ci: unify PR test suite naming (#21187)

test/registered/spec/eagle/test_eagle3_basic.py

@@ -12,8 +12,8 @@ from sglang.test.test_utils import (
     DEFAULT_TARGET_MODEL_EAGLE3,
 )
 
-register_cuda_ci(est_time=50, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=50, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=50, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=50, suite="stage-b-test-1-gpu-small")
 
 _is_hip = is_hip()
 

test/registered/spec/eagle/test_eagle_constrained_decoding.py

@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=100, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=100, suite="stage-b-test-1-gpu-large")
 
 
 class TestEagleConstrainedDecoding(

test/registered/spec/eagle/test_eagle_infer_a.py

@@ -22,7 +22,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=561, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=561, suite="stage-b-test-1-gpu-large")
 
 torch_dtype = torch.float16
 prefill_tolerance = 5e-2

test/registered/spec/eagle/test_eagle_infer_b.py

@@ -22,7 +22,7 @@ from sglang.test.kits.radix_cache_server_kit import run_radix_attention_test
 from sglang.test.server_fixtures.eagle_fixture import EagleServerBase
 from sglang.test.test_utils import DEFAULT_TARGET_MODEL_EAGLE, run_logprob_check
 
-register_cuda_ci(est_time=1100, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=1100, suite="stage-b-test-1-gpu-large")
 
 
 class TestEAGLEServerBasic(EagleServerBase):

test/registered/spec/eagle/test_eagle_infer_beta.py

@@ -19,7 +19,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=283, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=283, suite="stage-b-test-1-gpu-small")
 
 
 class TestEagleServerBase(CustomTestCase, MatchedStopMixin):

test/registered/spec/test_constrained_decoding_spec_reasoning.py

@@ -14,7 +14,7 @@ from sglang.test.test_utils import (
 )
 
 # Constrained decoding with EAGLE3 speculative reasoning (tp=2)
-register_cuda_ci(est_time=60, suite="stage-b-test-large-2-gpu")
+register_cuda_ci(est_time=60, suite="stage-b-test-2-gpu-large")
 
 
 class ServerWithGrammar(CustomTestCase):

test/registered/spec/test_ngram_speculative_decoding.py

@@ -12,7 +12,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=230, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=230, suite="stage-b-test-1-gpu-large")
 
 GSM_DATASET_PATH = None
 

test/registered/spec/test_standalone_speculative_decoding.py

@@ -18,7 +18,7 @@ from sglang.test.test_utils import (
 )
 
 # Standalone speculative decoding tests (FA3, Triton, FlashInfer backends)
-register_cuda_ci(est_time=308, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=308, suite="stage-b-test-1-gpu-large")
 
 GSM_DATASET_PATH = None
 

test/registered/spec/utils/test_build_eagle_tree.py

@@ -9,8 +9,8 @@ from sglang.srt.speculative.eagle_utils import (
 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=3, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=3, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=3, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=3, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestBuildEagleTree(unittest.TestCase):

test/registered/spec/utils/test_ngram_corpus.py

@@ -6,7 +6,7 @@ from sglang.srt.speculative.cpp_ngram.ngram_corpus import NgramCorpus
 from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
-register_cuda_ci(est_time=30, suite="stage-b-test-small-1-gpu")
+register_cuda_ci(est_time=30, suite="stage-b-test-1-gpu-small")
 
 
 def _make_corpus(match_type="BFS", **kwargs):