "itl_p95_ms": 55.32407406717539, + "itl_p99_ms": 61.596665140241385, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o2048_a1.json new file mode 100644 index 0000000..f314ef1 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 375.86731938296, + "request_tps": 0.21284106352031737, + "input_tps": 217.949249044805, + "output_tps": 435.89849808961, + "total_tps": 653.8477471344149, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 1012.9853350808844, + "ttft_p95_ms": 1060.0732530583628, + "ttft_p99_ms": 1060.6082381820306, + "tpot_p50_ms": 36.26880324916558, + "tpot_p95_ms": 36.36031736202677, + "tpot_p99_ms": 36.50945045857321, + "e2e_p50_ms": 75173.97456045728, + "e2e_p95_ms": 75301.6543291742, + "e2e_p99_ms": 75354.01330835419, + "itl_p50_ms": 34.3891941010952, + "itl_p95_ms": 56.740176025778055, + "itl_p99_ms": 61.57854507677257, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o256_a1.json new file mode 100644 index 0000000..d7ba41b --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 49.60790970898233, + "request_tps": 1.6126460572378174, + "input_tps": 1651.349562611525, + "output_tps": 412.83739065288125, + "total_tps": 2064.1869532644064, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 935.7815920375288, + "ttft_p95_ms": 987.9027201328427, + "ttft_p99_ms": 988.8454519142397, + "tpot_p50_ms": 35.24840932302907, + "tpot_p95_ms": 35.34214854765428, + "tpot_p99_ms": 35.575809728372484, + "e2e_p50_ms": 9909.394073998556, + "e2e_p95_ms": 9992.725722829346, + "e2e_p99_ms": 9993.638517349027, + "itl_p50_ms": 33.14175654668361, + "itl_p95_ms": 51.57231901539489, + "itl_p99_ms": 59.3762440350838, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o4096_a1.json new file mode 100644 index 0000000..6576ad0 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 751.3376580928452, + "request_tps": 0.1064767606658605, + "input_tps": 109.03220292184115, + "output_tps": 436.1288116873646, + "total_tps": 545.1610146092057, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 1016.063955030404, + "ttft_p95_ms": 1555.9732719440944, + "ttft_p99_ms": 1557.605766323395, + "tpot_p50_ms": 36.36836581573491, + "tpot_p95_ms": 36.73897300841474, + "tpot_p99_ms": 36.74271235461387, + "e2e_p50_ms": 149854.72987149842, + "e2e_p95_ms": 151522.32534443028, + "e2e_p99_ms": 151525.3242665343, + "itl_p50_ms": 34.41172291059047, + "itl_p95_ms": 56.844778708182275, + "itl_p99_ms": 61.78853524383158, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o512_a1.json new file mode 100644 index 0000000..b8dfecd --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 94.76783368107863, + "request_tps": 0.8441682888861142, + "input_tps": 864.4283278193809, + "output_tps": 432.21416390969046, + "total_tps": 1296.6424917290713, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 947.5187105126679, + "ttft_p95_ms": 1040.429971262347, + "ttft_p99_ms": 1041.2525061913766, + "tpot_p50_ms": 35.295242745540456, + "tpot_p95_ms": 35.35704588877044, + "tpot_p99_ms": 35.459834945952224, + "e2e_p50_ms": 18922.931105131283, + "e2e_p95_ms": 19052.50374833122, + "e2e_p99_ms": 19055.464482815005, + "itl_p50_ms": 33.14921457786113, + "itl_p95_ms": 51.78932250710204, + "itl_p99_ms": 59.60216530133038, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o1024_a1.json new file mode 100644 index 0000000..44d582c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 423.16662176605314, + "request_tps": 0.1890508274639578, + "input_tps": 3097.4087571694845, + "output_tps": 193.58804732309278, + "total_tps": 3290.996804492577, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 24554.481841041707, + "ttft_p95_ms": 47767.74098525056, + "ttft_p99_ms": 49284.376522910316, + "tpot_p50_ms": 58.8259498166778, + "tpot_p95_ms": 78.8021439746958, + "tpot_p99_ms": 81.38845075320822, + "e2e_p50_ms": 87506.93495490123, + "e2e_p95_ms": 88091.98981691152, + "e2e_p99_ms": 88095.21754639223, + "itl_p50_ms": 35.09103413671255, + "itl_p95_ms": 58.61164742382244, + "itl_p99_ms": 64.02320196386427, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o128_a1.json new file mode 100644 index 0000000..545af89 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 256.7601454681717, + "request_tps": 0.31157483516037693, + "input_tps": 5104.842099267616, + "output_tps": 39.88157890052825, + "total_tps": 5144.723678168144, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 25187.957516987808, + "ttft_p95_ms": 47894.87515146611, + "ttft_p99_ms": 49477.59126278339, + "tpot_p50_ms": 198.17720644458336, + "tpot_p95_ms": 368.67252661503744, + "tpot_p99_ms": 391.46561684295915, + "e2e_p50_ms": 54196.20531098917, + "e2e_p95_ms": 54292.84199364483, + "e2e_p99_ms": 54294.34734713286, + "itl_p50_ms": 35.08518589660525, + "itl_p95_ms": 58.37633586488664, + "itl_p99_ms": 64.08695708261803, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o2048_a1.json new file mode 100644 index 0000000..9561b67 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 614.5522862081416, + "request_tps": 0.13017606767620898, + "input_tps": 2132.804692807008, + "output_tps": 266.600586600876, + "total_tps": 2399.4052794078843, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 24607.956594438292, + "ttft_p95_ms": 47697.99113486661, + "ttft_p99_ms": 49292.4090649304, + "tpot_p50_ms": 48.10298325354711, + "tpot_p95_ms": 58.110758199371965, + "tpot_p99_ms": 59.46979306338756, + "e2e_p50_ms": 125611.4937895909, + "e2e_p95_ms": 126450.29024835676, + "e2e_p99_ms": 126453.36987355025, + "itl_p50_ms": 35.09154496714473, + "itl_p95_ms": 58.45766137354075, + "itl_p99_ms": 64.08793116686866, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o256_a1.json new file mode 100644 index 0000000..89b1789 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 278.9136179748457, + "request_tps": 0.2868271566690406, + "input_tps": 4699.376134865561, + "output_tps": 73.42775210727439, + "total_tps": 4772.803886972835, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 24543.6039295746, + "ttft_p95_ms": 47646.79518254706, + "ttft_p99_ms": 49249.08797713695, + "tpot_p50_ms": 123.02242240768072, + "tpot_p95_ms": 202.41332247026955, + "tpot_p99_ms": 213.29141370800562, + "e2e_p50_ms": 58673.51271153893, + "e2e_p95_ms": 59028.76297895564, + "e2e_p99_ms": 59031.15820220439, + "itl_p50_ms": 35.10860598180443, + "itl_p95_ms": 59.34609065297991, + "itl_p99_ms": 63.67699921829625, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o512_a1.json new file mode 100644 index 0000000..8058862 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 326.68043872504495, + "request_tps": 0.2448876348771317, + "input_tps": 4012.2390098269257, + "output_tps": 125.38246905709143, + "total_tps": 4137.621478884017, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 24534.058870980516, + "ttft_p95_ms": 47585.56273026624, + "ttft_p99_ms": 49259.400079401676, + "tpot_p50_ms": 80.12083880425615, + "tpot_p95_ms": 119.69057329970916, + "tpot_p99_ms": 125.067007240863, + "e2e_p50_ms": 68319.87011944875, + "e2e_p95_ms": 68537.59080437012, + "e2e_p99_ms": 68539.8418288771, + "itl_p50_ms": 35.064953030087054, + "itl_p95_ms": 58.41207688208669, + "itl_p99_ms": 63.36872420506552, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o1024_a1.json new file mode 100644 index 0000000..e68b46c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 219.81785867293365, + "request_tps": 0.363937673139796, + "input_tps": 1490.6887091806045, + "output_tps": 372.67217729515113, + "total_tps": 1863.3608864757555, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 3086.4613954909146, + "ttft_p95_ms": 11646.606297139078, + "ttft_p99_ms": 11760.603103360627, + "tpot_p50_ms": 37.052429024333115, + "tpot_p95_ms": 44.42799982865398, + "tpot_p99_ms": 45.902305760968, + "e2e_p50_ms": 43634.44979744963, + "e2e_p95_ms": 49578.70842582779, + "e2e_p99_ms": 49580.79158608103, + "itl_p50_ms": 34.58560153376311, + "itl_p95_ms": 56.85673162806779, + "itl_p99_ms": 61.24365669908002, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o128_a1.json new file mode 100644 index 0000000..3f3026f --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 81.84067250299267, + "request_tps": 0.9775090740740753, + "input_tps": 4003.8771674074123, + "output_tps": 125.12116148148164, + "total_tps": 4128.998328888893, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 7960.9383844072, + "ttft_p95_ms": 11711.344682960771, + "ttft_p99_ms": 11753.230824810453, + "tpot_p50_ms": 66.47893063738708, + "tpot_p95_ms": 110.99464661427078, + "tpot_p99_ms": 112.9407165098686, + "e2e_p50_ms": 16375.667929998599, + "e2e_p95_ms": 16515.728015080094, + "e2e_p99_ms": 16517.64271946624, + "itl_p50_ms": 34.67924799770117, + "itl_p95_ms": 57.22848706645891, + "itl_p99_ms": 62.60195288574323, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o2048_a1.json new file mode 100644 index 0000000..f39b007 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 407.2602603479754, + "request_tps": 0.19643458443906506, + "input_tps": 804.5960578624105, + "output_tps": 402.29802893120524, + "total_tps": 1206.8940867936158, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 3240.20006461069, + "ttft_p95_ms": 11741.664571419824, + "ttft_p99_ms": 11840.223265609238, + "tpot_p50_ms": 36.97641234489819, + "tpot_p95_ms": 40.52301196099482, + "tpot_p99_ms": 41.266952687252264, + "e2e_p50_ms": 80944.1796640167, + "e2e_p95_ms": 87096.1231123074, + "e2e_p99_ms": 87098.59982100548, + "itl_p50_ms": 34.466654527932405, + "itl_p95_ms": 56.807562068570405, + "itl_p99_ms": 61.39076543040574, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o256_a1.json new file mode 100644 index 0000000..22b6830 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 78.74828192405403, + "request_tps": 1.0158951794929716, + "input_tps": 4161.106655203212, + "output_tps": 260.0691659502007, + "total_tps": 4421.1758211534125, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 3091.937036952004, + "ttft_p95_ms": 11642.126731504686, + "ttft_p99_ms": 11714.910287000239, + "tpot_p50_ms": 37.18031867425524, + "tpot_p95_ms": 68.26280832984575, + "tpot_p99_ms": 73.90357631190186, + "e2e_p50_ms": 15660.35121458117, + "e2e_p95_ms": 21110.056800360326, + "e2e_p99_ms": 