Sync from v0.13

2026-01-19 10:38:50 +08:00
parent b2ef04d792
commit 5aef6c175a
3714 changed files with 854317 additions and 89342 deletions
--- a/tests/models/language/pooling_mteb_test/test_baai.py
+++ b/tests/models/language/pooling_mteb_test/test_baai.py
@@ -0,0 +1,114 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright contributors to the vLLM project
+import pytest
+
+from tests.models.language.pooling.embed_utils import correctness_test_embed_models
+from tests.models.utils import (
+    CLSPoolingEmbedModelInfo,
+    CLSPoolingRerankModelInfo,
+    EmbedModelInfo,
+    LASTPoolingEmbedModelInfo,
+    RerankModelInfo,
+)
+
+from .mteb_utils import mteb_test_embed_models, mteb_test_rerank_models
+
+MODELS = [
+    ########## BertModel
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-base-en",
+        architecture="BertModel",
+        mteb_score=0.779336792,
+        enable_test=True,
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-base-zh", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-small-en", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-small-zh", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-large-en", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-large-zh", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-large-zh-noinstruct", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-base-en-v1.5", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-base-zh-v1.5", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-small-en-v1.5", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-small-zh-v1.5", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-large-en-v1.5", architecture="BertModel", enable_test=False
+    ),
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-large-zh-v1.5", architecture="BertModel", enable_test=False
+    ),
+    ########## XLMRobertaModel
+    CLSPoolingEmbedModelInfo(
+        "BAAI/bge-m3",
+        architecture="XLMRobertaModel",
+        mteb_score=0.787343078,
+        enable_test=True,
+    ),
+    ########## Qwen2Model
+    LASTPoolingEmbedModelInfo(
+        "BAAI/bge-code-v1",
+        architecture="Qwen2Model",
+        mteb_score=0.75724465,
+        dtype="float32",
+        enable_test=True,
+    ),
+]
+
+RERANK_MODELS = [
+    ########## XLMRobertaForSequenceClassification
+    CLSPoolingRerankModelInfo(
+        "BAAI/bge-reranker-base",
+        architecture="XLMRobertaForSequenceClassification",
+        mteb_score=0.32398,
+        enable_test=True,
+    ),
+    CLSPoolingRerankModelInfo(
+        "BAAI/bge-reranker-large",
+        architecture="XLMRobertaForSequenceClassification",
+        enable_test=False,
+    ),
+    CLSPoolingRerankModelInfo(
+        "BAAI/bge-reranker-v2-m3",
+        architecture="XLMRobertaForSequenceClassification",
+        enable_test=False,
+    ),
+]
+
+
+@pytest.mark.parametrize("model_info", MODELS)
+def test_embed_models_mteb(hf_runner, vllm_runner, model_info: EmbedModelInfo) -> None:
+    mteb_test_embed_models(hf_runner, vllm_runner, model_info)
+
+
+@pytest.mark.parametrize("model_info", MODELS)
+def test_embed_models_correctness(
+    hf_runner, vllm_runner, model_info: EmbedModelInfo, example_prompts
+) -> None:
+    correctness_test_embed_models(hf_runner, vllm_runner, model_info, example_prompts)
+
+
+@pytest.mark.parametrize("model_info", RERANK_MODELS)
+def test_rerank_models_mteb(
+    hf_runner, vllm_runner, model_info: RerankModelInfo
+) -> None:
+    mteb_test_rerank_models(hf_runner, vllm_runner, model_info)