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 10, 16:30 UTC
[Bugfix][CI/Build] Fix failing pooling models test due to Triton kernel accuracy diff (#31776)
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Isotr0py committed
Jan 6, 2026 at 16:44 UTC
ee2e69d6cda898736cc2987cb8cdfaa2601c1375
1 file changed
+1
-1
tests/models/language/pooling/test_token_classification.py
+1
-1
@@ -78,7 +78,7 @@ def test_modernbert_models(
78
for hf_output, vllm_output in zip(hf_outputs, vllm_outputs):
79
hf_output = hf_output.detach().clone().cpu().float()
80
vllm_output = vllm_output.detach().clone().cpu().float()
81
- assert torch.allclose(hf_output, vllm_output, atol=1e-2)
81
+ torch.testing.assert_close(hf_output, vllm_output, atol=1.2e-2, rtol=1e-3)
82
83
84
@pytest.mark.parametrize("model", ["bd2lcco/Qwen3-0.6B-finetuned"])