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 6, 21:45 UTC
[ci] Update rtol for test_classification (#36556)
Signed-off-by: angelayi <yiangela7@gmail.com> Co-authored-by: Richard Zou <zou3519@users.noreply.github.com>
Angela Yi committed
Mar 11, 2026 at 03:08 UTC
13e79fc8111b9eb3a2a5a367ea08f5d7fbf57281
1 file changed
+3
-1
tests/models/language/pooling/test_classification.py
+3
-1
@@ -45,5 +45,7 @@ def test_models(
45
# half datatype tests in
46
# tests/models/language/pooling/test_embedding.py
47
assert torch.allclose(
48
- hf_output, vllm_output, 1e-3 if dtype == "float" else 1e-2
48
+ hf_output,
49
+ vllm_output,
50
+ rtol=2e-3 if dtype == "float" else 1e-2,
51
)