[CI] Split pooling from entrypoints Test (#24632)

Signed-off-by: wang.yuqi <noooop@126.com>

.buildkite/test-pipeline.yaml

@@ -113,7 +113,7 @@ steps:
   - tests/entrypoints/
   commands:
   - pytest -v -s entrypoints/openai/tool_parsers
-  - pytest -v -s entrypoints/ --ignore=entrypoints/llm --ignore=entrypoints/openai --ignore=entrypoints/offline_mode --ignore=entrypoints/test_chat_utils.py
+  - pytest -v -s entrypoints/ --ignore=entrypoints/llm --ignore=entrypoints/openai --ignore=entrypoints/offline_mode --ignore=entrypoints/test_chat_utils.py  --ignore=entrypoints/pooling
 
 - label: Entrypoints Integration Test (LLM) # 30min
   timeout_in_minutes: 40
@@ -148,6 +148,19 @@ steps:
   - pytest -v -s entrypoints/openai --ignore=entrypoints/openai/test_chat_with_tool_reasoning.py --ignore=entrypoints/openai/test_oot_registration.py --ignore=entrypoints/openai/test_tensorizer_entrypoint.py --ignore=entrypoints/openai/correctness/ --ignore=entrypoints/openai/test_collective_rpc.py --ignore=entrypoints/openai/tool_parsers/
   - pytest -v -s entrypoints/test_chat_utils.py
 
+- label: Entrypoints Integration Test (Pooling)
+  timeout_in_minutes: 50
+  mirror_hardwares: [amdexperimental]
+  working_dir: "/vllm-workspace/tests"
+  fast_check: true
+  torch_nightly: true
+  source_file_dependencies:
+  - vllm/
+  - tests/entrypoints/pooling
+  commands:
+  - export VLLM_WORKER_MULTIPROC_METHOD=spawn
+  - pytest -v -s entrypoints/pooling
+
 - label: Distributed Tests (4 GPUs) # 35min
   timeout_in_minutes: 50
   mirror_hardwares: [amdexperimental]

tests/entrypoints/pooling/llm/test_classify.py

@@ -6,11 +6,10 @@ import weakref
 import pytest
 import torch
 
+from tests.models.utils import softmax
 from vllm import LLM, PoolingParams
 from vllm.distributed import cleanup_dist_env_and_memory
 
-from ...models.utils import softmax
-
 MODEL_NAME = "jason9693/Qwen2.5-1.5B-apeach"
 
 prompts = ["The chef prepared a delicious meal."]

tests/entrypoints/pooling/llm/test_reward.py

@@ -6,11 +6,10 @@ import weakref
 import pytest
 import torch
 
+from tests.models.utils import softmax
 from vllm import LLM, PoolingParams
 from vllm.distributed import cleanup_dist_env_and_memory
 
-from ...models.utils import softmax
-
 MODEL_NAME = "internlm/internlm2-1_8b-reward"
 
 prompts = ["The chef prepared a delicious meal."]

tests/entrypoints/pooling/llm/test_score.py

@@ -6,11 +6,10 @@ import weakref
 import pytest
 import torch
 
+from tests.models.utils import softmax
 from vllm import LLM, PoolingParams
 from vllm.distributed import cleanup_dist_env_and_memory
 
-from ...models.utils import softmax
-
 MODEL_NAME = "tomaarsen/Qwen3-Reranker-0.6B-seq-cls"
 
 

tests/entrypoints/pooling/openai/test_classification.py

@@ -6,10 +6,9 @@ import requests
 import torch
 import torch.nn.functional as F
 
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import ClassificationResponse
 
-from ...utils import RemoteOpenAIServer
-
 MODEL_NAME = "jason9693/Qwen2.5-1.5B-apeach"
 DTYPE = "float32"  # Use float32 to avoid NaN issue
 

tests/entrypoints/pooling/openai/test_embedding.py

@@ -11,14 +11,13 @@ import requests
 import torch
 import torch.nn.functional as F
 
