sglang/ GB200 perf

dsr1-fp4-1k1k-mid-curve

concurrency 8,192 · 11 days ago

cron
passed
re-run
Commit

chore: bump tokenspeed_mla 0.1.1 -> 0.1.6 (#28116)

kpham-sgl·11 days ago
PR #28116 · chore: bump tokenspeed_mla 0.1.1 -> 0.1.6
Run
GitHub Actions27455007552-1
Slurm job5301
GPUs48·prefill 16 / decode 32
ISL / OSL1024 / 1024

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

81,920

duration

693

max_concurrency

8,192

mean_e2el_ms

67,035ms

mean_itl_ms

1,196ms

mean_tpot_ms

24.40ms

mean_ttft_ms

44,572ms

median_e2el_ms

69,458ms

median_itl_ms

1,040ms

median_tpot_ms

22.04ms

median_ttft_ms

49,574ms

num_prompts

81,920

output_throughput

108,900tok/s

p99_e2el_ms

85,169ms

p99_itl_ms

3,283ms

p99_tpot_ms

41.47ms

p99_ttft_ms

54,865ms

peak_output_tokens_per_s

152,976s

request_throughput

118tok/s

std_e2el_ms

8,281ms

std_itl_ms

613ms

std_tpot_ms

6.94ms

std_ttft_ms

10,988ms

total_input_tokens

75,511,905

total_output_tokens

75,496,082

total_token_throughput

217,823tok/s