biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Bugfix] Fix ModernBert cuda graph capturing in v1 (#21901)

Browse Source 
Signed-off-by: Isotr0py <mozf@mail2.sysu.edu.cn>
Signed-off-by: Isotr0py <2037008807@qq.com>
This commit is contained in:
Isotr0py
2025-08-09 13:17:22 +08:00
committed by GitHub
parent 35afe1b30b
commit 429e4e2d42
5 changed files with 39 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
tests/models/language/pooling/mteb_utils.py
View File

@@ -162,7 +162,8 @@ def mteb_test_embed_models(hf_runner,
vllm_runner,
model_info: EmbedModelInfo,
vllm_extra_kwargs=None,
hf_model_callback=None):
hf_model_callback=None,
atol=MTEB_RERANK_TOL):
if not model_info.enable_test:
# A model family has many models with the same architecture,
# and we don't need to test each one.
@@ -198,7 +199,7 @@ def mteb_test_embed_models(hf_runner,
print("SentenceTransformers:", st_dtype, st_main_score)
print("Difference:", st_main_score - vllm_main_score)
assert st_main_score == pytest.approx(vllm_main_score, abs=MTEB_EMBED_TOL)
assert st_main_score == pytest.approx(vllm_main_score, abs=atol)
def run_mteb_rerank(cross_encoder, tasks, languages):