21111.51674753055, + "itl_p50_ms": 34.62257899809629, + "itl_p95_ms": 57.151046535000205, + "itl_p99_ms": 62.4327381211333, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o4096_a1.json new file mode 100644 index 0000000..8a493ce --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 781.9880918110721, + "request_tps": 0.10230334814270799, + "input_tps": 419.0345139925319, + "output_tps": 419.0345139925319, + "total_tps": 838.0690279850638, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 3097.252030042, + "ttft_p95_ms": 11633.983958070166, + "ttft_p99_ms": 11746.600034988951, + "tpot_p50_ms": 36.86781218838315, + "tpot_p95_ms": 38.49309932329363, + "tpot_p99_ms": 38.90570051973158, + "e2e_p50_ms": 155683.48012503702, + "e2e_p95_ms": 162603.52390279295, + "e2e_p99_ms": 162604.7951752739, + "itl_p50_ms": 34.45248049683869, + "itl_p95_ms": 56.79855167400092, + "itl_p99_ms": 61.650155324023224, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o512_a1.json new file mode 100644 index 0000000..60e3c51 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i4096_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 126.69021714199334, + "request_tps": 0.63146154300404, + "input_tps": 2586.466480144548, + "output_tps": 323.3083100180685, + "total_tps": 2909.7747901626167, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 3073.0733450036496, + "ttft_p95_ms": 11019.555248494726, + "ttft_p99_ms": 11065.65095537575, + "tpot_p50_ms": 38.32130294193836, + "tpot_p95_ms": 53.03848213837399, + "tpot_p99_ms": 58.19536544753461, + "e2e_p50_ms": 25832.93817844242, + "e2e_p95_ms": 30626.584287162405, + "e2e_p99_ms": 30627.329029503744, + "itl_p50_ms": 34.66650005429983, + "itl_p95_ms": 56.83219095226377, + "itl_p99_ms": 62.253236598335214, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o1024_a1.json new file mode 100644 index 0000000..1b02781 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1330.2602253910154, + "request_tps": 0.06013860932847547, + "input_tps": 3941.2439009509685, + "output_tps": 61.58193595235888, + "total_tps": 4002.8258369033274, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 118513.18043493666, + "ttft_p95_ms": 230933.78256635042, + "ttft_p99_ms": 232188.58536266722, + "tpot_p50_ms": 137.7084055620283, + "tpot_p95_ms": 237.46356306838157, + "tpot_p99_ms": 250.69030466432596, + "e2e_p50_ms": 271416.1142479861, + "e2e_p95_ms": 272710.0276091136, + "e2e_p99_ms": 272711.17356425617, + "itl_p50_ms": 36.84138902463019, + "itl_p95_ms": 61.22844358906149, + "itl_p99_ms": 67.15500121004875, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o128_a1.json new file mode 100644 index 0000000..67f2585 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1164.4257685828488, + "request_tps": 0.06870339197093095, + "input_tps": 4502.545496206931, + "output_tps": 8.794034172279162, + "total_tps": 4511.33953037921, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 120625.26360049378, + "ttft_p95_ms": 230919.13024774985, + "ttft_p99_ms": 233213.97572936953, + "tpot_p50_ms": 847.999912252151, + "tpot_p95_ms": 1666.0538613161907, + "tpot_p99_ms": 1748.680016982211, + "e2e_p50_ms": 236925.28252699412, + "e2e_p95_ms": 241650.54931361228, + "e2e_p99_ms": 241654.8121647397, + "itl_p50_ms": 36.83545487001538, + "itl_p95_ms": 60.686023067682996, + "itl_p99_ms": 88.91421182081103, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o256_a1.json new file mode 100644 index 0000000..2392e90 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1179.4044030888472, + "request_tps": 0.06783084732470125, + "input_tps": 4445.362410271621, + "output_tps": 17.36469691512352, + "total_tps": 4462.727107186744, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 118520.80365340225, + "ttft_p95_ms": 219668.39720717393, + "ttft_p99_ms": 232060.81151018385, + "tpot_p50_ms": 456.4013590040964, + "tpot_p95_ms": 834.1273255841166, + "tpot_p99_ms": 887.1042354636113, + "e2e_p50_ms": 241645.95780789386, + "e2e_p95_ms": 242439.91744236555, + "e2e_p99_ms": 242443.70790595654, + "itl_p50_ms": 36.84396401513368, + "itl_p95_ms": 61.602629092521966, + "itl_p99_ms": 67.18409784138203, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o512_a1.json new file mode 100644 index 0000000..5573e96 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c16_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 80, + "failed": 0, + "duration_s": 1229.7762659930158, + "request_tps": 0.0650524832949202, + "input_tps": 4263.27954521589, + "output_tps": 33.30687144699914, + "total_tps": 4296.586416662889, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 117602.89832751732, + "ttft_p95_ms": 220174.25819333873, + "ttft_p99_ms": 232176.57839990687, + "tpot_p50_ms": 250.0348996674941, + "tpot_p95_ms": 436.4520301115598, + "tpot_p99_ms": 462.66546465371283, + "e2e_p50_ms": 251397.16331893578, + "e2e_p95_ms": 252849.9990378623, + "e2e_p99_ms": 252851.82005017297, + "itl_p50_ms": 36.86228208243847, + "itl_p95_ms": 60.973369050771, + "itl_p99_ms": 67.14509757934138, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o1024_a1.json new file mode 100644 index 0000000..3dacd46 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 150.84414941398427, + "request_tps": 0.03314679435314225, + "input_tps": 543.0770786818827, + "output_tps": 33.942317417617666, + "total_tps": 577.0193960995003, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 3190.0415259879082, + "ttft_p95_ms": 3315.6291431281716, + "ttft_p99_ms": 3326.89755034633, + "tpot_p50_ms": 26.902348835946206, + "tpot_p95_ms": 26.929367462743006, + "tpot_p99_ms": 26.932126218243663, + "e2e_p50_ms": 30694.867955055088, + "e2e_p95_ms": 30857.89103670977, + "e2e_p99_ms": 30877.166467448696, + "itl_p50_ms": 24.886570870876312, + "itl_p95_ms": 45.76608098577708, + "itl_p99_ms": 50.4940598970279, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o128_a1.json new file mode 100644 index 0000000..1e36f82 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 30.202961774077266, + "request_tps": 0.16554667841520834, + "input_tps": 2712.3167791547735, + "output_tps": 21.189974837146668, + "total_tps": 2733.5067539919205, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 3189.695225097239, + "ttft_p95_ms": 3287.1102603152394, + "ttft_p99_ms": 3298.7277361750603, + "tpot_p50_ms": 27.00854029953773, + "tpot_p95_ms": 27.13199242340826, + "tpot_p99_ms": 27.15187601517798, + "e2e_p50_ms": 6534.987838938832, + "e2e_p95_ms": 6732.873298088089, + "e2e_p99_ms": 6747.015990102664, + "itl_p50_ms": 24.80782102793455, + "itl_p95_ms": 45.76855925843107, + "itl_p99_ms": 49.268019353039556, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o2048_a1.json new file mode 100644 index 0000000..28f72c8 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 288.6096170139499, + "request_tps": 0.017324440022933558, + "input_tps": 283.8436253357434, + "output_tps": 35.480453166967926, + "total_tps": 319.32407850271136, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 3130.5205351673067, + "ttft_p95_ms": 3326.790696242824, + "ttft_p99_ms": 3348.429289693013, + "tpot_p50_ms": 26.910882942859878, + "tpot_p95_ms": 26.939971389955993, + "tpot_p99_ms": 26.940729667464, + "e2e_p50_ms": 58267.43282494135, + "e2e_p95_ms": 58423.72469645925, + "e2e_p99_ms": 58437.077996963635, + "itl_p50_ms": 24.81332910247147, + "itl_p95_ms": 48.45747402869165, + "itl_p99_ms": 51.16548259742558, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o256_a1.json new file mode 100644 index 0000000..da643dc --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 47.27609982597642, + "request_tps": 0.10576168546908539, + "input_tps": 1732.799454725495, + "output_tps": 27.07499148008586, + "total_tps": 1759.8744462055809, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 3095.8918230608106, + "ttft_p95_ms": 3272.758443793282, + "ttft_p99_ms": 3304.284312753007, + "tpot_p50_ms": 27.063766945445657, + "tpot_p95_ms": 27.16058336417465, + "tpot_p99_ms": 27.17001092872199, + "e2e_p50_ms": 9997.152394149452, + "e2e_p95_ms": 10198.707201657817, + "e2e_p99_ms": 10232.637099577114, + "itl_p50_ms": 24.803407955914736, + "itl_p95_ms": 48.14038993790746, + "itl_p99_ms": 50.25792038068175, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o512_a1.json new file mode 100644 index 0000000..a070944 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 81.74180358601734, + "request_tps": 0.06116821235463043, + "input_tps": 1002.179991218265, + "output_tps": 31.31812472557078, + "total_tps": 1033.4981159438357, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 3152.1379488985986, + "ttft_p95_ms": 3280.934850871563, + "ttft_p99_ms": 3303.358238041401, + "tpot_p50_ms": 26.90626537553062, + "tpot_p95_ms": 26.921899768354372, + "tpot_p99_ms": 26.923089013626313, + "e2e_p50_ms": 16891.912296880037, + "e2e_p95_ms": 17034.689639275894, + "e2e_p99_ms": 17060.389525359496, + "itl_p50_ms": 24.79128399863839, + "itl_p95_ms": 48.06917109526691, + "itl_p99_ms": 50.64208318479359, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o1024_a1.json new file mode 100644 index 0000000..abc5b72 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 137.4567740750499, + "request_tps": 0.0363750716081119, + "input_tps": 148.99229330682635, + "output_tps": 37.24807332670659, + "total_tps": 186.24036663353294, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 250.88916602544487, + "ttft_p95_ms": 253.6237696185708, + "ttft_p99_ms": 253.63623090088367, + "tpot_p50_ms": 26.61167698223162, + "tpot_p95_ms": 26.682842084191165, + "tpot_p99_ms": 26.683082640242063, + "e2e_p50_ms": 27477.38489904441, + "e2e_p95_ms": 27547.797128511593, + "e2e_p99_ms": 27548.662202609703, + "itl_p50_ms": 24.606115883216262, + "itl_p95_ms": 47.627381072379656, + "itl_p99_ms": 50.5407242104411, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o128_a1.json new file mode 100644 index 0000000..a108222 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 18.347577747888863, + "request_tps": 0.2725155368574643, + "input_tps": 1116.2236389681739, + "output_tps": 34.88198871775543, + "total_tps": 1151.1056276859294, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 