dsr1-fp4-1k1k-mid-curve

concurrency 2,048 · 13 days ago

cron

passed

re-run

Commit

SHA9788c8e

[RL] Handle Mooncake buffers across memory release (#27696)

zhuzilin·13 days ago

PR #27696 · [RL] Handle Mooncake buffers across memory release

Run

GitHub Actions27321587026-1

Slurm job5293

GPUs48·prefill 16 / decode 32

ISL / OSL1024 / 1024

Logscron/27321587026-1/1k1k/dsr1-fp4-1k1k-mid-curve/2026-06-11/5293/sa-bench_isl_1024_osl_1024/

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

20,480

duration

212

max_concurrency

2,048

mean_e2el_ms

20,487ms

mean_itl_ms

994ms

mean_tpot_ms

20.29ms

mean_ttft_ms

1,787ms

median_e2el_ms

19,964ms

median_itl_ms

1,020ms

median_tpot_ms

20.50ms

median_ttft_ms

823ms

num_prompts

20,480

output_throughput

88,937tok/s

p99_e2el_ms

33,361ms

p99_itl_ms

1,557ms

p99_tpot_ms

22.23ms

p99_ttft_ms

12,784ms

peak_output_tokens_per_s

103,377s

request_throughput

96.43tok/s

std_e2el_ms

3,149ms

std_itl_ms

203ms

std_tpot_ms

1.31ms

std_ttft_ms

2,806ms

total_input_tokens

18,880,692

total_output_tokens

18,888,974

total_token_throughput

177,836tok/s