ci: unify PR test suite naming (#21187)

test/registered/unit/README.md

@@ -14,8 +14,8 @@ Tests can use CPU or GPU — the key criterion is **no server process**.
 3. Register for CI at the **top of the file** (after imports, before test classes):
    ```python
    from sglang.test.ci.ci_register import register_cpu_ci
-   register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
-   # or: register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+   register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
+   # or: register_cuda_ci(est_time=10, suite="stage-b-test-1-gpu-small")
    ```
 4. Run locally:
    ```bash
@@ -39,7 +39,7 @@ Tests can use CPU or GPU — the key criterion is **no server process**.
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 

test/registered/unit/entrypoints/openai/test_serving_embedding.py

@@ -23,7 +23,7 @@ from sglang.srt.entrypoints.openai.serving_embedding import OpenAIServingEmbeddi
 from sglang.srt.managers.io_struct import EmbeddingReqInput
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=10, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=10, suite="stage-a-test-cpu")
 
 
 # Mock TokenizerManager for embedding tests

test/registered/unit/entrypoints/test_ssl_cert_refresher.py

@@ -8,7 +8,7 @@ from sglang.srt.entrypoints.ssl_utils import SSLCertRefresher
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=9, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=9, suite="stage-a-test-cpu")
 
 
 def _make_temp_pem(content: bytes) -> str:

test/registered/unit/function_call/test_function_call_parser.py

@@ -18,7 +18,7 @@ from sglang.srt.function_call.pythonic_detector import PythonicDetector
 from sglang.srt.function_call.qwen3_coder_detector import Qwen3CoderDetector
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-cpu-only")
+register_cpu_ci(1.0, "stage-a-test-cpu")
 
 
 class TestPythonicDetector(unittest.TestCase):

test/registered/unit/function_call/test_glm47_moe_detector.py

@@ -10,7 +10,7 @@ from sglang.srt.function_call.glm47_moe_detector import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-cpu-only")
+register_cpu_ci(1.0, "stage-a-test-cpu")
 
 
 class TestGlm47MoeDetector(unittest.TestCase):

test/registered/unit/function_call/test_json_schema_constraint.py

@@ -18,7 +18,7 @@ from sglang.srt.function_call.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-cpu-only")
+register_cpu_ci(1.0, "stage-a-test-cpu")
 
 
 class TestJsonSchemaConstraint(unittest.TestCase):

test/registered/unit/function_call/test_parallel_tool_calls.py

@@ -23,7 +23,7 @@ from sglang.srt.entrypoints.openai.protocol import Function, Tool
 from sglang.srt.function_call.json_array_parser import JsonArrayParser
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-cpu-only")
+register_cpu_ci(1.0, "stage-a-test-cpu")
 
 
 class TestParallelToolCalls(unittest.TestCase):

test/registered/unit/function_call/test_unknown_tool_name.py

@@ -9,7 +9,7 @@ from sglang.srt.function_call.base_format_detector import BaseFormatDetector
 from sglang.srt.function_call.core_types import StreamingParseResult
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(1.0, "stage-a-cpu-only")
+register_cpu_ci(1.0, "stage-a-test-cpu")
 
 
 class DummyDetector(BaseFormatDetector):

test/registered/unit/layers/test_conv_layer.py

@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=10, suite="stage-b-test-1-gpu-small")
 
 import unittest
 

test/registered/unit/layers/test_mamba_state_scatter_triton.py

@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=10, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=10, suite="stage-b-test-1-gpu-small")
 
 import os
 import unittest

test/registered/unit/managers/test_io_struct.py

@@ -9,8 +9,8 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cuda_ci(est_time=8, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=8, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=8, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestGenerateReqInputNormalization(CustomTestCase):

test/registered/unit/managers/test_prefill_adder.py

@@ -12,8 +12,8 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=1, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=2, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=1, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=2, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestPrefillAdder(CustomTestCase):

test/registered/unit/managers/test_profile_merger_http_api.py

@@ -4,8 +4,8 @@ import unittest
 from sglang.srt.managers.io_struct import ProfileReqInput
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=9, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=9, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestProfileMergerHTTPAPI(unittest.TestCase):

test/registered/unit/mem_cache/test_evict_policy.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock

test/registered/unit/mem_cache/test_mamba_unittest.py

@@ -21,8 +21,8 @@ from sglang.srt.server_args import ServerArgs, set_global_server_args_for_schedu
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=9, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=9, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestMamba(unittest.TestCase):

test/registered/unit/mem_cache/test_nsa_pool_host_unit.py

@@ -11,7 +11,7 @@ from sglang.srt.mem_cache.memory_pool_host import (
 from sglang.srt.utils import is_cuda, is_hip, is_npu, is_xpu
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=3, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=3, suite="stage-b-test-1-gpu-small")
 
 
 class TestNSAHiCacheTransfer(unittest.TestCase):

test/registered/unit/mem_cache/test_radix_cache_slru_accuracy.py

@@ -13,7 +13,7 @@ from sglang.srt.mem_cache.memory_pool import MHATokenToKVPool, ReqToTokenPool
 from sglang.srt.mem_cache.radix_cache import RadixCache, RadixKey
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=5, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=5, suite="stage-b-test-1-gpu-small")
 
