[CI] fix test: use minimum supported layered output count (#2350)

Signed-off-by: David Chen <530634352@qq.com>
This commit is contained in:
WeiQing Chen authored and GitHub committed 2026-03-31 14:59:00 +08:00
1 parent c1a978ae8d
commit 0cbd1cbeb8
1 file changed
+3 -3
@@ -7,7 +7,7 @@ and are supported by Qwen-Image-Layered model.
Kept cases (maximum feature coverage, no cpu-offload):
sp_001 : cache_dit + Ulysses-SP 2 (2×H100)
cfg_parallel_002 : cache_dit + CFG-Parallel 2 (2×H100)
layers_guard_001 : layerwise offload + layers=2 (1×H100, issue #1969 guard)
layers_guard_001 : layerwise offload + layers=3 (1×H100, issue #1969 guard)
Total distinct features covered: cache_dit, Ulysses-SP, CFG-Parallel, layerwise-offload.
"""
@@ -102,8 +102,8 @@ LAYERS_GUARD_CASES = [
model=MODEL,
server_args=["--enable-layerwise-offload"],
),
2,
id="layers_guard_001_layers2",
3,
id="layers_guard_001_layers3",
marks=SINGLE_CARD_FEATURE_MARKS,
),
]