All pull requests

Pull requests

• feat(llm-bench): per-row speculation acceptance on replay (stacked on #151)

  #152·igor-fw opened Nov 10, 2026Collaborator·

  ·

• feat(llm-bench): measurement fields for the runner's checks

  #151·igor-fw opened Nov 10, 2026Collaborator·

  ·

• feat(llm-bench): add synchronized cohort decode lpt

  #150·mianwei-fai opened Nov 9, 2026·

  ·

• feat(llm-bench): per-class reasoning-effort tier shares, assigned per prefix (stacked on #146)

  #148·igor-fw opened Nov 3, 2026Collaborator·

  ·
  1

• fix(llm-bench): add a translation lead-in to the limerick prompt

  #142·janderson4 opened Oct 27, 2026Contributor·

  ·

• fix(llm-bench): pin truststore<0.10.5 to fix gevent TLS handshake crash

  #139·dwsun-fireworks-ai opened Oct 12, 2026·

  ·

• fix(llm-bench): count inference requests in failure checks

  #136·kiteretsu903 opened Sep 14, 2026·

  ·
  3

• feat(llm-bench): add video input load testing (video_url + slicing)

  #135·jonathancaevans opened Aug 17, 2026Contributor·

  ·
  3

• feat(llm-bench): route generation sweeps and sustained load

  #134·ishaan-shivhare opened Aug 5, 2026Contributor·

  ·
  6

• feat(load_test): add --session-mode for progressively growing prompts

  #131·nithiyn opened Jul 26, 2026·

  ·
  2

• Bump pillow from 10.0.0 to 12.3.0 in /llm_bench

  dependenciespython
  #129·dependabot[bot] opened Jul 20, 2026Bot·

  ·

• feat(llm-bench): add ragged prompt batches

  #128·janderson4 opened Jul 17, 2026Contributor·10 of 14 tasks·

  ·
  24

• feat(load_test): add optional per-user x-session-affinity

  #127·yunyichi opened Jul 3, 2026·

  ·
  1

• Make load_test.py always strict (remove conditional branch)

  #122·ishaan-shivhare opened Jun 6, 2026Contributor·1 of 2 tasks·

  ·
  5

• Make load_test.py unconditionally strict for limericks/code

  #121·ishaan-shivhare opened Jun 6, 2026Contributor·1 of 2 tasks·

  ·
  3

• Strict prompt construction in load_test.py (minimal)

  #120·ishaan-shivhare opened Jun 6, 2026Contributor·

  ·
  4

• Make load_test.py use strict token-id prompts (PER-75)

  #119·ishaan-shivhare opened Jun 5, 2026Contributor·

  ·
  4

• fix: add units to result summary output

  #118·hashwnath opened Jun 1, 2026·3 tasks·

  ·
  1

  2

• llm_bench: add --targets for multi-deployment load testing

  #116·aidando73 opened May 26, 2026Contributor·4 tasks·

  ·
  4

• llm_bench: auto-detect TRT-LLM rejection of Fireworks request extras

  #109·kevinle-fw opened May 8, 2026Contributor·

  ·
  2

• FAME v0: MiniMax M2.7 attention kernel profiling

  #104·vedularaghu opened May 1, 2026Contributor·4 tasks done·

  ·

• Add GLM 5.1 TPS benchmarking support

  #83·teofeliu opened Apr 13, 2026·

  ·

• Improve embeddings load test with batch support, new API params, and proper metrics

  #64·andrefoo opened Mar 3, 2026·5 tasks done·

  ·

• Prompt cache length default

  #61·aidando73 opened Feb 22, 2026Contributor·

  ·
  2

• test aiperf

  #48·tantara opened Jan 30, 2026Contributor·

  ·