 
 class TestSLRUAccuracy(unittest.TestCase):

test/registered/unit/mem_cache/test_radix_cache_unit.py

@@ -21,8 +21,8 @@ from sglang.srt.mem_cache.common import available_and_evictable_str
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
 # CPU-based unit test, runs quickly on any GPU runner
-register_cuda_ci(est_time=5, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=5, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=5, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=5, suite="stage-b-test-1-gpu-small-amd")
 
 import random
 import time

test/registered/unit/mem_cache/test_swa_unittest.py

@@ -17,8 +17,8 @@ from sglang.srt.mem_cache.swa_radix_cache import SWARadixCache
 from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=8, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=8, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestSWA(unittest.TestCase):

test/registered/unit/model_executor/test_model_hooks.py

@@ -9,8 +9,8 @@ from sglang.srt.server_args import ServerArgs
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=6, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=10, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=6, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 HOOK_CALLS = []
 

test/registered/unit/model_loader/test_modelopt_export.py

@@ -19,8 +19,8 @@ from sglang.srt.configs.model_config import ModelConfig
 from sglang.srt.model_loader.loader import ModelOptModelLoader
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=9, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=9, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=9, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=9, suite="stage-b-test-1-gpu-small-amd")
 
 # Note: PYTHONPATH=python should be set when running tests
 

test/registered/unit/model_loader/test_modelopt_loader.py

@@ -30,7 +30,7 @@ CALIBRATION_BATCH_SIZE = 36
 CALIBRATION_NUM_SAMPLES = 512
 DEFAULT_DEVICE = "cuda:0"
 
-register_cuda_ci(est_time=11, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=11, suite="stage-b-test-1-gpu-small")
 
 
 class TestModelOptModelLoader(CustomTestCase):

test/registered/unit/observability/test_cpu_monitor.py

@@ -3,7 +3,7 @@ import unittest
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=60, suite="stage-a-cpu-only", nightly=True)
+register_cpu_ci(est_time=60, suite="stage-a-test-cpu", nightly=True)
 
 
 class TestCpuMonitor(unittest.TestCase):

test/registered/unit/observability/test_metrics_utils.py

@@ -6,7 +6,7 @@ from sglang.srt.observability.utils import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=1, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=1, suite="stage-a-test-cpu")
 
 
 class TestMetricsUtils(unittest.TestCase):

test/registered/unit/parser/test_harmony_parser.py

@@ -12,7 +12,7 @@ from sglang.srt.parser.harmony_parser import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=6, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=6, suite="stage-a-test-cpu")
 
 
 class TestEvent(CustomTestCase):

test/registered/unit/parser/test_jinja_template_utils.py

@@ -11,7 +11,7 @@ from sglang.srt.parser.jinja_template_utils import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=7, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=7, suite="stage-a-test-cpu")
 
 
 class TestTemplateContentFormatDetection(CustomTestCase):

test/registered/unit/parser/test_reasoning_parser.py

@@ -14,7 +14,7 @@ from sglang.srt.parser.reasoning_parser import (
 from sglang.test.ci.ci_register import register_cpu_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 
 class TestStreamingParseResult(CustomTestCase):

test/registered/unit/sampling/test_custom_logit_processor.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import json
 import unittest

test/registered/unit/sampling/test_penaltylib.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock

test/registered/unit/sampling/test_sampling_batch_info.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock, patch

test/registered/unit/sampling/test_sampling_params.py

@@ -2,7 +2,7 @@
 
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 from unittest.mock import MagicMock

test/registered/unit/server_args/test_server_args.py

@@ -10,7 +10,7 @@ from sglang.test.test_utils import (
     CustomTestCase,
 )
 
-register_cpu_ci(est_time=1, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=1, suite="stage-a-test-cpu")
 
 # Mock get_device() so all tests run on CPU-only CI runners
 _mock_device = patch("sglang.srt.server_args.get_device", return_value="cuda")

test/registered/unit/utils/test_gauge_histogram.py

@@ -1,6 +1,6 @@
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=5, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=5, suite="stage-a-test-cpu")
 
 import unittest
 

test/registered/unit/utils/test_json_response.py

@@ -10,7 +10,7 @@ from sglang.srt.utils.json_response import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=2, suite="stage-a-cpu-only")
+register_cpu_ci(est_time=2, suite="stage-a-test-cpu")
 
 
 class TestJSONResponseUtils(unittest.TestCase):

test/registered/unit/utils/test_patch_tokenizer.py

@@ -10,7 +10,7 @@ from sglang.srt.utils.patch_tokenizer import (
 )
 from sglang.test.ci.ci_register import register_cpu_ci
 
-register_cpu_ci(est_time=30, suite="stage-a-cpu-only", nightly=True)
+register_cpu_ci(est_time=30, suite="stage-a-test-cpu", nightly=True)
 
 
 class TestPatchTokenizerEndToEndTest(unittest.TestCase):

test/registered/unit/utils/test_profile_merger.py

@@ -17,8 +17,8 @@ from sglang.srt.managers.io_struct import ProfileReq, ProfileReqInput, ProfileRe
 from sglang.srt.utils.profile_merger import ProfileMerger
 from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 
-register_cuda_ci(est_time=8, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=8, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=8, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=8, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestProfileMerger(unittest.TestCase):