ci(test): migrate OpenAI server tests to registered CI system (#16326)

Co-authored-by: Kangyan-Zhou <zky314343421@gmail.com>

python/sglang/srt/distributed/device_communicators/pynccl_allocator.py

@@ -3,7 +3,6 @@ import tempfile
 from contextlib import nullcontext
 
 import torch
-import torch.utils.cpp_extension
 from packaging import version
 from torch.cuda.memory import CUDAPluggableAllocator
 
@@ -90,6 +89,8 @@ def restore_symmetric_memory_context(saved_context):
 def get_nccl_mem_pool():
     global _allocator, _mem_pool, _cur_device
     if _mem_pool is None:
+        import torch.utils.cpp_extension
+
         out_dir = tempfile.gettempdir()
         nccl_allocator_libname = "nccl_allocator"
         torch.utils.cpp_extension.load_inline(

test/registered/openai_server/basic/test_openai_embedding.py

@@ -4,6 +4,7 @@ import unittest
 import openai
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_EMBEDDING_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=70, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=141, suite="stage-b-test-small-1-gpu")
+
 
 class TestOpenAIEmbedding(CustomTestCase):
     @classmethod
@@ -69,7 +73,6 @@ class TestOpenAIEmbedding(CustomTestCase):
         self.assertEqual(len(response.data), 1)
         self.assertTrue(len(response.data[0].embedding) > 0)
 
-        client = openai.Client(api_key=self.api_key, base_url=self.base_url)
         response = client.embeddings.create(
             model=self.model,
             input=[15339, 314, 703, 284, 612, 262, 10658, 10188, 286, 2061],

test/registered/openai_server/basic/test_openai_server.py

@@ -17,6 +17,7 @@ import requests
 from sglang.srt.sampling.custom_logit_processor import CustomLogitProcessor
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.runners import TEST_RERANK_QUERY_DOCS
 from sglang.test.test_utils import (
     DEFAULT_SMALL_CROSS_ENCODER_MODEL_NAME_FOR_TEST,
@@ -27,6 +28,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=184, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=149, suite="stage-b-test-small-1-gpu")
+
 
 class TestOpenAIServer(CustomTestCase):
     @classmethod

test/registered/openai_server/basic/test_protocol.py

@@ -28,6 +28,10 @@ from sglang.srt.entrypoints.openai.protocol import (
     ModelList,
     UsageInfo,
 )
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=3, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu")
 
 
 class TestModelCard(unittest.TestCase):

test/registered/openai_server/basic/test_serving_chat.py

@@ -21,6 +21,10 @@ from sglang.srt.entrypoints.openai.protocol import (
 from sglang.srt.entrypoints.openai.serving_chat import OpenAIServingChat
 from sglang.srt.managers.io_struct import GenerateReqInput
 from sglang.srt.utils import get_or_create_event_loop
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu")
 
 
 class _MockTokenizerManager:

test/registered/openai_server/basic/test_serving_completions.py

@@ -11,6 +11,10 @@ from unittest.mock import AsyncMock, Mock
 from sglang.srt.entrypoints.openai.protocol import CompletionRequest
 from sglang.srt.entrypoints.openai.serving_completions import OpenAIServingCompletion
 from sglang.srt.managers.tokenizer_manager import TokenizerManager
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu")
 
 
 class _MockTemplateManager:

test/registered/openai_server/basic/test_serving_embedding.py

@@ -14,6 +14,10 @@ from sglang.srt.entrypoints.openai.protocol import (
 )
 from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbedding
 from sglang.srt.managers.io_struct import EmbeddingReqInput
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
+
+register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu")
 
