[AMD] Register 3 CPU-bound / triton unit + light-integration tests for AMD 1-GPU PR CI (#31088)

test/registered/attention/test_chunk_gated_delta_rule.py

@@ -6,10 +6,15 @@ from sglang.kernels.ops.attention.fla.chunk import chunk_gated_delta_rule
 from sglang.kernels.ops.attention.fla.fused_recurrent import (
     fused_recurrent_gated_delta_rule,
 )
-from sglang.srt.utils import get_device
-from sglang.test.ci.ci_register import register_cuda_ci, register_xpu_ci
+from sglang.srt.utils import get_device, is_hip
+from sglang.test.ci.ci_register import (
+    register_amd_ci,
+    register_cuda_ci,
+    register_xpu_ci,
+)
 
 register_cuda_ci(est_time=11, stage="base-b", runner_config="1-gpu-large")
+register_amd_ci(est_time=11, stage="stage-b", runner_config="1-gpu-large-amd")
 register_xpu_ci(est_time=900, suite="stage-b-test-1-gpu-xpu")
 
 
@@ -183,6 +188,11 @@ class TestChunkGatedDeltaRule(unittest.TestCase):
     def test_dim_k_gt_v(self):
         self._check_shape(B=4, T_per_seq=128, H=16, K=128, V=64, pool_size=32)
 
+    @unittest.skipIf(
+        is_hip(),
+        "K=V=256 head dim exceeds the FLA chunk triton kernel's shared-memory "
+        "budget on ROCm (out-of-resource at launch); smaller head dims pass.",
+    )
     def test_dim_256x256(self):
         self._check_shape(B=4, T_per_seq=128, H=16, K=256, V=256, pool_size=32)
 

test/registered/observability/test_encoder_server_metrics.py

@@ -13,7 +13,7 @@ from prometheus_client.samples import Sample
 from sglang.srt.disaggregation.encode_server import MINIMUM_PNG_PICTURE_BASE64
 from sglang.srt.utils import kill_process_tree
 from sglang.srt.utils.network import get_zmq_socket_on_host
-from sglang.test.ci.ci_register import register_cuda_ci
+from sglang.test.ci.ci_register import register_amd_ci, register_cuda_ci
 from sglang.test.test_utils import (
     DEFAULT_SMALL_VLM_MODEL_NAME_FOR_TEST,
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
@@ -23,6 +23,7 @@ from sglang.test.test_utils import (
 )
 
 register_cuda_ci(est_time=180, stage="base-b", runner_config="1-gpu-small")
+register_amd_ci(est_time=180, stage="stage-b", runner_config="1-gpu-small-amd")
 
 _MODEL_NAME = DEFAULT_SMALL_VLM_MODEL_NAME_FOR_TEST
 

test/registered/unit/spec/test_eagle_worker_v2_topk1_fastpath.py

@@ -16,10 +16,15 @@ from sglang.srt.model_executor.forward_batch_info import ForwardMode
 from sglang.srt.speculative.adaptive_runtime_state import SpecRuntimeState
 from sglang.srt.speculative.eagle_utils import organize_draft_results
 from sglang.srt.speculative.eagle_worker_v2 import EagleDraftWorker, EAGLEWorkerV2
-from sglang.test.ci.ci_register import register_cpu_ci, register_cuda_ci
+from sglang.test.ci.ci_register import (
+    register_amd_ci,
+    register_cpu_ci,
+    register_cuda_ci,
+)
 from sglang.test.test_utils import CustomTestCase
 
 register_cuda_ci(est_time=20, stage="base-b", runner_config="1-gpu-small")
+register_amd_ci(est_time=20, stage="stage-b", runner_config="1-gpu-small-amd")
 register_cpu_ci(est_time=20, suite="base-a-test-cpu")
 
 DEVICE = "cuda" if torch.cuda.is_available() else "cpu"