sglang/ GB200 perf

dsr1-fp4-1k1k-mid-curve

concurrency 4,096 · 12 days ago

cron
passed
re-run
Commit

add lfm2.5 to new cookbook. (#27409)

vincentzed·12 days ago
PR #27409 · add lfm2.5 to new cookbook.
Run
GitHub Actions27392321842-1
Slurm job5299
GPUs48·prefill 16 / decode 32
ISL / OSL1024 / 1024

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

40,960

duration

308

max_concurrency

4,096

mean_e2el_ms

30,049ms

mean_itl_ms

1,156ms

mean_tpot_ms

23.60ms

mean_ttft_ms

8,329ms

median_e2el_ms

29,811ms

median_itl_ms

1,148ms

median_tpot_ms

23.84ms

median_ttft_ms

8,719ms

num_prompts

40,960

output_throughput

122,420tok/s

p99_e2el_ms

47,762ms

p99_itl_ms

2,225ms

p99_tpot_ms

28.57ms

p99_ttft_ms

23,133ms

peak_output_tokens_per_s

159,281s

request_throughput

133tok/s

std_e2el_ms

4,289ms

std_itl_ms

332ms

std_tpot_ms

3.15ms

std_ttft_ms

4,101ms

total_input_tokens

37,769,666

total_output_tokens

37,742,239

total_token_throughput

244,929tok/s