+from tests.models.language.pooling.embed_utils import (
+    run_embedding_correctness_test)
+from tests.models.utils import check_embeddings_close
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import EmbeddingResponse
 from vllm.transformers_utils.tokenizer import get_tokenizer
 
-from ...models.language.pooling.embed_utils import (
-    run_embedding_correctness_test)
-from ...models.utils import check_embeddings_close
-from ...utils import RemoteOpenAIServer
-
 MODEL_NAME = "intfloat/multilingual-e5-small"
 DUMMY_CHAT_TEMPLATE = """{% for message in messages %}{{message['role'] + ': ' + message['content'] + '\\n'}}{% endfor %}"""  # noqa: E501
 DTYPE = "bfloat16"

tests/entrypoints/pooling/openai/test_embedding_dimensions.py

@@ -9,13 +9,12 @@ from typing import Optional
 import openai
 import pytest
 
-from vllm.entrypoints.openai.protocol import EmbeddingResponse
-
-from ...conftest import HfRunner
-from ...models.language.pooling.embed_utils import (
+from tests.conftest import HfRunner
+from tests.models.language.pooling.embed_utils import (
     run_embedding_correctness_test)
-from ...models.utils import EmbedModelInfo
-from ...utils import RemoteOpenAIServer
+from tests.models.utils import EmbedModelInfo
+from tests.utils import RemoteOpenAIServer
+from vllm.entrypoints.openai.protocol import EmbeddingResponse
 
 MODELS = [
     EmbedModelInfo("intfloat/multilingual-e5-small", is_matryoshka=False),

tests/entrypoints/pooling/openai/test_embedding_long_text.py

@@ -14,10 +14,9 @@ import openai
 import pytest
 import pytest_asyncio
 
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import EmbeddingResponse
 
-from ...utils import RemoteOpenAIServer
-
 
 def _generate_random_text(word_count: int) -> str:
     """Generate random text with approximately the specified word count."""

tests/entrypoints/pooling/openai/test_pooling.py

@@ -8,11 +8,10 @@ import pytest
 import requests
 
 from tests.models.utils import check_embeddings_close
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import PoolingResponse
 from vllm.transformers_utils.tokenizer import get_tokenizer
 
-from ...utils import RemoteOpenAIServer
-
 MODEL_NAME = "internlm/internlm2-1_8b-reward"
 DUMMY_CHAT_TEMPLATE = """{% for message in messages %}{{message['role'] + ': ' + message['content'] + '\\n'}}{% endfor %}"""  # noqa: E501
 

tests/entrypoints/pooling/openai/test_rerank.py

@@ -6,10 +6,9 @@ import requests
 import torch
 import torch.nn.functional as F
 
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import RerankResponse
 
-from ...utils import RemoteOpenAIServer
-
 MODEL_NAME = "BAAI/bge-reranker-base"
 DTYPE = "bfloat16"
 

tests/entrypoints/pooling/openai/test_score.py

@@ -8,10 +8,9 @@ import torch
 import torch.nn.functional as F
 from torch import tensor
 
+from tests.utils import RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import ScoreResponse
 
-from ...utils import RemoteOpenAIServer
-
 MODELS = [
     {
         "name": "BAAI/bge-reranker-v2-m3",

tests/entrypoints/pooling/openai/test_vision_embedding.py

@@ -7,11 +7,10 @@ import pytest
 import requests
 from transformers import AutoProcessor
 
+from tests.utils import VLLM_PATH, RemoteOpenAIServer
 from vllm.entrypoints.openai.protocol import EmbeddingResponse
 from vllm.multimodal.utils import encode_image_base64, fetch_image
 
-from ...utils import VLLM_PATH, RemoteOpenAIServer
-
 MODEL_NAME = "TIGER-Lab/VLM2Vec-Full"
 MAXIMUM_IMAGES = 2