6 Commits

Author SHA1 Message Date
yy-fighting
4ea6e92b42 sglang tp_dp_matrix: fix timeout calling bash function (rc=127)
timeout(1) cannot execute shell functions directly; it needs a standalone
command. Export run_bench_serving via export -f and wrap both warmup and
scenario runs in bash -c so the function is visible in the subshell.
2026-07-10 02:51:29 +00:00
yy-fighting
06cfab20d2 sglang tp_dp_matrix: keep 1M ISL (128 Y, 256 P) in scope 2026-07-10 02:18:34 +00:00
yy-fighting
f7f4e928d9 sglang tp_dp_matrix: exclude 1M ISL from mandatory matrix per current scope 2026-07-10 02:17:38 +00:00
yy-fighting
7852dd2040 sglang tp_dp_matrix: run SGLang server/client in Docker on H200
- config.env: switch to /data/models/DeepSeek-V4-Flash, USE_DOCKER=1,
  CONTEXT_LENGTH=1M, CONCURRENCY_SAMPLES=2, add DATASET_PATH.
- start_sglang_docker.sh: add --ipc host --shm-size 16g for NCCL.
- start_sglang_dp.sh: use --model-path for Docker CLI.
- run_bench.sh: mount dataset into client container and pass --dataset-path
to avoid HF download on offline nodes.
2026-07-10 02:17:00 +00:00
Root User
db419e17b8 dsv4_h200_sglang_tp_dp_matrix: sample only low/high concurrency, use RUNTIME_BASE for self-contained runtime files 2026-07-09 15:36:03 +00:00
Root User
8d61e682cd add dsv4_h200_sglang_tp_dp_matrix experiment (Docker-based SGLang TP×DP matrix) 2026-07-09 14:49:31 +00:00