Merge stage-c-test-large-4-gpu suites into partitioned suites (#18325)

test/README.md

@@ -100,7 +100,8 @@ If a test cannot run on 5090 due to any of the above, use `stage-b-test-large-1-
 **Per-Commit (CUDA)**:
 - Stage A: `stage-a-test-1` (locked), `stage-a-test-2`, `stage-a-test-cpu`
 - Stage B: `stage-b-test-small-1-gpu` (5090), `stage-b-test-large-1-gpu` (H100), `stage-b-test-large-2-gpu`
-- Stage C: `stage-c-test-large-4-gpu`, `stage-c-test-large-4-gpu-b200`, `stage-c-test-large-8-gpu-b200`
+- Stage C (4-GPU): `stage-c-test-4-gpu-h100`, `stage-c-test-4-gpu-b200`, `stage-c-test-4-gpu-gb200`, `stage-c-test-deepep-4-gpu`
+- Stage C (8-GPU): `stage-c-test-8-gpu-h20`, `stage-c-test-8-gpu-h200`, `stage-c-test-8-gpu-b200`, `stage-c-test-deepep-8-gpu-h200`
 
 **Per-Commit (AMD)**:
 - `stage-a-test-1`, `stage-b-test-small-1-gpu-amd`, `stage-b-test-large-2-gpu-amd`

test/registered/attention/test_local_attn.py

@@ -16,7 +16,7 @@ from sglang.test.test_utils import (
 )
 
 # Local attention with FA3 (requires SM 90+ / H100, tp=4)
-register_cuda_ci(est_time=200, suite="stage-c-test-large-4-gpu")
+register_cuda_ci(est_time=200, suite="stage-c-test-4-gpu-h100")
 
 
 @unittest.skipIf(get_device_sm() < 90, "Test requires CUDA SM 90 or higher")

test/registered/distributed/test_dp_attention_large.py

@@ -23,7 +23,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=350, suite="stage-c-test-large-4-gpu")
+register_cuda_ci(est_time=350, suite="stage-c-test-4-gpu-h100")
 
 
 class TestDPAttentionDP2TP4(

test/registered/moe/test_cutedsl_moe.py

@@ -12,7 +12,7 @@ from sglang.srt.layers.moe.flashinfer_cutedsl_moe import flashinfer_cutedsl_moe_
 from sglang.srt.layers.moe.topk import TopKConfig, select_experts
 from sglang.test.ci.ci_register import register_cuda_ci
 
-register_cuda_ci(est_time=300, suite="stage-c-test-large-4-gpu-b200")
+register_cuda_ci(est_time=300, suite="stage-c-test-4-gpu-b200")
 
 SKIP_TEST = torch.cuda.get_device_capability() < (10, 0)
 SKIP_REASON = "Nvfp4 Requires compute capability of 10 or above."

test/registered/rl/test_return_routed_experts.py

@@ -21,7 +21,7 @@ from sglang.test.test_utils import (
     popen_launch_server,
 )
 
-register_cuda_ci(est_time=360, suite="stage-c-test-large-4-gpu")
+register_cuda_ci(est_time=360, suite="stage-c-test-4-gpu-h100")
 
 SHAREGPT_URL = (
     "https://huggingface.co/datasets/anon8231489123/"

test/registered/spec/eagle/test_eagle_dp_attention.py

@@ -20,7 +20,7 @@ from sglang.test.test_utils import (
 )
 
 # EAGLE3 with DP attention (tp=2, dp=2, requires 4 GPUs)
-register_cuda_ci(est_time=200, suite="stage-c-test-large-4-gpu")
+register_cuda_ci(est_time=200, suite="stage-c-test-4-gpu-h100")
 
 
 class TestEAGLE3EngineDPAttention(CustomTestCase):

test/registered/spec/eagle/test_eagle_infer_beta_dp_attention.py

@@ -17,7 +17,7 @@ from sglang.test.test_utils import (
 )
 
 # EAGLE with DP attention on B200 (tp=2, dp=2, requires 4 B200 GPUs)
-register_cuda_ci(est_time=300, suite="stage-c-test-large-4-gpu-b200")
+register_cuda_ci(est_time=300, suite="stage-c-test-4-gpu-b200")
 
 
 def test_gsm8k(base_url: str):

test/run_suite.py

@@ -32,7 +32,6 @@ PER_COMMIT_SUITES = {
         "stage-b-test-small-1-gpu",
         "stage-b-test-large-1-gpu",
         "stage-b-test-large-2-gpu",
-        "stage-c-test-large-4-gpu",
         "stage-c-test-4-gpu-h100",
         "stage-c-test-4-gpu-b200",
         "stage-c-test-4-gpu-gb200",