dsr1-fp4-1k1k-mid-curve

concurrency 2,048 · 9 days ago

cron

passed

re-run

Commit

SHA0417951

[Bug Fix] Validate tokenizer-dependent features with skip_tokenizer_init (#27882)

rllin·9 days ago

PR #27882 · [Bug Fix] Validate tokenizer-dependent features with skip_tokenizer_init

Run

GitHub Actions27522040539-1

Slurm job5308

GPUs48·prefill 16 / decode 32

ISL / OSL1024 / 1024

Logscron/27522040539-1/1k1k/dsr1-fp4-1k1k-mid-curve/2026-06-15/5308/sa-bench_isl_1024_osl_1024/

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

20,480

duration

215

max_concurrency

2,048

mean_e2el_ms

20,755ms

mean_itl_ms

1,004ms

mean_tpot_ms

20.50ms

mean_ttft_ms

1,867ms

median_e2el_ms

20,189ms

median_itl_ms

1,041ms

median_tpot_ms

20.91ms

median_ttft_ms

819ms

num_prompts

20,480

output_throughput

87,912tok/s

p99_e2el_ms

33,747ms

p99_itl_ms

1,539ms

p99_tpot_ms

21.83ms

p99_ttft_ms

14,168ms

peak_output_tokens_per_s

106,437s

request_throughput

95.32tok/s

std_e2el_ms

3,369ms

std_itl_ms

194ms

std_tpot_ms

1.37ms

std_ttft_ms

3,141ms

total_input_tokens

18,880,692

total_output_tokens

18,888,974

total_token_throughput

175,785tok/s