dsr1-fp4-1k1k-mid-curve

concurrency 512 · 15 days ago

cron

passed

re-run

Commit

SHA0f86738

test: fix gemma GSM8K thresholds in nightly text eval (#27342)

kpham-sgl·15 days ago

PR #27342 · test: fix gemma GSM8K thresholds in nightly text eval

Run

GitHub Actions27181662432-1

Slurm job5284

GPUs48·prefill 16 / decode 32

ISL / OSL1024 / 1024

Logscron/27181662432-1/1k1k/dsr1-fp4-1k1k-mid-curve/2026-06-09/5284/sa-bench_isl_1024_osl_1024/

Metrics

28 captured

best_of

1.00

burstiness

1.00

completed

5,120

duration

162

max_concurrency

512

mean_e2el_ms

15,628ms

mean_itl_ms

774ms

mean_tpot_ms

15.80ms

mean_ttft_ms

1,062ms

median_e2el_ms

15,581ms

median_itl_ms

770ms

median_tpot_ms

15.95ms

median_ttft_ms

772ms

num_prompts

5,120

output_throughput

29,096tok/s

p99_e2el_ms

19,167ms

p99_itl_ms

1,063ms

p99_tpot_ms

16.55ms

p99_ttft_ms

4,289ms

peak_output_tokens_per_s

35,461s

request_throughput

31.55tok/s

std_e2el_ms

1,284ms

std_itl_ms

153ms

std_tpot_ms

0.73ms

std_ttft_ms

976ms

total_input_tokens

4,717,859

total_output_tokens

4,722,209

total_token_throughput

58,166tok/s