 
 # Mock TokenizerManager for embedding tests

test/registered/openai_server/features/test_enable_thinking.py

@@ -12,6 +12,7 @@ import unittest
 import requests
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_ENABLE_THINKING_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -20,6 +21,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=70, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=70, suite="stage-b-test-small-1-gpu")
+
 
 class TestEnableThinking(CustomTestCase):
     @classmethod

test/registered/openai_server/features/test_json_mode.py

@@ -4,6 +4,7 @@ import unittest
 import openai
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=109, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=120, suite="stage-b-test-small-1-gpu")
+
 
 class TestJSONModeMixin:
     """Mixin class containing JSON mode test methods"""

test/registered/openai_server/features/test_openai_server_ebnf.py

@@ -4,6 +4,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -12,6 +13,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=7, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=20, suite="stage-b-test-small-1-gpu")
+
 
 # -------------------------------------------------------------------------
 #    EBNF Test Class: TestOpenAIServerEBNF
@@ -96,3 +100,9 @@ class TestOpenAIServerEBNF(CustomTestCase):
         self.assertRegex(
             text, pattern, f"Text '{text}' not matching the EBNF strict JSON shape"
         )
+
+
+if __name__ == "__main__":
+    import unittest
+
+    unittest.main()

test/registered/openai_server/features/test_openai_server_hidden_states.py

@@ -5,6 +5,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_DRAFT_MODEL_EAGLE,
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
@@ -15,6 +16,13 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=186, suite="stage-b-test-small-1-gpu")
+register_amd_ci(
+    est_time=186,
+    suite="stage-b-test-small-1-gpu",
+    disabled="see https://github.com/sgl-project/sglang/issues/11127",
+)
+
 
 class BaseTestOpenAIServerWithHiddenStates(ABC):
 

test/registered/openai_server/features/test_reasoning_content.py

@@ -14,6 +14,7 @@ import unittest
 import openai
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_REASONING_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -22,6 +23,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=89, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=89, suite="stage-b-test-small-1-gpu")
+
 
 class TestReasoningContentAPI(CustomTestCase):
     @classmethod

test/registered/openai_server/function_call/test_openai_function_calling.py

@@ -5,6 +5,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -13,6 +14,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=60, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=73, suite="stage-b-test-small-1-gpu")
+
 
 class TestOpenAIServerFunctionCalling(CustomTestCase):
     # NOTE: this system_message is for Llama3.2 system prompt. Without this,

test/registered/openai_server/function_call/test_tool_choice.py

@@ -14,6 +14,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
@@ -21,6 +22,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=120, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=258, suite="stage-b-test-small-1-gpu")
+
 
 class TestToolChoiceLlama32(CustomTestCase):
 

test/registered/openai_server/validation/test_large_max_new_tokens.py

@@ -11,6 +11,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -21,6 +22,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=41, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=41, suite="stage-b-test-small-1-gpu")
+
 
 class TestLargeMaxNewTokens(CustomTestCase):
     @classmethod

test/registered/openai_server/validation/test_matched_stop.py

@@ -2,6 +2,7 @@ import unittest
 
 from sglang.srt.sampling.sampling_params import MAX_LEN, get_max_seq_length
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.kits.matched_stop_kit import MatchedStopMixin
 from sglang.test.test_utils import (
     DEFAULT_MODEL_NAME_FOR_TEST,
@@ -10,6 +11,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=40, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=60, suite="stage-b-test-small-1-gpu")
+
 
 class TestMatchedStop(CustomTestCase, MatchedStopMixin):
     @classmethod

test/registered/openai_server/validation/test_openai_server_ignore_eos.py

@@ -2,6 +2,7 @@ import openai
 
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.hf_transformers_utils import get_tokenizer
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -10,6 +11,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=47, suite="stage-b-test-small-1-gpu")
+
 
 class TestOpenAIServerIgnoreEOS(CustomTestCase):
     @classmethod
@@ -82,3 +86,9 @@ class TestOpenAIServerIgnoreEOS(CustomTestCase):
             "length",
             f"Expected finish_reason='length' for ignore_eos=True, got {response_ignore_eos.choices[0].finish_reason}",
         )
+
+
+if __name__ == "__main__":
+    import unittest
+
+    unittest.main()

