From 3a5483d339a7eb790bfa0b8d3deffad6e5636db2 Mon Sep 17 00:00:00 2001 From: WeiQing Chen <40507679+david6666666@users.noreply.github.com> Date: Thu, 26 Mar 2026 22:03:03 +0800 Subject: [PATCH] [CI] fix Wan22 timeout and i2i accuracy threshold (#2235) Signed-off-by: David Chen <530634352@qq.com> --- .buildkite/test-nightly.yml | 2 +- tests/e2e/accuracy/test_gedit_bench_h100_smoke.py | 12 ++++++------ 2 files changed, 7 insertions(+), 7 deletions(-) diff --git a/.buildkite/test-nightly.yml b/.buildkite/test-nightly.yml index e7f46f487..cbaf83da9 100644 --- a/.buildkite/test-nightly.yml +++ b/.buildkite/test-nightly.yml @@ -112,7 +112,7 @@ steps: type: DirectoryOrCreate - label: ":full_moon: Diffusion Model Wan22 completed Test with H100" - timeout_in_minutes: 45 + timeout_in_minutes: 90 depends_on: upload-nightly-pipeline if: build.env("NIGHTLY") == "1" || build.pull_request.labels includes "nightly-test" commands: diff --git a/tests/e2e/accuracy/test_gedit_bench_h100_smoke.py b/tests/e2e/accuracy/test_gedit_bench_h100_smoke.py index 80359fb13..ac5f2cb3c 100644 --- a/tests/e2e/accuracy/test_gedit_bench_h100_smoke.py +++ b/tests/e2e/accuracy/test_gedit_bench_h100_smoke.py @@ -106,9 +106,9 @@ def test_gedit_bench_h100_smoke( group_summary = language_summary["by_group"][group] assert set(group_summary) == {"count", "Q_SC", "Q_PQ", "Q_O"} - assert summary["languages"]["en"]["overall"]["Q_SC"] >= 7.1 - assert summary["languages"]["en"]["overall"]["Q_PQ"] >= 5.9 - assert summary["languages"]["en"]["overall"]["Q_O"] >= 6.3 - assert summary["languages"]["cn"]["overall"]["Q_SC"] >= 7.0 - assert summary["languages"]["cn"]["overall"]["Q_PQ"] >= 5.8 - assert summary["languages"]["cn"]["overall"]["Q_O"] >= 6.2 + assert summary["languages"]["en"]["overall"]["Q_SC"] >= 7.0 + assert summary["languages"]["en"]["overall"]["Q_PQ"] >= 5.8 + assert summary["languages"]["en"]["overall"]["Q_O"] >= 6.2 + assert summary["languages"]["cn"]["overall"]["Q_SC"] >= 6.9 + assert summary["languages"]["cn"]["overall"]["Q_PQ"] >= 5.7 + assert summary["languages"]["cn"]["overall"]["Q_O"] >= 6.1