A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 02:15 UTC
Renable google/gemma-3-1b-it accuracy test. (#20866)
Signed-off-by: Qiliang Cui <derrhein@gmail.com>
QiliangCui committed
Jul 12, 2025 at 21:48 UTC
99b4f080d83ae284941b01922d7fe3b9a39034fd
1 file changed
+2
-3
tests/entrypoints/llm/test_accuracy.py
+2
-3
@@ -71,9 +71,8 @@ def test_lm_eval_accuracy_v1_engine(model, monkeypatch: pytest.MonkeyPatch):
71
# Limit compilation time for TPU V1
72
73
if model == "google/gemma-3-1b-it":
74
- pytest.skip(
75
- "Temporarily disabled due to test failures"
76
- "(timeout or accuracy mismatch). Re-enable once fixed.")
74
+ # TPU + google/gemma-3-1b-it + xet doesn't work well.
75
+ m.setenv("HF_HUB_DISABLE_XET", "1")
76
77
more_args = "max_model_len=2048,max_num_seqs=64"
78