[CI] Bump mteb version to `mteb[bm25s]>=2, <3` for pooling model unit tests (#34961)

Signed-off-by: yewentao256 <zhyanwentao@126.com>

requirements/nightly_torch_test.txt

@@ -28,7 +28,7 @@ num2words # required for smolvlm test
 opencv-python-headless >= 4.13.0 # required for video test
 datamodel_code_generator # required for minicpm3 test
 lm-eval[api]>=0.4.11 # required for model evaluation test
-mteb>=1.38.11, <2 # required for mteb test
+mteb[bm25s]>=2, <3 # required for mteb test
 transformers==4.57.5
 tokenizers==0.22.0
 schemathesis>=3.39.15 # Required for openai schema test.

requirements/rocm-test.txt

@@ -70,7 +70,7 @@ ray[cgraph,default]>=2.48.0
 torchgeo==0.7.0
     # via terratorch
 # MTEB Benchmark Test
-mteb==2.1.2
+mteb[bm25s]>=2, <3
 
 # Utilities
 num2words==0.5.14

requirements/test.txt

@@ -491,7 +491,7 @@ msgpack==1.1.0
     # via
     #   librosa
     #   ray
-mteb==2.1.2
+mteb==2.8.3
     # via -r requirements/test.in
 multidict==6.1.0
     # via

tests/models/language/pooling_mteb_test/mteb_score_utils.py

@@ -191,6 +191,9 @@ def run_mteb_rerank(cross_encoder: mteb.CrossEncoderProtocol, tasks, languages):
         mteb_tasks: list[mteb.abstasks.AbsTaskRetrieval] = mteb.get_tasks(
             tasks=tasks, languages=languages, eval_splits=eval_splits
         )
+        for task in mteb_tasks:
+            if not task.data_loaded:
+                task.load_data()
 
         mteb.evaluate(
             bm25s,