250.82844705320895, + "ttft_p95_ms": 256.7141311708838, + "ttft_p99_ms": 257.63806560076773, + "tpot_p50_ms": 26.909995739766227, + "tpot_p95_ms": 27.076458471907873, + "tpot_p99_ms": 27.091512166432977, + "e2e_p50_ms": 3668.39790600352, + "e2e_p95_ms": 3695.424357103184, + "e2e_p99_ms": 3698.260110737756, + "itl_p50_ms": 24.74083099514246, + "itl_p95_ms": 47.6031189551577, + "itl_p99_ms": 49.66991530265658, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o2048_a1.json new file mode 100644 index 0000000..5d2888c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 273.5867548820097, + "request_tps": 0.018275738539156837, + "input_tps": 74.8574250563864, + "output_tps": 37.4287125281932, + "total_tps": 112.28613758457962, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 261.7046369705349, + "ttft_p95_ms": 349.8987853061408, + "ttft_p99_ms": 362.0478419493884, + "tpot_p50_ms": 26.592822536383906, + "tpot_p95_ms": 26.61543222684295, + "tpot_p99_ms": 26.61759176356131, + "e2e_p50_ms": 54736.32405605167, + "e2e_p95_ms": 54748.984178341925, + "e2e_p99_ms": 54749.01293158531, + "itl_p50_ms": 24.60544602945447, + "itl_p95_ms": 47.4006032804027, + "itl_p99_ms": 49.956067339517176, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o256_a1.json new file mode 100644 index 0000000..f3c2188 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 35.27845326298848, + "request_tps": 0.1417295696817192, + "input_tps": 580.5243174163219, + "output_tps": 36.28276983852012, + "total_tps": 616.807087254842, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 268.0197979789227, + "ttft_p95_ms": 291.9198756571859, + "ttft_p99_ms": 292.9051029961556, + "tpot_p50_ms": 26.539895318302456, + "tpot_p95_ms": 26.827862831928275, + "tpot_p99_ms": 26.862892086781997, + "e2e_p50_ms": 7060.824715998024, + "e2e_p95_ms": 7112.919608317316, + "e2e_p99_ms": 7118.816237747669, + "itl_p50_ms": 24.650061037391424, + "itl_p95_ms": 44.851708575151854, + "itl_p99_ms": 49.219665322452784, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o4096_a1.json new file mode 100644 index 0000000..d651564 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 548.1068854129408, + "request_tps": 0.009122308318081111, + "input_tps": 37.36497487086023, + "output_tps": 37.36497487086023, + "total_tps": 74.72994974172046, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 270.29989496804774, + "ttft_p95_ms": 280.77137977816164, + "ttft_p99_ms": 282.4734351504594, + "tpot_p50_ms": 26.703159420515647, + "tpot_p95_ms": 26.721241031938273, + "tpot_p99_ms": 26.721443862750927, + "e2e_p50_ms": 109619.73772197962, + "e2e_p95_ms": 109701.13868489861, + "e2e_p99_ms": 109702.0101480931, + "itl_p50_ms": 24.71006289124489, + "itl_p95_ms": 45.58566838968543, + "itl_p99_ms": 49.87936850637197, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o512_a1.json new file mode 100644 index 0000000..e63cb01 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i4096_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 69.34047695202753, + "request_tps": 0.07210795511918958, + "input_tps": 295.3541841682005, + "output_tps": 36.91927302102506, + "total_tps": 332.27345718922555, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 248.92402300611138, + "ttft_p95_ms": 266.0276424139738, + "ttft_p99_ms": 268.3548469096422, + "tpot_p50_ms": 26.661601453862808, + "tpot_p95_ms": 26.700651652677212, + "tpot_p99_ms": 26.707691252833037, + "e2e_p50_ms": 13878.469962859526, + "e2e_p95_ms": 13891.861443966627, + "e2e_p99_ms": 13893.72499614954, + "itl_p50_ms": 24.624896002933383, + "itl_p95_ms": 45.65126097295369, + "itl_p99_ms": 49.571911790408194, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o1024_a1.json new file mode 100644 index 0000000..82f7676 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 201.46440248796716, + "request_tps": 0.024818280243323058, + "input_tps": 1626.49081402642, + "output_tps": 25.41391896916281, + "total_tps": 1651.9047329955827, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 14628.885317826644, + "ttft_p95_ms": 14868.979735858738, + "ttft_p99_ms": 14913.150340802968, + "tpot_p50_ms": 27.836642022547792, + "tpot_p95_ms": 27.844008100505842, + "tpot_p99_ms": 27.84499085022142, + "e2e_p50_ms": 43054.00717793964, + "e2e_p95_ms": 43330.51605997607, + "e2e_p99_ms": 43373.7068128027, + "itl_p50_ms": 25.475302943959832, + "itl_p95_ms": 50.770317669957876, + "itl_p99_ms": 54.30494689382613, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o128_a1.json new file mode 100644 index 0000000..ad69822 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 76.5485741731245, + "request_tps": 0.0653180030328436, + "input_tps": 4280.680646760438, + "output_tps": 8.360704388203981, + "total_tps": 4289.041351148642, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 14596.343518933281, + "ttft_p95_ms": 14810.734046529979, + "ttft_p99_ms": 14839.107076432556, + "tpot_p50_ms": 27.468161299956712, + "tpot_p95_ms": 27.539072871046862, + "tpot_p99_ms": 27.552164999592144, + "e2e_p50_ms": 18084.800004027784, + "e2e_p95_ms": 18259.232654888183, + "e2e_p99_ms": 18286.724362093955, + "itl_p50_ms": 25.447878055274487, + "itl_p95_ms": 46.60883932374417, + "itl_p99_ms": 51.75249738618731, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o256_a1.json new file mode 100644 index 0000000..1757a72 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 94.3989737580996, + "request_tps": 0.052966677506607864, + "input_tps": 3471.224177073053, + "output_tps": 13.559469441691613, + "total_tps": 3484.7836465147443, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 14599.749388871714, + "ttft_p95_ms": 14722.214041789994, + "ttft_p99_ms": 14730.883328346536, + "tpot_p50_ms": 27.881732368476545, + "tpot_p95_ms": 27.903486282619486, + "tpot_p99_ms": 27.905507876263826, + "e2e_p50_ms": 21712.56085904315, + "e2e_p95_ms": 21806.284560635686, + "e2e_p99_ms": 21809.799600169063, + "itl_p50_ms": 25.44583403505385, + "itl_p95_ms": 49.89882283844054, + "itl_p99_ms": 52.76037988718599, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o512_a1.json new file mode 100644 index 0000000..f09bea2 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c1_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 5, + "failed": 0, + "duration_s": 130.4824561579153, + "request_tps": 0.03831932772593422, + "input_tps": 2511.295461846825, + "output_tps": 19.619495795678322, + "total_tps": 2530.9149576425034, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 14583.937218878418, + "ttft_p95_ms": 14795.338988443837, + "ttft_p99_ms": 14826.081829732284, + "tpot_p50_ms": 27.982182919691134, + "tpot_p95_ms": 28.00536987685221, + "tpot_p99_ms": 28.00938046322279, + "e2e_p50_ms": 28872.23194888793, + "e2e_p95_ms": 29090.035974793136, + "e2e_p99_ms": 29125.418794117868, + "itl_p50_ms": 25.67037590779364, + "itl_p95_ms": 50.41482762899249, + "itl_p99_ms": 53.483981639146805, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o1024_a1.json new file mode 100644 index 0000000..671234e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 235.65979528496973, + "request_tps": 0.6789448314954245, + "input_tps": 695.2395074513147, + "output_tps": 695.2395074513147, + "total_tps": 1390.4790149026294, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 2981.6380555275828, + "ttft_p95_ms": 5691.65476138005, + "ttft_p99_ms": 5695.868596371729, + "tpot_p50_ms": 42.80786131479516, + "tpot_p95_ms": 45.161864508836935, + "tpot_p99_ms": 45.66859811666182, + "e2e_p50_ms": 47101.06354195159, + "e2e_p95_ms": 49224.63555838913, + "e2e_p99_ms": 49227.704358948395, + "itl_p50_ms": 40.25210742838681, + "itl_p95_ms": 59.54401966882868, + "itl_p99_ms": 67.25824875058606, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o2048_a1.json new file mode 100644 index 0000000..2f2208a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 455.1479200378526, + "request_tps": 0.35153406827981004, + "input_tps": 359.9708859185255, + "output_tps": 719.941771837051, + "total_tps": 1079.9126577555764, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 2997.489857953042, + "ttft_p95_ms": 4408.118210406974, + "ttft_p99_ms": 4409.418380276766, + "tpot_p50_ms": 43.03901399925722, + "tpot_p95_ms": 43.72752447672071, + "tpot_p99_ms": 43.95574388355626, + "e2e_p50_ms": 91594.98853096738, + "e2e_p95_ms": 92511.16100725485, + "e2e_p99_ms": 92513.411904776, + "itl_p50_ms": 40.91528104618192, + "itl_p95_ms": 61.7140918620862, + "itl_p99_ms": 67.57471586111933, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o256_a1.json new file mode 100644 index 0000000..211cb67 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 72.94000257505104, + "request_tps": 2.1935836900385253, + "input_tps": 2246.22969859945, + "output_tps": 561.5574246498625, + "total_tps": 2807.7871232493126, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 2986.9613859336823, + "ttft_p95_ms": 5638.966712448744, + "ttft_p99_ms": 5717.785984254442, + "tpot_p50_ms": 43.25931125649196, + "tpot_p95_ms": 52.76475823822194, + "tpot_p99_ms": 60.14857301456559, + "e2e_p50_ms": 14733.08369051665, + "e2e_p95_ms": 16383.586864836978, + "e2e_p99_ms": 16391.066697821952, + "itl_p50_ms": 40.04870797507465, + "itl_p95_ms": 61.18088683579117, + "itl_p99_ms": 67.89261806756254, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o4096_a1.json new file mode 100644 index 0000000..1a19825 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 898.5537388259545, + "request_tps": 0.17806391881364284, + "input_tps": 182.33745286517026, + "output_tps": 729.349811460681, + "total_tps": 911.6872643258514, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 3016.2806854350492, + "ttft_p95_ms": 3879.865962965414, + "ttft_p99_ms": 3881.218241781462, + "tpot_p50_ms": 43.30877847119516, + "tpot_p95_ms": 43.5328147706794, + "tpot_p99_ms": 43.585850752513146, + "e2e_p50_ms": 179637.5885854941, + "e2e_p95_ms": 181239.3637237721, + "e2e_p99_ms": 181245.79159579705, + "itl_p50_ms": 40.978090022690594, + "itl_p95_ms": 61.92717199446633, + "itl_p99_ms": 67.58996407035738, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o512_a1.json new file mode 100644 index 0000000..3f1045a --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c32_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 160, + "failed": 0, + "duration_s": 116.47414470394142, + "request_tps": 1.3736954274847377, + "input_tps": 1406.6641177443714, + "output_tps": 703.3320588721857, + "total_tps": 2109.996176616557, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 1728.7794339936227, + "ttft_p95_ms": 1780.919153173454, + "ttft_p99_ms": 1782.5818243529648, + "tpot_p50_ms": 42.263241610289306, + "tpot_p95_ms": 42.444828835041704, + "tpot_p99_ms": 42.508655317581514, + "e2e_p50_ms": 23282.43359492626, + "e2e_p95_ms": 23419.968224375043, + "e2e_p99_ms": 23422.798421205953, + "itl_p50_ms": 40.06265802308917, + "itl_p95_ms": 58.03677138173953, + "itl_p99_ms": 67.34221158549191, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c48_i1024_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c48_i1024_o512_a1.json new file mode 100644 index 0000000..6ff5985 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c48_i1024_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 240, + "failed": 0, + "duration_s": 538.8399277571589, + "request_tps": 0.44540129199215867, + "input_tps": 456.0909229999705, + "output_tps": 228.04546149998524, + "total_tps": 684.1363844999557, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 2837.4407634837553, + "ttft_p95_ms": 8041.054585680831, + "ttft_p99_ms": 8290.32544139307, + "tpot_p50_ms": 202.26081837298543, + "tpot_p95_ms": 209.53028491557413, + "tpot_p99_ms": 212.22897237452446, + "e2e_p50_ms": 106162.95209643431, + "e2e_p95_ms": 111388.48089557141, + "e2e_p99_ms": 111393.2417512266, + "itl_p50_ms": 197.14024697896093, + "itl_p95_ms": 226.85785598587245, + "itl_p99_ms": 233.4809394343756, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i1024_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i1024_o128_a1.json new file mode 100644 index 0000000..8616aa5 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i1024_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 23.175778079079464, + "request_tps": 1.7259398956752865, + "input_tps": 1767.3624531714934, + "output_tps": 220.92030664643667, + "total_tps": 1988.28275981793, + "mean_input_tokens": 1024.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 586.4921135362238, + "ttft_p95_ms": 622.7808163734153, + "ttft_p99_ms": 623.125078517478, + "tpot_p50_ms": 31.85279993340373, + "tpot_p95_ms": 32.22969000018752, + "tpot_p99_ms": 32.44051498659762, + "e2e_p50_ms": 4641.481174970977, + "e2e_p95_ms": 4670.240912691224, + "e2e_p99_ms": 4670.475004990585, + "itl_p50_ms": 29.695762554183602, + "itl_p95_ms": 49.66246752301232, + "itl_p99_ms": 57.65656981850043, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o1024_a1.json new file mode 100644 index 0000000..c45713c --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 288.24989658407867, + "request_tps": 0.13876847996832684, + "input_tps": 2273.582775801067, + "output_tps": 142.09892348756668, + "total_tps": 2415.6816992886334, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 14033.50293962285, + "ttft_p95_ms": 24491.35324291419, + "ttft_p99_ms": 24759.231015304103, + "tpot_p50_ms": 41.92089678206183, + "tpot_p95_ms": 52.53722619232089, + "tpot_p99_ms": 52.69310522824981, + "e2e_p50_ms": 58040.07501259912, + "e2e_p95_ms": 58448.35202746326, + "e2e_p99_ms": 58450.28124490753, + "itl_p50_ms": 30.73286300059408, + "itl_p95_ms": 50.72684104088693, + "itl_p99_ms": 57.92215333785862, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o128_a1.json new file mode 100644 index 0000000..8500c91 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 141.1061867091339, + "request_tps": 0.283474459432832, + "input_tps": 4644.44554334752, + "output_tps": 36.2847308074025, + "total_tps": 4680.730274154922, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 14016.848506988026, + "ttft_p95_ms": 24515.615026152227, + "ttft_p99_ms": 24870.744992317632, + "tpot_p50_ms": 105.65934030661785, + "tpot_p95_ms": 191.19213953821398, + "tpot_p99_ms": 193.04246251707838, + "e2e_p50_ms": 28676.901251543313, + "e2e_p95_ms": 29117.543387238402, + "e2e_p99_ms": 29119.045873775613, + "itl_p50_ms": 30.759848072193563, + "itl_p95_ms": 50.73059933492914, + "itl_p99_ms": 57.838678041007384, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o2048_a1.json new file mode 100644 index 0000000..e34d1e6 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 457.1198655979242, + "request_tps": 0.08750440094673856, + "input_tps": 1433.6721051113645, + "output_tps": 179.20901313892057, + "total_tps": 1612.881118250285, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 14047.149811987765, + "ttft_p95_ms": 24552.844235277735, + "ttft_p99_ms": 24749.485357583035, + "tpot_p50_ms": 37.538202461908504, + "tpot_p95_ms": 42.66441069203499, + "tpot_p99_ms": 42.9089602763204, + "e2e_p50_ms": 91750.90344448108, + "e2e_p95_ms": 92103.39087992907, + "e2e_p99_ms": 92105.43445243035, + "itl_p50_ms": 30.728504993021488, + "itl_p95_ms": 51.14136257907376, + "itl_p99_ms": 58.35309494053946, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o256_a1.json new file mode 100644 index 0000000..90b29f5 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 162.15329257398844, + "request_tps": 0.24668015903376442, + "input_tps": 4041.607725609196, + "output_tps": 63.15012071264369, + "total_tps": 4104.75784632184, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 14056.083535542712, + "ttft_p95_ms": 24572.28690140182, + "ttft_p99_ms": 24781.60693206126, + "tpot_p50_ms": 69.22504512739239, + "tpot_p95_ms": 111.92492204695465, + "tpot_p99_ms": 112.1908097408584, + "e2e_p50_ms": 32945.15154813416, + "e2e_p95_ms": 33157.10190224927, + "e2e_p99_ms": 33157.680223954376, + "itl_p50_ms": 30.78671346884221, + "itl_p95_ms": 52.047876664437354, + "itl_p99_ms": 57.51584505662322, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o512_a1.json new file mode 100644 index 0000000..14e5fa5 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i16384_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 204.11131394607946, + "request_tps": 0.19597149823143506, + "input_tps": 3210.797027023832, + "output_tps": 100.33740709449475, + "total_tps": 3311.134434118327, + "mean_input_tokens": 16384.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 14109.92513189558, + "ttft_p95_ms": 24601.094605890103, + "ttft_p99_ms": 24752.130593049806, + "tpot_p50_ms": 50.97825450885589, + "tpot_p95_ms": 72.14095964029964, + "tpot_p99_ms": 72.50778853802152, + "e2e_p50_ms": 41347.00059145689, + "e2e_p95_ms": 41603.95504550543, + "e2e_p99_ms": 41605.48491582973, + "itl_p50_ms": 30.709165963344276, + "itl_p95_ms": 51.023998111486435, + "itl_p99_ms": 57.31804177630693, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o1024_a1.json new file mode 100644 index 0000000..b31d1ff --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 196.3215377200395, + "request_tps": 0.2037473853584074, + "input_tps": 834.5492904280367, + "output_tps": 208.63732260700917, + "total_tps": 1043.1866130350459, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 5105.625024996698, + "ttft_p95_ms": 5958.576378412545, + "ttft_p99_ms": 6064.859193183947, + "tpot_p50_ms": 33.55810541205767, + "tpot_p95_ms": 36.34212366745689, + "tpot_p99_ms": 36.55665332609869, + "e2e_p50_ms": 39269.71339690499, + "e2e_p95_ms": 39593.57306190068, + "e2e_p99_ms": 39594.54825331457, + "itl_p50_ms": 30.5747389793396, + "itl_p95_ms": 50.719153042882674, + "itl_p99_ms": 56.99894693214446, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o128_a1.json new file mode 100644 index 0000000..24ab7ce --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 49.58700227085501, + "request_tps": 0.8066629997415711, + "input_tps": 3304.091646941475, + "output_tps": 103.2528639669211, + "total_tps": 3407.3445109083964, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 5025.09073051624, + "ttft_p95_ms": 5915.067831950728, + "ttft_p99_ms": 5934.279049527831, + "tpot_p50_ms": 39.28249386619542, + "tpot_p95_ms": 62.01645719083613, + "tpot_p99_ms": 62.2953052225332, + "e2e_p50_ms": 10006.435959017836, + "e2e_p95_ms": 10091.394168324769, + "e2e_p99_ms": 10092.632040372118, + "itl_p50_ms": 30.577882542274892, + "itl_p95_ms": 49.62789301062003, + "itl_p99_ms": 57.752098962664604, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o2048_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o2048_a1.json new file mode 100644 index 0000000..8c1ad96 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o2048_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 363.24991058488376, + "request_tps": 0.11011702641741698, + "input_tps": 451.03934020573996, + "output_tps": 225.51967010286998, + "total_tps": 676.5590103086099, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 2048.0, + "ttft_p50_ms": 5055.334709468298, + "ttft_p95_ms": 5906.876079260837, + "ttft_p99_ms": 6023.611045922153, + "tpot_p50_ms": 33.129859284787166, + "tpot_p95_ms": 34.53019272802082, + "tpot_p99_ms": 34.57053666370664, + "e2e_p50_ms": 72728.55528548826, + "e2e_p95_ms": 73087.71414748626, + "e2e_p99_ms": 73088.92423222307, + "itl_p50_ms": 30.50815698225051, + "itl_p95_ms": 50.1406642026268, + "itl_p99_ms": 56.95488589582965, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o256_a1.json new file mode 100644 index 0000000..0b74966 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 70.42232585605234, + "request_tps": 0.5680016885804441, + "input_tps": 2326.534916425499, + "output_tps": 145.4084322765937, + "total_tps": 2471.943348702093, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 5035.245958482847, + "ttft_p95_ms": 5984.106912557036, + "ttft_p99_ms": 6042.94591952581, + "tpot_p50_ms": 35.78011738455982, + "tpot_p95_ms": 46.93394760007733, + "tpot_p99_ms": 47.50900067536927, + "e2e_p50_ms": 14144.402876030654, + "e2e_p95_ms": 14412.95803807443, + "e2e_p99_ms": 14413.140350417234, + "itl_p50_ms": 30.597370583564043, + "itl_p95_ms": 50.928640994243324, + "itl_p99_ms": 56.54409846989438, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o4096_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o4096_a1.json new file mode 100644 index 0000000..9935df5 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o4096_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 697.7969978589099, + "request_tps": 0.057323261812152054, + "input_tps": 234.79608038257481, + "output_tps": 234.79608038257481, + "total_tps": 469.59216076514963, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 4096.0, + "ttft_p50_ms": 5059.23117196653, + "ttft_p95_ms": 6020.107874658424, + "ttft_p99_ms": 6025.99322974449, + "tpot_p50_ms": 32.921226551871115, + "tpot_p95_ms": 33.65806957458343, + "tpot_p99_ms": 33.68422295676342, + "e2e_p50_ms": 139956.67370257434, + "e2e_p95_ms": 140090.03641876625, + "e2e_p99_ms": 140090.8339079097, + "itl_p50_ms": 30.493864906020463, + "itl_p95_ms": 50.366084929555655, + "itl_p99_ms": 57.08304369123653, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o512_a1.json new file mode 100644 index 0000000..a0afdcb --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i4096_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 112.55219594296068, + "request_tps": 0.3553906671023215, + "input_tps": 1455.6801724511088, + "output_tps": 181.9600215563886, + "total_tps": 1637.6401940074975, + "mean_input_tokens": 4096.