ci: unify PR test suite naming (#21187)

test/registered/attention/test_chunk_gated_delta_rule.py

@@ -8,7 +8,7 @@ from sglang.srt.layers.attention.fla.fused_recurrent import (
 )
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=60, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=60, suite="stage-b-test-1-gpu-large")
 
 
 @unittest.skipIf(not torch.cuda.is_available(), "Test requires CUDA")

test/registered/attention/test_create_kvindices.py

@@ -9,8 +9,8 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 # Triton kernel unit test for KV indices creation
-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-amd")
+register_cuda_ci(est_time=10, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=10, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestCreateKvIndices(CustomTestCase):

test/registered/attention/test_fa3.py

@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
 
 # FlashAttention3 integration tests (requires SM 90+ / H100)
 # Multiple test classes: FA3, FA3+MLA, FA3+SpecDecode variants
-register_cuda_ci(est_time=300, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=300, suite="stage-b-test-1-gpu-large")
 
 GSM_DATASET_PATH = None
 

test/registered/attention/test_hybrid_attn_backend.py

@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
 
 # Hybrid attention backend tests (FA3 prefill + FlashInfer decode, requires SM 90+ / H100)
 # Multiple test classes: base, MLA, TorchCompile, SpecDecode variants
-register_cuda_ci(est_time=200, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=200, suite="stage-b-test-1-gpu-large")
 
 GSM_DATASET_PATH = None
 

test/registered/attention/test_kda_kernels.py

@@ -8,7 +8,7 @@ from sglang.srt.layers.attention.fla.fused_sigmoid_gating_recurrent import (
 from sglang.srt.layers.attention.fla.kda import fused_kda_gate, fused_recurrent_kda
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=30, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=30, suite="stage-b-test-1-gpu-large")
 
 
 @unittest.skipIf(not torch.cuda.is_available(), "Test requires CUDA")

test/registered/attention/test_normal_decode_set_metadata.py

@@ -21,7 +21,7 @@ from sglang.test.ci.ci_register import register_cuda_ci
 from sglang.test.test_utils import CustomTestCase
 
 # Register this test for CUDA CI in stage-b (fast attention/kernel tests)
-register_cuda_ci(est_time=25, suite="stage-b-test-large-1-gpu")
+register_cuda_ci(est_time=25, suite="stage-b-test-1-gpu-large")
 
 
 def reference_normal_decode_set_metadata(

test/registered/attention/test_torch_native_attention_backend.py

@@ -18,8 +18,8 @@ from sglang.test.test_utils import (
 )
 
 # Torch native attention backend integration test with MMLU eval
-register_cuda_ci(est_time=169, suite="stage-b-test-small-1-gpu")
-register_amd_ci(est_time=150, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=169, suite="stage-b-test-1-gpu-small")
+register_amd_ci(est_time=150, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestTorchNativeAttnBackend(CustomTestCase):

test/registered/attention/test_triton_attention_backend.py

@@ -20,8 +20,8 @@ from sglang.test.test_utils import (
 )
 
 # Triton attention backend integration test with latency benchmark and MMLU eval
-register_cuda_ci(est_time=200, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=1400, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=200, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=1400, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestTritonAttnBackend(CustomTestCase):

test/registered/attention/test_triton_attention_kernels.py

@@ -23,8 +23,8 @@ from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import CustomTestCase, is_in_amd_ci
 
 # Triton attention kernel unit tests (decode, extend, prefill)
-register_cuda_ci(est_time=30, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=30, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=30, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=30, suite="stage-b-test-1-gpu-small-amd")
 
 
 def extend_attention_fwd_torch(

test/registered/attention/test_triton_sliding_window.py

@@ -16,8 +16,8 @@ from sglang.test.test_utils import (
 )
 
 # Sliding window attention with Triton backend (Gemma-3 model)
-register_cuda_ci(est_time=100, suite="stage-b-test-large-1-gpu")
-register_amd_ci(est_time=200, suite="stage-b-test-small-1-gpu-amd")
+register_cuda_ci(est_time=100, suite="stage-b-test-1-gpu-large")
+register_amd_ci(est_time=200, suite="stage-b-test-1-gpu-small-amd")
 
 
 class TestSlidingWindowAttentionTriton(CustomTestCase):

test/registered/attention/test_wave_attention_kernels.py

@@ -25,7 +25,7 @@ from sglang.srt.utils import get_device
 from sglang.test.ci.ci_register import register_amd_ci
 
 # Wave attention kernel unit tests (AMD only - requires wave_lang)
-register_amd_ci(est_time=60, suite="stage-a-test-small-1-gpu-amd")
+register_amd_ci(est_time=60, suite="stage-a-test-1-gpu-small-amd")
 
 
 class TestWaveAttention(unittest.TestCase):