test/registered/openai_server/validation/test_request_length_validation.py

@@ -3,6 +3,7 @@ import unittest
 import openai
 
 from sglang.srt.utils import kill_process_tree
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -11,6 +12,9 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
+register_cuda_ci(est_time=38, suite="stage-b-test-small-1-gpu")
+register_amd_ci(est_time=31, suite="stage-b-test-small-1-gpu")
+
 
 class TestRequestLengthValidation(CustomTestCase):
     @classmethod

test/srt/run_suite.py

@@ -10,23 +10,6 @@ from sglang.test.ci.ci_utils import TestFile, run_unittest_files
 suites = {
     "per-commit-1-gpu": [
         TestFile("debug_utils/test_tensor_dump_forward_hook.py", 9),
-        TestFile("openai_server/basic/test_openai_embedding.py", 70),
-        TestFile("openai_server/basic/test_openai_server.py", 184),
-        TestFile("openai_server/basic/test_protocol.py", 3),
-        TestFile("openai_server/basic/test_serving_chat.py", 10),
-        TestFile("openai_server/basic/test_serving_completions.py", 10),
-        TestFile("openai_server/basic/test_serving_embedding.py", 10),
-        TestFile("openai_server/features/test_enable_thinking.py", 70),
-        TestFile("openai_server/features/test_json_mode.py", 109),
-        TestFile("openai_server/features/test_openai_server_ebnf.py", 7),
-        TestFile("openai_server/features/test_openai_server_hidden_states.py", 186),
-        TestFile("openai_server/features/test_reasoning_content.py", 89),
-        TestFile("openai_server/function_call/test_openai_function_calling.py", 60),
-        TestFile("openai_server/function_call/test_tool_choice.py", 120),
-        TestFile("openai_server/validation/test_large_max_new_tokens.py", 41),
-        TestFile("openai_server/validation/test_matched_stop.py", 40),
-        TestFile("openai_server/validation/test_openai_server_ignore_eos.py", 6),
-        TestFile("openai_server/validation/test_request_length_validation.py", 38),
         TestFile("ops/test_repeat_interleave.py", 60),
         TestFile("test_abort.py", 131),
         TestFile("test_chunked_prefill.py", 312),
@@ -155,22 +138,6 @@ suite_amd = {
         # TestFile("lora/test_lora_backend.py", 99), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/13107
         # TestFile("lora/test_lora_cuda_graph.py", 250), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/13107
         # TestFile("lora/test_lora_qwen3.py", 97), # Disabled temporarily, see https://github.com/sgl-project/sglang/issues/13107
-        TestFile("openai_server/basic/test_openai_embedding.py", 141),
-        TestFile("openai_server/basic/test_openai_server.py", 149),
-        TestFile("openai_server/basic/test_protocol.py", 10),
-        TestFile("openai_server/basic/test_serving_chat.py", 10),
-        TestFile("openai_server/basic/test_serving_completions.py", 10),
-        TestFile("openai_server/basic/test_serving_embedding.py", 10),
-        TestFile("openai_server/features/test_enable_thinking.py", 70),
-        TestFile("openai_server/features/test_json_mode.py", 120),
-        TestFile("openai_server/features/test_openai_server_ebnf.py", 20),
-        TestFile("openai_server/features/test_reasoning_content.py", 89),
-        TestFile("openai_server/function_call/test_openai_function_calling.py", 73),
-        TestFile("openai_server/function_call/test_tool_choice.py", 120),
-        TestFile("openai_server/validation/test_large_max_new_tokens.py", 41),
-        TestFile("openai_server/validation/test_matched_stop.py", 60),
-        TestFile("openai_server/validation/test_openai_server_ignore_eos.py", 85),
-        TestFile("openai_server/validation/test_request_length_validation.py", 31),
         TestFile("ops/test_repeat_interleave.py", 75),
         TestFile("rotary_embedding/test_mrope.py", 15),
         TestFile("test_abort.py", 51),