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 5065.312996157445, + "ttft_p95_ms": 6016.085800586734, + "ttft_p99_ms": 6074.749361765571, + "tpot_p50_ms": 34.33472787484767, + "tpot_p95_ms": 40.08658976554269, + "tpot_p99_ms": 40.18997862394681, + "e2e_p50_ms": 22591.4841449121, + "e2e_p95_ms": 22795.752924506087, + "e2e_p99_ms": 22796.309465037193, + "itl_p50_ms": 30.596577562391758, + "itl_p95_ms": 50.246595905628055, + "itl_p99_ms": 56.772089656442404, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o1024_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o1024_a1.json new file mode 100644 index 0000000..e7bb451 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o1024_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 739.3616181469988, + "request_tps": 0.054100725569510504, + "input_tps": 3545.5451509234404, + "output_tps": 55.399142983178756, + "total_tps": 3600.9442939066194, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 1024.0, + "ttft_p50_ms": 60294.08791498281, + "ttft_p95_ms": 116841.35833900655, + "ttft_p99_ms": 116967.29428508785, + "tpot_p50_ms": 82.37472311334132, + "tpot_p95_ms": 131.37778040708542, + "tpot_p99_ms": 132.5520263926535, + "e2e_p50_ms": 150254.07721009105, + "e2e_p95_ms": 151760.6172351283, + "e2e_p99_ms": 151763.06750814896, + "itl_p50_ms": 31.72238008119166, + "itl_p95_ms": 53.61441783607006, + "itl_p99_ms": 61.79662807378918, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o128_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o128_a1.json new file mode 100644 index 0000000..933f9d6 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o128_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 587.0072881928645, + "request_tps": 0.06814225445674156, + "input_tps": 4465.770788077015, + "output_tps": 8.72220857046292, + "total_tps": 4474.492996647478, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 128.0, + "ttft_p50_ms": 60264.91823955439, + "ttft_p95_ms": 116942.56678273669, + "ttft_p99_ms": 116968.76564948354, + "tpot_p50_ms": 419.5593314018012, + "tpot_p95_ms": 818.3646683727837, + "tpot_p99_ms": 829.4111081236225, + "e2e_p50_ms": 119654.99441663269, + "e2e_p95_ms": 121336.96261633886, + "e2e_p99_ms": 121339.05124852201, + "itl_p50_ms": 31.66217007674277, + "itl_p95_ms": 56.3370821531862, + "itl_p99_ms": 61.347583890892565, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o256_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o256_a1.json new file mode 100644 index 0000000..44cf69e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o256_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 608.7442546100356, + "request_tps": 0.06570903905388674, + "input_tps": 4306.307583435521, + "output_tps": 16.821513997795005, + "total_tps": 4323.129097433316, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 256.0, + "ttft_p50_ms": 60408.123059547506, + "ttft_p95_ms": 116816.32042041747, + "ttft_p99_ms": 117158.49739520112, + "tpot_p50_ms": 225.85036875497477, + "tpot_p95_ms": 424.8813510864662, + "tpot_p99_ms": 430.2625958223681, + "e2e_p50_ms": 123890.77430358157, + "e2e_p95_ms": 125879.0796181187, + "e2e_p99_ms": 125880.81930729328, + "itl_p50_ms": 31.708508962765336, + "itl_p95_ms": 55.004988797009, + "itl_p99_ms": 60.499521191231906, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o512_a1.json b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o512_a1.json new file mode 100644 index 0000000..cf82421 --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/metrics/sglang_c8_i65536_o512_a1.json @@ -0,0 +1,25 @@ +{ + "status": "COMPLETED", + "completed": 40, + "failed": 0, + "duration_s": 652.6434917969164, + "request_tps": 0.061289203834498415, + "input_tps": 4016.649262497688, + "output_tps": 31.38007236326319, + "total_tps": 4048.029334860951, + "mean_input_tokens": 65536.0, + "mean_output_tokens": 512.0, + "ttft_p50_ms": 60481.61956341937, + "ttft_p95_ms": 116843.18738668226, + "ttft_p99_ms": 117324.80349385645, + "tpot_p50_ms": 131.188325970543, + "tpot_p95_ms": 229.34224289243244, + "tpot_p99_ms": 231.58556326105642, + "e2e_p50_ms": 132822.63546145987, + "e2e_p95_ms": 134675.67310340237, + "e2e_p99_ms": 134677.4176377803, + "itl_p50_ms": 31.73411381430924, + "itl_p95_ms": 53.03707402199507, + "itl_p99_ms": 61.14733610767871, + "validation_errors": [] +} diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/server_cmd.txt b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/server_cmd.txt new file mode 100644 index 0000000..cff8c7e --- /dev/null +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/adaptive_results/adaptive_20260716-074243/tp8_dp1/server_cmd.txt @@ -0,0 +1 @@ +--host 0.0.0.0 --port 30014 --model-path /models --attention-backend nsa --nsa-prefill klxdsa --nsa-decode klxdsa --trust-remote-code --disable-custom-all-reduce --page-size 64 --mem-fraction-static 0.8 --tensor-parallel-size 8 --disable-shared-experts-fusion --quantization w8a8_int8 --kv-cache-dtype float16 --disable-piecewise-cuda-graph --cuda-graph-max-bs 32 --watchdog-timeout 3000000 --tool-call-parser deepseekv4 --reasoning-parser deepseek-v4 --constrained-json-disable-any-whitespace --enable-metrics --enable-request-time-stats-logging --context-length 140000 diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency.sh index 16b6056..a18ad88 100755 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency.sh @@ -7,15 +7,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh index 91176a6..dc82e78 100755 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -7,15 +7,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_bench.sh b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_bench.sh index bb643c8..bdbe1e5 100755 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_bench.sh +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -519,7 +519,7 @@ PY # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ + "$PYTHON" "${SCRIPT_DIR}/../../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/start_sglang_docker.sh b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/start_sglang_docker.sh index f19ebb4..c2b7a13 100755 --- a/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/start_sglang_docker.sh +++ b/experiments/p800/dsv4_p800_sglang_tp_dp_matrix/start_sglang_docker.sh @@ -15,9 +15,9 @@ fi SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" diff --git a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh index 0979a46..7972a9e 100755 --- a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh index 50f0b6b..c98b531 100755 --- a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="sglang" ENGINE_PORT="$SGLANG_PORT" diff --git a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_bench.sh b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_bench.sh index 0671523..a8bc490 100755 --- a/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_bench.sh +++ b/experiments/pro6000/dsv4_pro6000_sglang_tp_dp_matrix/run_bench.sh @@ -6,9 +6,9 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" @@ -522,7 +522,7 @@ run_parallel_config() { # Parse results. log "parsing ${config_label} results" - "$PYTHON" "${SCRIPT_DIR}/../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ + "$PYTHON" "${SCRIPT_DIR}/../../../scripts/common/parse_backend.py" "$result_root" --backend sglang \ >> "${phase_log_dir}/parse.log" 2>&1 || { log "WARNING: parser failed for ${config_label}; see ${phase_log_dir}/parse.log" } diff --git a/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh b/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh index ca23f6a..f76a93d 100755 --- a/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh +++ b/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh b/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh index aee8e20..fb18812 100755 --- a/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh +++ b/experiments/pro6000/dsv4_pro6000_vllm_tp_dp_matrix/run_adaptive_concurrency_add16.sh @@ -6,15 +6,15 @@ SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" EXPERIMENT_NAME="$(basename "$SCRIPT_DIR")" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/lib.sh" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/platform.sh" +source "${SCRIPT_DIR}/../../../scripts/common/platform.sh" # shellcheck source=/dev/null source "${SCRIPT_DIR}/config.env" # shellcheck source=/dev/null source "${SCRIPT_DIR}/adaptive_config.env" # shellcheck source=/dev/null -source "${SCRIPT_DIR}/../../scripts/common/adaptive_bench_lib.sh" +source "${SCRIPT_DIR}/../../../scripts/common/adaptive_bench_lib.sh" ENGINE="vllm" ENGINE_PORT="$VLLM_PORT" diff --git a/platforms/README.md b/platforms/README.md index 671fa70..f38ec98 100644 --- a/platforms/README.md +++ b/platforms/README.md @@ -20,6 +20,8 @@ PLATFORM=kunlun_p800 bash experiments/dsv4_p800_sglang/run_bench.sh |---|---|---|---| | `kunlun_p800.env` | Kunlun P800 XPU | `sglang-xpu` | Docker-based SGLang serving image | | `nvidia_h200.env` | NVIDIA H200 | `vllm-dspark` | Native host virtual environments | +| `nvidia_h20.env` | NVIDIA H20 | `vllm` / `sglang` | Docker-based(vllm-openai / sglang 官方镜像) | +| `nvidia_rtx6000d.env` | NVIDIA RTX 6000D | `vllm` / `sglang` | Docker-based,SM120 部署见 `envs/SM120_DSV4_DEPLOYMENT_GUIDE.md` | ## What belongs here diff --git a/scripts/common/adaptive_bench_lib.sh.bak.20260716-084832 b/scripts/common/adaptive_bench_lib.sh.bak.20260716-084832 deleted file mode 100755 index fcd06f4..0000000 --- a/scripts/common/adaptive_bench_lib.sh.bak.20260716-084832 +++ /dev/null @@ -1,778 +0,0 @@ -#!