engineering/arborist
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

bench: bump cross-model bench timeout default 180s → 600s

Browse source 
Real-world: shared endpoints can queue (hermes was queued behind
opencompletion this morning, single calls pushed past 180s but
completed in 121s on retry). 600s tolerates ~10× the typical
non-queued latency before declaring timeout.
This commit is contained in:
russell@unturf.com 2026-05-30 22:19:43 -04:00
parent d8ed8fbd5b
commit ae842555d8
No known key found for this signature in database
1 changed files with 1 additions and 1 deletions
Download patch file Download diff file Expand all files Collapse all files

2
Makefile
View file

@ -432,7 +432,7 @@ bench-emergent-pending: bootstrap ## print log entries awaiting teacher review
# options" pattern that uncloseai-cli's tool_arborist supports.
BENCH_CM_QUESTIONS ?= bench/qa_questions_smoke.txt
BENCH_CM_OUT ?= bench/cross_model_results
BENCH_CM_TIMEOUT ?= 180
BENCH_CM_TIMEOUT ?= 600
BENCH_CM_LIMIT ?= 0
bench-cross-model: bootstrap ## cross-model self-play (Hermes + Qwen on smoke fixture) [BENCH_CM_QUESTIONS=... LIMIT=N BURN=1]
PYTHONUNBUFFERED=1 $(PY) bench/cross_model_selfplay.py \