All pull requests

Pull requests

• merge queue: checking #1288 on main (6168d28)

  #1327·mergify[bot] opened Oct 16, 2026Bot·7 of 47 tasks·

  ·

• Support multi-model round-robin for multi-LoRA benchmarking

  #1309·bbenshab opened Oct 15, 2026·

  ·
  4

• Support stacking of --config files

  clifeatureinternal
  #1299·dbutenhof opened Oct 13, 2026Collaborator·3 tasks done·

  ·
  1

  

• Add a ramp option to sweep for finding the peak rate

  #1295·theminjunchoi opened Oct 12, 2026Contributor·7 of 8 tasks·

  ·
  1

  1

• Write JSON/YAML outputs to paths without a known suffix

  #1294·Windyskr opened Oct 11, 2026·5 tasks done·

  ·
  1

  2

• Report worker loop lag before the first token

  #1293·QHarshil opened Oct 11, 2026Contributor·8 tasks done·

  ·
  6

• fix(openai): preserve mixed tool response text in history

  #1291·git-jxj opened Oct 11, 2026Contributor·3 tasks done·

  ·
  1

• fix(csv): retain zero-valued measurement statistics

  #1290·git-jxj opened Oct 11, 2026Contributor·3 tasks done·

  ·
  3

• fix: keep resized image dimensions positive

  #1289·git-jxj opened Oct 11, 2026Contributor·3 tasks done·

  ·

• Pass rampup_duration to constant sweep steps

  queued
  #1288·Jackmeson1 opened Oct 11, 2026·5 tasks done·

  ·
  1

  4

• docs: translate benchmark guide to Chinese

  #1285·tangming1996 opened Oct 10, 2026Contributor·2 of 3 tasks·

  ·
  5

• docs: translate analyze guide into Chinese

  #1284·tangming1996 opened Oct 10, 2026Contributor·2 of 3 tasks·

  ·
  4

• docs: translate server getting started guide to Chinese

  #1283·tangming1996 opened Oct 10, 2026Contributor·2 of 3 tasks·

  ·
  7

• fix(trace): preserve minimum overlap for unequal session lengths

  #1281·git-jxj opened Oct 10, 2026Contributor·3 tasks done·

  ·

• Scheduler data prefetch

  #1280·jaredoconnell opened Oct 9, 2026Collaborator·3 tasks done·

  ·
  1

  39

• fix: follow redirects when downloading remote media

  #1271·git-jxj opened Oct 9, 2026Contributor·3 tasks done·

  ·
  2

• feat: report output-token goodput and per-objective SLO attainment

  #1270·chinmay29 opened Oct 8, 2026·3 tasks done·

  ·
  16

• feat(console): print per-turn latency table for multi-turn workloads

  #1266·huangpengtao00-dotcom opened Oct 8, 2026Contributor·5 of 6 tasks·

  ·

• feat(metrics): add conversation-level metrics for multi-turn workloads

  #1265·huangpengtao00-dotcom opened Oct 8, 2026Contributor·9 of 10 tasks·

  ·

• Scrape server metrics from Prometheus endpoints during benchmarks

  #1250·theminjunchoi opened Oct 7, 2026Contributor·7 tasks done·

  ·
  1

  3

• Add target_moe constraint to stop at a target margin of error

  #1247·theminjunchoi opened Oct 7, 2026Contributor·7 tasks done·

  ·
  1

  1

• Minimum prefetch and related warnings

  #1239·jaredoconnell opened Oct 5, 2026Collaborator·3 tasks done·

  ·

• fix(scheduler): drain terminal updates before completing

  #1223·xujiantop-crypto opened Oct 2, 2026·6 tasks done·

  ·
  1

  3

• Optimize synthetic generation on Mooncake and WEKA

  #1222·jaredoconnell opened Oct 2, 2026Collaborator·3 tasks done·

  ·
  1

• Compute the per-modality token total per request

  #1211·David-Wu1119 opened Oct 1, 2026Contributor·5 tasks done·

  ·
  1