/usr/bin/env bash -# Shared adaptive-concurrency search loop for serving benchmarks. - -ADAPTIVE_COMMON_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" -ADAPTIVE_HELPER="${ADAPTIVE_COMMON_DIR}/adaptive_concurrency.py" - -adaptive_timestamp() { - date --iso-8601=seconds -} - -adaptive_json_number_or_null() { - local value="${1:-}" - if [[ -n "$value" ]]; then - printf '%s' "$value" - else - printf 'null' - fi -} - -adaptive_warmup_request_count() { - local concurrency="$1" - if (( BENCH_WARMUP_MAX_REQUESTS > 0 && concurrency > BENCH_WARMUP_MAX_REQUESTS )); then - printf '%s' "$BENCH_WARMUP_MAX_REQUESTS" - else - printf '%s' "$concurrency" - fi -} -export -f adaptive_warmup_request_count - -adaptive_validate_config() { - local name - for name in SEARCH_START_CONCURRENCY SEARCH_MAX_CONCURRENCY \ - NUM_PROMPTS_MULTIPLIER PLATEAU_PATIENCE MAX_POINT_RETRIES \ - BENCH_WARMUP_MAX_REQUESTS; do - if ! [[ "${!name}" =~ ^[0-9]+$ ]]; then - log "ERROR: ${name} must be a non-negative integer, got ${!name}" - return 1 - fi - done - if (( SEARCH_START_CONCURRENCY < 1 )); then - log "ERROR: SEARCH_START_CONCURRENCY must be at least 1" - return 1 - fi - if (( SEARCH_MAX_CONCURRENCY < SEARCH_START_CONCURRENCY )); then - log "ERROR: SEARCH_MAX_CONCURRENCY must be >= SEARCH_START_CONCURRENCY" - return 1 - fi - if [[ -n "${SEARCH_ADDEND:-}" ]]; then - if ! [[ "${SEARCH_ADDEND}" =~ ^[0-9]+$ ]] || (( SEARCH_ADDEND < 1 )); then - log "ERROR: SEARCH_ADDEND must be at least 1" - return 1 - fi - else - if ! [[ "${SEARCH_MULTIPLIER:-}" =~ ^[0-9]+$ ]] || (( SEARCH_MULTIPLIER < 2 )); then - log "ERROR: SEARCH_MULTIPLIER must be at least 2" - return 1 - fi - fi - if (( NUM_PROMPTS_MULTIPLIER < 1 || PLATEAU_PATIENCE < 1 )); then - log "ERROR: NUM_PROMPTS_MULTIPLIER and PLATEAU_PATIENCE must be at least 1" - return 1 - fi - if ! [[ "$TPS_MIN_GAIN_PCT" =~ ^[0-9]+([.][0-9]+)?$ ]]; then - log "ERROR: TPS_MIN_GAIN_PCT must be a non-negative number" - return 1 - fi - if ! [[ "$RANDOM_RANGE_RATIO" =~ ^[0-9]+([.][0-9]+)?$ ]] || \ - ! awk -v ratio="$RANDOM_RANGE_RATIO" 'BEGIN { exit !(ratio >= 0 && ratio <= 1) }'; then - log "ERROR: RANDOM_RANGE_RATIO must be between 0.0 and 1.0" - return 1 - fi - if ! [[ "${TTFT_SLO_MS:-4000}" =~ ^[0-9]+$ ]] || (( TTFT_SLO_MS < 1 )); then - log "ERROR: TTFT_SLO_MS must be a positive integer, got ${TTFT_SLO_MS}" - return 1 - fi - if [[ "${ENABLE_TTFT_SLO_STOP:-1}" != "0" && "${ENABLE_TTFT_SLO_STOP:-1}" != "1" ]]; then - log "ERROR: ENABLE_TTFT_SLO_STOP must be 0 or 1, got ${ENABLE_TTFT_SLO_STOP}" - return 1 - fi -} - -adaptive_start_gpu_monitor() { - local csv_path="$1" - mkdir -p "$(dirname "$csv_path")" - nvidia-smi \ - --query-gpu=timestamp,index,memory.used,memory.total,utilization.gpu \ - --format=csv \ - -l "$GPU_MEM_SAMPLE_INTERVAL_S" \ - > "$csv_path" 2>/dev/null & - echo $! -} - -adaptive_stop_gpu_monitor() { - local pid="$1" - if kill -0 "$pid" 2>/dev/null; then - kill "$pid" 2>/dev/null || true - sleep 1 - kill -9 "$pid" 2>/dev/null || true - fi -} - -adaptive_warmup() { - local log_path="$1" - log "running synthetic warmup" - timeout "$SCENARIO_TIMEOUT_S" bash -c \ - 'engine_run_bench 1024 128 1 1 /dev/null' \ - >> "$log_path" 2>&1 -} - -adaptive_restart_server() { - local tp="$1" - local dp="$2" - log "restarting ${ENGINE} server tp=${tp} dp=${dp}" - engine_stop_server "$tp" "$dp" - sleep "$SERVER_RESTART_COOLDOWN_S" - engine_start_server "$tp" "$dp" - adaptive_warmup "${ADAPTIVE_LOG_DIR}/${ENGINE}_tp${tp}_dp${dp}.warmup.log" -} - -adaptive_append_point_from_metrics() { - local metrics_file="$1" - local tp="$2" - local dp="$3" - local mark="$4" - local isl="$5" - local osl="$6" - local concurrency="$7" - local num_prompts="$8" - local attempt="$9" - local gain_pct="${10:-}" - local plateau_streak="${11:-0}" - local raw_file="${12}" - local detail_log="${13}" - local gain_json - gain_json="$(adaptive_json_number_or_null "$gain_pct")" - - jq -c \ - --arg timestamp "$(adaptive_timestamp)" \ - --arg engine "$ENGINE" \ - --argjson tp "$tp" \ - --argjson dp "$dp" \ - --arg mark "$mark" \ - --argjson isl "$isl" \ - --argjson osl "$osl" \ - --argjson concurrency "$concurrency" \ - --argjson num_prompts "$num_prompts" \ - --argjson warmup_requests "$(adaptive_warmup_request_count "$concurrency")" \ - --argjson attempt "$attempt" \ - --argjson gain_pct "$gain_json" \ - --argjson plateau_streak "$plateau_streak" \ - --arg raw_file "$raw_file" \ - --arg detail_log "$detail_log" \ - '. + { - timestamp: $timestamp, - engine: $engine, - tp: $tp, - dp: $dp, - mark: $mark, - isl: $isl, - osl: $osl, - concurrency: $concurrency, - num_prompts: $num_prompts, - warmup_requests: $warmup_requests, - attempt: $attempt, - gain_pct: $gain_pct, - plateau_streak: $plateau_streak, - error_type: "", - raw_file: $raw_file, - detail_log: $detail_log - }' "$metrics_file" >> "$ADAPTIVE_POINTS_JSONL" -} - -adaptive_append_failed_point() { - local tp="$1" - local dp="$2" - local mark="$3" - local isl="$4" - local osl="$5" - local concurrency="$6" - local num_prompts="$7" - local attempt="$8" - local status="$9" - local error_type="${10}" - local raw_file="${11}" - local detail_log="${12}" - - jq -nc \ - --arg timestamp "$(adaptive_timestamp)" \ - --arg engine "$ENGINE" \ - --argjson tp "$tp" \ - --argjson dp "$dp" \ - --arg mark "$mark" \ - --argjson isl "$isl" \ - --argjson osl "$osl" \ - --argjson concurrency "$concurrency" \ - --argjson num_prompts "$num_prompts" \ - --argjson warmup_requests "$(adaptive_warmup_request_count "$concurrency")" \ - --argjson attempt "$attempt" \ - --arg status "$status" \ - --arg error_type "$error_type" \ - --arg raw_file "$raw_file" \ - --arg detail_log "$detail_log" \ - '{ - timestamp: $timestamp, - engine: $engine, - tp: $tp, - dp: $dp, - mark: $mark, - isl: $isl, - osl: $osl, - concurrency: $concurrency, - num_prompts: $num_prompts, - warmup_requests: $warmup_requests, - attempt: $attempt, - status: $status, - completed: 0, - failed: $num_prompts, - error_type: $error_type, - validation_errors: [], - gain_pct: null, - plateau_streak: 0, - raw_file: $raw_file, - detail_log: $detail_log - }' >> "$ADAPTIVE_POINTS_JSONL" -} - -adaptive_append_shape_summary() { - local tp="$1" - local dp="$2" - local mark="$3" - local isl="$4" - local osl="$5" - local status="$6" - local stop_reason="$7" - local tested_points="$8" - local max_successful_concurrency="$9" - local saturation_concurrency="${10:-}" - local stop_probe_concurrency="${11:-}" - local best_tps_concurrency="${12:-}" - local best_total_tps="${13:-}" - local last_total_tps="${14:-}" - local saturation_json stop_probe_json best_c_json best_tps_json last_tps_json - saturation_json="$(adaptive_json_number_or_null "$saturation_concurrency")" - stop_probe_json="$(adaptive_json_number_or_null "$stop_probe_concurrency")" - best_c_json="$(adaptive_json_number_or_null "$best_tps_concurrency")" - best_tps_json="$(adaptive_json_number_or_null "$best_total_tps")" - last_tps_json="$(adaptive_json_number_or_null "$last_total_tps")" - - jq -nc \ - --arg timestamp "$(adaptive_timestamp)" \ - --arg engine "$ENGINE" \ - --argjson tp "$tp" \ - --argjson dp "$dp" \ - --arg mark "$mark" \ - --argjson isl "$isl" \ - --argjson osl "$osl" \ - --arg status "$status" \ - --arg stop_reason "$stop_reason" \ - --argjson tested_points "$tested_points" \ - --argjson search_cap "$SEARCH_MAX_CONCURRENCY" \ - --argjson max_successful_concurrency "$max_successful_concurrency" \ - --argjson saturation_concurrency "$saturation_json" \ - --argjson stop_probe_concurrency "$stop_probe_json" \ - --argjson best_tps_concurrency "$best_c_json" \ - --argjson best_total_tps "$best_tps_json" \ - --argjson last_total_tps "$last_tps_json" \ - '{ - timestamp: $timestamp, - engine: $engine, - tp: $tp, - dp: $dp, - mark: $mark, - isl: $isl, - osl: $osl, - status: $status, - stop_reason: $stop_reason, - tested_points: $tested_points, - search_cap: $search_cap, - max_successful_concurrency: $max_successful_concurrency, - saturation_concurrency: $saturation_concurrency, - stop_probe_concurrency: $stop_probe_concurrency, - best_tps_concurrency: $best_tps_concurrency, - best_total_tps: $best_total_tps, - last_total_tps: $last_total_tps - }' >> "$ADAPTIVE_SHAPES_JSONL" -} - -adaptive_run_point() { - local tp="$1" - local dp="$2" - local isl="$3" - local osl="$4" - local concurrency="$5" - local num_prompts="$6" - local config_root="$7" - local attempt bench_rc parse_rc gpu_pid - - POINT_RAW_FILE="" - POINT_DETAIL_LOG="" - POINT_METRICS_FILE="" - POINT_ATTEMPT=0 - POINT_ERROR_TYPE="" - - for (( attempt = 1; attempt <= MAX_POINT_RETRIES + 1; attempt++ )); do - POINT_ATTEMPT="$attempt" - POINT_RAW_FILE="${config_root}/raw_outputs/${ENGINE}_adaptive_c${concurrency}_i${isl}_o${osl}_a${attempt}.jsonl" - POINT_DETAIL_LOG="${config_root}/logs/${ENGINE}_c${concurrency}_i${isl}_o${osl}_a${attempt}.log" - POINT_METRICS_FILE="${config_root}/metrics/${ENGINE}_c${concurrency}_i${isl}_o${osl}_a${attempt}.json" - local gpu_csv="${config_root}/gpu_logs/gpu_c${concurrency}_i${isl}_o${osl}_a${attempt}.csv" - - log "probe ${ENGINE} tp=${tp} dp=${dp} isl=${isl} osl=${osl} c=${concurrency} warmup=$(adaptive_warmup_request_count "$concurrency") n=${num_prompts} attempt=${attempt}" - gpu_pid="$(adaptive_start_gpu_monitor "$gpu_csv")" - bench_rc=0 - timeout "$SCENARIO_TIMEOUT_S" bash -c \ - 'engine_run_bench "$@"' _ \ - "$isl" "$osl" "$concurrency" "$num_prompts" "$POINT_RAW_FILE" \ - > "$POINT_DETAIL_LOG" 2>&1 || bench_rc=$? - adaptive_stop_gpu_monitor "$gpu_pid" - - if (( bench_rc == 0 )); then - parse_rc=0 - "$PYTHON" "$ADAPTIVE_HELPER" parse-result \ - --input "$POINT_RAW_FILE" \ - --output "$POINT_METRICS_FILE" \ - --expected-prompts "$num_prompts" \ - --isl "$isl" \ - --osl "$osl" \ - --input-tolerance-pct "$INPUT_LENGTH_TOLERANCE_PCT" \ - --output-tolerance-pct "$OUTPUT_LENGTH_TOLERANCE_PCT" \ - >> "$POINT_DETAIL_LOG" 2>&1 || parse_rc=$? - if (( parse_rc == 0 )); then - return 0 - fi - POINT_ERROR_TYPE="workload validation failed" - return 21 - fi - - if engine_detect_oom "$POINT_DETAIL_LOG" "$tp" "$dp"; then - POINT_ERROR_TYPE="detected out of memory" - return 20 - fi - - POINT_ERROR_TYPE="benchmark rc=${bench_rc}" - if (( attempt <= MAX_POINT_RETRIES )); then - log "probe failed without OOM; restarting server before retry" - if ! adaptive_restart_server "$tp" "$dp"; then - POINT_ERROR_TYPE="benchmark failed and server restart failed" - return 22 - fi - continue - fi - return 22 - done -} - -adaptive_run_shape() { - local tp="$1" - local dp="$2" - local mark="$3" - local isl="$4" - local osl="$5" - local config_root="$6" - local concurrency="$SEARCH_START_CONCURRENCY" - local previous_tps="" - local current_tps="" - local best_tps="" - local best_concurrency="" - local max_successful_concurrency=0 - local last_total_tps="" - local plateau_streak=0 - local plateau_start_concurrency="" - local saturation_concurrency="" - local stop_probe_concurrency="" - local stop_reason="SEARCH_CAP_REACHED" - local shape_status="COMPLETED" - local tested_points=0 - local gain_pct="" - local meaningful_gain=1 - local point_rc=0 - - ADAPTIVE_RESTART_NEEDED=0 - ADAPTIVE_FATAL_WORKLOAD=0 - - while (( concurrency <= SEARCH_MAX_CONCURRENCY )); do - local num_prompts=$((concurrency * NUM_PROMPTS_MULTIPLIER)) - point_rc=0 - adaptive_run_point "$tp" "$dp" "$isl" "$osl" "$concurrency" "$num_prompts" "$config_root" || point_rc=$? - tested_points=$((tested_points + 1)) - stop_probe_concurrency="$concurrency" - - if (( point_rc == 0 )); then - current_tps="$(jq -r '.total_tps' "$POINT_METRICS_FILE")" - gain_pct="" - meaningful_gain=1 - if [[ -n "$previous_tps" ]]; then - IFS=$'\t' read -r gain_pct meaningful_gain < <( - "$PYTHON" "$ADAPTIVE_HELPER" gain \ - --previous "$previous_tps" \ - --current "$current_tps" \ - --threshold-pct "$TPS_MIN_GAIN_PCT" - ) - if (( meaningful_gain == 1 )); then - plateau_streak=0 - plateau_start_concurrency="" - else - plateau_streak=$((plateau_streak + 1)) - if [[ -z "$plateau_start_concurrency" ]]; then - plateau_start_concurrency="$concurrency" - fi - fi - fi - - if [[ -z "$best_tps" ]] || awk -v current="$current_tps" -v best="$best_tps" 'BEGIN { exit !(current > best) }'; then - best_tps="$current_tps" - best_concurrency="$concurrency" - fi - max_successful_concurrency="$concurrency" - last_total_tps="$current_tps" - - # TTFT SLO check: stop searching if TTFT P95 exceeds the SLO threshold - local ttft_p95_ms="" - if [[ "${ENABLE_TTFT_SLO_STOP:-1}" == "1" ]]; then - ttft_p95_ms="$(jq -r '.ttft_p95_ms' "$POINT_METRICS_FILE")" - if awk -v ttft="$ttft_p95_ms" -v slo="${TTFT_SLO_MS:-4000}" 'BEGIN { exit !(ttft > slo) }'; then - adaptive_append_point_from_metrics \ - "$POINT_METRICS_FILE" "$tp" "$dp" "$mark" "$isl" "$osl" \ - "$concurrency" "$num_prompts" "$POINT_ATTEMPT" "$gain_pct" \ - "$plateau_streak" "$POINT_RAW_FILE" "$POINT_DETAIL_LOG" - stop_reason="TTFT_SLO_EXCEEDED" - shape_status="TTFT_BOUNDARY" - log "probe result c=${concurrency} total_tps=${current_tps} ttft_p95=${ttft_p95_ms}ms > SLO=${TTFT_SLO_MS}ms, stopping shape" - break - fi - fi - - adaptive_append_point_from_metrics \ - "$POINT_METRICS_FILE" "$tp" "$dp" "$mark" "$isl" "$osl" \ - "$concurrency" "$num_prompts" "$POINT_ATTEMPT" "$gain_pct" \ - "$plateau_streak" "$POINT_RAW_FILE" "$POINT_DETAIL_LOG" - - log "probe result c=${concurrency} total_tps=${current_tps} gain_pct=${gain_pct:-first} plateau_streak=${plateau_streak}/${PLATEAU_PATIENCE}" - if (( plateau_streak >= PLATEAU_PATIENCE )); then - saturation_concurrency="$plateau_start_concurrency" - stop_reason="TPS_PLATEAU" - break - fi - previous_tps="$current_tps" - elif (( point_rc == 20 )); then - adaptive_append_failed_point \ - "$tp" "$dp" "$mark" "$isl" "$osl" "$concurrency" "$num_prompts" \ - "$POINT_ATTEMPT" "OOM" "$POINT_ERROR_TYPE" "$POINT_RAW_FILE" "$POINT_DETAIL_LOG" - shape_status="OOM_BOUNDARY" - stop_reason="OOM" - ADAPTIVE_RESTART_NEEDED=1 - break - elif (( point_rc == 21 )); then - adaptive_append_point_from_metrics \ - "$POINT_METRICS_FILE" "$tp" "$dp" "$mark" "$isl" "$osl" \ - "$concurrency" "$num_prompts" "$POINT_ATTEMPT" "" 0 \ - "$POINT_RAW_FILE" "$POINT_DETAIL_LOG" - shape_status="INVALID_WORKLOAD" - stop_reason="INVALID_WORKLOAD" - ADAPTIVE_FATAL_WORKLOAD=1 - break - else - adaptive_append_failed_point \ - "$tp" "$dp" "$mark" "$isl" "$osl" "$concurrency" "$num_prompts" \ - "$POINT_ATTEMPT" "FAILED" "$POINT_ERROR_TYPE" "$POINT_RAW_FILE" "$POINT_DETAIL_LOG" - shape_status="FAILED" - stop_reason="BENCHMARK_FAILED" - ADAPTIVE_RESTART_NEEDED=1 - break - fi - - if (( concurrency == SEARCH_MAX_CONCURRENCY )); then - stop_reason="SEARCH_CAP_REACHED" - break - fi - local next_concurrency - if [[ -n "${SEARCH_ADDEND:-}" ]]; then - next_concurrency=$((concurrency + SEARCH_ADDEND)) - else - next_concurrency=$((concurrency * SEARCH_MULTIPLIER)) - fi - if (( next_concurrency > SEARCH_MAX_CONCURRENCY )); then - next_concurrency="$SEARCH_MAX_CONCURRENCY" - fi - concurrency="$next_concurrency" - done - - adaptive_append_shape_summary \ - "$tp" "$dp" "$mark" "$isl" "$osl" "$shape_status" "$stop_reason" \ - "$tested_points" "$max_successful_concurrency" "$saturation_concurrency" \ - "$stop_probe_concurrency" "$best_concurrency" "$best_tps" "$last_total_tps" - - log "shape done tp=${tp} dp=${dp} isl=${isl} osl=${osl} stop=${stop_reason} saturation_c=${saturation_concurrency:-none} best_c=${best_concurrency:-none} best_total_tps=${best_tps:-none}" -} - -adaptive_matches_filter() { - local value="$1" - local filter="${2:-}" - [[ -z "$filter" || " $filter " == *" $value "* ]] -} - -adaptive_print_sequence() { - local concurrency="$SEARCH_START_CONCURRENCY" - local -a sequence=() - while (( concurrency <= SEARCH_MAX_CONCURRENCY )); do - sequence+=("$concurrency") - if (( concurrency == SEARCH_MAX_CONCURRENCY )); then - break - fi - local next_concurrency - if [[ -n "${SEARCH_ADDEND:-}" ]]; then - next_concurrency=$((concurrency + SEARCH_ADDEND)) - else - next_concurrency=$((concurrency * SEARCH_MULTIPLIER)) - fi - if (( next_concurrency > SEARCH_MAX_CONCURRENCY )); then - next_concurrency="$SEARCH_MAX_CONCURRENCY" - fi - concurrency="$next_concurrency" - done - printf '%s ' "${sequence[@]}" -} - -adaptive_cleanup_active_server() { - if [[ -n "${ADAPTIVE_ACTIVE_TP:-}" && -n "${ADAPTIVE_ACTIVE_DP:-}" ]]; then - engine_stop_server "$ADAPTIVE_ACTIVE_TP" "$ADAPTIVE_ACTIVE_DP" || true - ADAPTIVE_ACTIVE_TP="" - ADAPTIVE_ACTIVE_DP="" - fi -} - -adaptive_shape_already_tested() { - local tp="$1" - local dp="$2" - local mark="$3" - local isl="$4" - local osl="$5" - local resume_shapes_jsonl="$6" - if [[ ! -f "$resume_shapes_jsonl" ]]; then - return 1 - fi - local count - count="$(jq -s -r --argjson tp "$tp" --argjson dp "$dp" --arg mark "$mark" --argjson isl "$isl" --argjson osl "$osl" ' - [ .[] | select(.tp == $tp and .dp == $dp and .mark == $mark and .isl == $isl and .osl == $osl) ] | length - ' "$resume_shapes_jsonl" 2>/dev/null)" - if [[ -z "$count" ]]; then - return 1 - fi - if (( count > 0 )); then - return 0 - fi - return 1 -} - -adaptive_main() { - local resume_mode=0 - local resume_shapes_jsonl="" - if [[ -n "${RESUME_RUN_ID:-}" ]]; then - ADAPTIVE_RUN_ROOT="${RESULT_BASE}/${RESUME_RUN_ID}" - if [[ ! -d "$ADAPTIVE_RUN_ROOT" ]]; then - log "ERROR: RESUME_RUN_ID=${RESUME_RUN_ID} does not exist under ${RESULT_BASE}" - return 1 - fi - resume_mode=1 - RUN_ID="$RESUME_RUN_ID" - resume_shapes_jsonl="${ADAPTIVE_RUN_ROOT}/adaptive_shapes.jsonl" - if [[ ! -f "$resume_shapes_jsonl" ]]; then - log "WARNING: no adaptive_shapes.jsonl found in ${ADAPTIVE_RUN_ROOT}; will resume with empty history" - fi - log "RESUME mode: continuing run_id=${RESUME_RUN_ID}" - else - RUN_ID="${RUN_ID:-adaptive_$(date '+%Y%m%d-%H%M%S')}" - ADAPTIVE_RUN_ROOT="${RESULT_BASE}/${RUN_ID}" - fi - RESULT_BASE="${RESULT_BASE:-${SCRIPT_DIR}/adaptive_results}" - MATRIX_FILE="${MATRIX_FILE:-${SCRIPT_DIR}/matrix.json}" - MATRIX_MODE="${MATRIX_MODE:-Y}" - ADAPTIVE_LOG_DIR="${ADAPTIVE_RUN_ROOT}/logs" - ADAPTIVE_POINTS_JSONL="${ADAPTIVE_RUN_ROOT}/adaptive_points.jsonl" - ADAPTIVE_SHAPES_JSONL="${ADAPTIVE_RUN_ROOT}/adaptive_shapes.jsonl" - local shapes_tsv="${ADAPTIVE_RUN_ROOT}/shapes.tsv" - mkdir -p "$ADAPTIVE_LOG_DIR" - if (( resume_mode == 0 )); then - : > "$ADAPTIVE_POINTS_JSONL" - : > "$ADAPTIVE_SHAPES_JSONL" - fi - log_init "${ADAPTIVE_LOG_DIR}/orchestrator.log" - - adaptive_validate_config - "$PYTHON" "$ADAPTIVE_HELPER" shapes \ - --matrix "$MATRIX_FILE" \ - --mode "$MATRIX_MODE" \ - > "$shapes_tsv" - - local tokenize_prompt_json=true - if [[ "$BENCH_DATASET_NAME" == "random" ]]; then - local required_dataset_rows=$((SEARCH_MAX_CONCURRENCY * NUM_PROMPTS_MULTIPLIER)) - local dataset_capacity=0 - local dataset_error="" - if [[ ! -f "$DATASET_PATH" ]]; then - dataset_error="DATASET_PATH does not exist: ${DATASET_PATH}" - elif ! dataset_capacity="$( - "$PYTHON" "$ADAPTIVE_HELPER" dataset-capacity --path "$DATASET_PATH" 2>/dev/null - )"; then - dataset_error="DATASET_PATH is not a valid ShareGPT JSON file: ${DATASET_PATH}" - fi - if [[ -n "$dataset_error" ]]; then - if [[ "$DRY_RUN" == "1" ]]; then - log "WARNING: ${dataset_error}" - else - log "ERROR: ${dataset_error}" - return 1 - fi - fi - if [[ -z "$dataset_error" ]] && (( dataset_capacity < required_dataset_rows )); then - if [[ "$DRY_RUN" == "1" ]]; then - log "WARNING: DATASET_PATH has ${dataset_capacity} usable conversations, but a real run may request ${required_dataset_rows}." - else - log "ERROR: DATASET_PATH has ${dataset_capacity} usable conversations, but adaptive search may request ${required_dataset_rows}. Set DATASET_PATH to a larger ShareGPT file or explicitly use BENCH_DATASET_NAME=random-ids." - return 1 - fi - fi - tokenize_prompt_json=false - elif [[ "$BENCH_DATASET_NAME" != random* ]]; then - log "ERROR: BENCH_DATASET_NAME must be random or start with random" - return 1 - fi - - log "experiment=${EXPERIMENT_NAME} engine=${ENGINE} run_id=${RUN_ID}" - log "search_start=${SEARCH_START_CONCURRENCY} search_cap=${SEARCH_MAX_CONCURRENCY} multiplier=${SEARCH_MULTIPLIER} min_gain_pct=${TPS_MIN_GAIN_PCT} patience=${PLATEAU_PATIENCE} prompts_multiplier=${NUM_PROMPTS_MULTIPLIER} warmup_cap=${BENCH_WARMUP_MAX_REQUESTS}" - log "dataset=${BENCH_DATASET_NAME} random_range_ratio=${RANDOM_RANGE_RATIO} tokenize_prompt=${tokenize_prompt_json} matrix=${MATRIX_FILE} mode=${MATRIX_MODE}" - - jq -n \ - --arg experiment "$EXPERIMENT_NAME" \ - --arg engine "$ENGINE" \ - --arg run_id "$RUN_ID" \ - --arg model "$MODEL_PATH" \ - --arg hardware "$HARDWARE" \ - --arg matrix "$MATRIX_FILE" \ - --arg dataset "$BENCH_DATASET_NAME" \ - --argjson tokenize_prompt "$tokenize_prompt_json" \ - --argjson random_range_ratio "$RANDOM_RANGE_RATIO" \ - --argjson search_start "$SEARCH_START_CONCURRENCY" \ - --argjson search_cap "$SEARCH_MAX_CONCURRENCY" \ - --argjson search_multiplier "$SEARCH_MULTIPLIER" \ - --argjson prompts_multiplier "$NUM_PROMPTS_MULTIPLIER" \ - --argjson min_gain_pct "$TPS_MIN_GAIN_PCT" \ - --argjson plateau_patience "$PLATEAU_PATIENCE" \ - --argjson warmup_max_requests "$BENCH_WARMUP_MAX_REQUESTS" \ - --argjson ttft_slo_ms "${TTFT_SLO_MS:-4000}" \ - --argjson enable_ttft_slo_stop "${ENABLE_TTFT_SLO_STOP:-1}" \ - '{ - experiment: $experiment, - engine: $engine, - run_id: $run_id, - model: $model, - hardware: $hardware, - matrix: $matrix, - dataset: $dataset, - tokenize_prompt: $tokenize_prompt, - random_range_ratio: $random_range_ratio, - search: { - start_concurrency: $search_start, - max_concurrency: $search_cap, - multiplier: $search_multiplier, - num_prompts_multiplier: $prompts_multiplier, - min_tps_gain_pct: $min_gain_pct, - plateau_patience: $plateau_patience, - warmup_max_requests: $warmup_max_requests, - ttft_slo_ms: $ttft_slo_ms, - enable_ttft_slo_stop: $enable_ttft_slo_stop - } - }' > "${ADAPTIVE_RUN_ROOT}/run_manifest.json" - - ADAPTIVE_ACTIVE_TP="" - ADAPTIVE_ACTIVE_DP="" - trap adaptive_cleanup_active_server EXIT - trap 'adaptive_cleanup_active_server; exit 130' INT - trap 'adaptive_cleanup_active_server; exit 143' TERM - - local abort_all=0 - local cfg tp dp config_root server_args shape_count mark isl osl - for cfg in "${PARALLEL_CONFIGS[@]}"; do - read -r tp dp <<< "$cfg" - if ! adaptive_matches_filter "$tp" "$TP_LIST"; then - continue - fi - config_root="${ADAPTIVE_RUN_ROOT}/tp${tp}_dp${dp}" - mkdir -p "${config_root}/raw_outputs" "${config_root}/metrics" \ - "${config_root}/logs" "${config_root}/gpu_logs" - server_args="$(engine_build_server_args "$tp" "$dp")" - printf '%s\n' "$server_args" > "${config_root}/server_cmd.txt" - - if [[ "$DRY_RUN" == "1" ]]; then - log "DRY_RUN server tp=${tp} dp=${dp}: ${server_args}" - shape_count=0 - while IFS=$'\t' read -r mark isl osl; do - [[ "$mark" == "mark" ]] && continue - adaptive_matches_filter "$isl" "$ISL_LIST" || continue - adaptive_matches_filter "$osl" "$OSL_LIST" || continue - if (( GRID_LIMIT > 0 && shape_count >= GRID_LIMIT )); then - break - fi - shape_count=$((shape_count + 1)) - if (( resume_mode == 1 )) && adaptive_shape_already_tested "$tp" "$dp" "$mark" "$isl" "$osl" "$resume_shapes_jsonl"; then - log "DRY_RUN RESUME skip already tested shape tp=${tp} dp=${dp} isl=${isl} osl=${osl} mark=${mark}" - continue - fi - log "DRY_RUN tp=${tp} dp=${dp} isl=${isl} osl=${osl} C=[$(adaptive_print_sequence)]" - done < "$shapes_tsv" - continue - fi - - ADAPTIVE_ACTIVE_TP="$tp" - ADAPTIVE_ACTIVE_DP="$dp" - if ! engine_start_server "$tp" "$dp"; then - log "ERROR: failed to start ${ENGINE} tp=${tp} dp=${dp}; skipping config" - adaptive_cleanup_active_server - continue - fi - adaptive_warmup "${ADAPTIVE_LOG_DIR}/${ENGINE}_tp${tp}_dp${dp}.warmup.log" || { - log "ERROR: warmup failed for tp=${tp} dp=${dp}; skipping config" - adaptive_cleanup_active_server - continue - } - - shape_count=0 - while IFS=$'\t' read -r mark isl osl; do - [[ "$mark" == "mark" ]] && continue - adaptive_matches_filter "$isl" "$ISL_LIST" || continue - adaptive_matches_filter "$osl" "$OSL_LIST" || continue - if (( GRID_LIMIT > 0 && shape_count >= GRID_LIMIT )); then - log "GRID_LIMIT=${GRID_LIMIT} reached for tp=${tp} dp=${dp}" - break - fi - shape_count=$((shape_count + 1)) - if (( resume_mode == 1 )) && adaptive_shape_already_tested "$tp" "$dp" "$mark" "$isl" "$osl" "$resume_shapes_jsonl"; then - log "RESUME skip already tested shape tp=${tp} dp=${dp} isl=${isl} osl=${osl} mark=${mark}" - continue - fi - log "===== shape START tp=${tp} dp=${dp} isl=${isl} osl=${osl} mark=${mark} =====" - adaptive_run_shape "$tp" "$dp" "$mark" "$isl" "$osl" "$config_root" - - if (( ADAPTIVE_FATAL_WORKLOAD == 1 )); then - log "ERROR: invalid synthetic workload; aborting remaining searches" - abort_all=1 - break - fi - if (( ADAPTIVE_RESTART_NEEDED == 1 )); then - if ! adaptive_restart_server "$tp" "$dp"; then - log "ERROR: server restart failed; skipping remaining shapes for tp=${tp} dp=${dp}" - break - fi - fi - done < "$shapes_tsv" - - adaptive_cleanup_active_server - if (( abort_all == 1 )); then - break - fi - done - - "$PYTHON" "$ADAPTIVE_HELPER" summarize \ - --points "$ADAPTIVE_POINTS_JSONL" \ - --shapes "$ADAPTIVE_SHAPES_JSONL" \ - --output-dir "$ADAPTIVE_RUN_ROOT" \ - >> "${ADAPTIVE_LOG_DIR}/summarize.log" 2>&1 - log "adaptive results saved to ${ADAPTIVE_RUN_ROOT}" - - trap - EXIT INT TERM - adaptive_cleanup_active_server - if (( abort_all == 1 )); then - return 2 - fi -} diff --git a/scripts/common/dp_proxy.py b/scripts/common/dp_proxy.py deleted file mode 100644 index ff69bb1..0000000 --- a/scripts/common/dp_proxy.py +++ /dev/null @@ -1,173 +0,0 @@ -#!/usr/bin/env python3 -"""Lightweight round-robin reverse proxy for vLLM data-parallel child servers. - -vLLM's --data-parallel-multi-port-external-lb only exposes aggregated /health on -a supervisor port; it does not forward /v1/models, /v1/completions, etc. This -proxy sits in front of the child API servers and round-robins requests across -them. - -Usage: - python3 dp_proxy.py --port 30040 --backends 30030,30031,30032,30033 -""" -import argparse -import asyncio -import itertools -from contextlib import asynccontextmanager - -import httpx -from fastapi import FastAPI, Request, Response -from fastapi.responses import StreamingResponse -from uvicorn import Config, Server - - -HOP_BY_HOP_HEADERS = { - "connection", - "keep-alive", - "proxy-authenticate", - "proxy-authorization", - "te", - "trailers", - "transfer-encoding", - "upgrade", - "content-encoding", - "content-length", -} - - -class DPProxy: - def __init__(self, backends: list[int], timeout: float = 60.0): - self.backend_urls = [f"http://127.0.0.1:{port}" for port in backends] - self.cycle = itertools.cycle(self.backend_urls) - self.lock = asyncio.Lock() - self.timeout = timeout - self.client = httpx.AsyncClient( - timeout=httpx.Timeout(timeout), - limits=httpx.Limits(max_connections=100, max_keepalive_connections=20), - ) - - async def close(self): - await self.client.aclose() - - async def _next_backend(self) -> str: - async with self.lock: - return next(self.cycle) - - async def _is_healthy(self, base_url: str) -> bool: - try: - r = await self.client.get(f"{base_url}/health", timeout=2.0) - return r.status_code == 200 - except Exception: - return False - - async def health(self) -> Response: - results = await asyncio.gather( - *(self._is_healthy(url) for url in self.backend_urls) - ) - if all(results): - return Response(status_code=200) - return Response(status_code=503) - - async def ready(self) -> Response: - # Bench clients use /v1/models as the ready signal. We forward that - # to a healthy backend; this endpoint just checks /health. - return await self.health() - - async def proxy(self, request: Request) -> Response: - path = request.url.path - if request.url.query: - path = f"{path}?{request.url.query}" - - body = await request.body() - headers = dict(request.headers) - headers.pop("host", None) - - tried = set() - while len(tried) < len(self.backend_urls): - base_url = await self._next_backend() - if base_url in tried: - # Avoid infinite loop when only one backend remains. - break - tried.add(base_url) - - target = f"{base_url}{path}" - try: - backend_req = self.client.build_request( - method=request.method, - url=target, - headers=headers, - content=body, - ) - backend_resp = await self.client.send( - backend_req, follow_redirects=False, stream=True - ) - except Exception: - continue - - if backend_resp.status_code >= 500: - await backend_resp.aclose() - continue - - response_headers = { - k: v - for k, v in backend_resp.headers.items() - if k.lower() not in HOP_BY_HOP_HEADERS - } - return StreamingResponse( - backend_resp.aiter_bytes(), - status_code=backend_resp.status_code, - headers=response_headers, - background=None, - ) - - return Response( - content=b"All DP backends are unavailable", - status_code=503, - ) - - -def make_app(proxy: DPProxy) -> FastAPI: - app = FastAPI() - - @app.get("/health", include_in_schema=False) - async def health(): - return await proxy.health() - - @app.get("/ready", include_in_schema=False) - @app.get("/readyz", include_in_schema=False) - async def ready(): - return await proxy.ready() - - @app.api_route("/{path:path}", methods=["GET", "POST", "PUT", "DELETE", "OPTIONS"]) - async def catch_all(request: Request): - return await proxy.proxy(request) - - return app - - -async def main(): - parser = argparse.ArgumentParser() - parser.add_argument("--port", type=int, required=True) - parser.add_argument( - "--backends", - type=lambda s: [int(x.strip()) for x in s.split(",")], - required=True, - help="Comma-separated list of child vLLM server ports", - ) - parser.add_argument("--timeout", type=float, default=300.0) - parser.add_argument("--log-level", default="warning") - args = parser.parse_args() - - proxy = DPProxy(args.backends, timeout=args.timeout) - app = make_app(proxy) - - config = Config(app, host="0.0.0.0", port=args.port, log_level=args.log_level) - server = Server(config) - - try: - await server.serve() - finally: - await proxy.close() - - -if __name__ == "__main__": - asyncio.run(main())