[AMD] CI - add 2 pp test cases to performance-test-2-gpu-amd (#16514)

.github/workflows/pr-test-amd.yml

@@ -794,6 +794,16 @@ jobs:
         run: |
           bash scripts/ci/amd_ci_exec.sh python3 -m unittest test_bench_serving.TestBenchServing.test_moe_offline_throughput_without_radix_cache
 
+      - name: Benchmark offline PP decode throughput (PP=2)
+        timeout-minutes: 10
+        run: |
+          bash scripts/ci/amd_ci_exec.sh python3 -m unittest test_bench_serving.TestBenchServing.test_pp_offline_throughput_default_decode
+
+      - name: Benchmark offline PP prefill throughput (PP=2)
+        timeout-minutes: 10
+        run: |
+          bash scripts/ci/amd_ci_exec.sh python3 -m unittest test_bench_serving.TestBenchServing.test_pp_long_context_prefill
+
   accuracy-test-1-gpu-amd:
     needs: [check-changes, stage-a-test-1-amd]
     if: |

test/srt/test_bench_serving.py

@@ -432,7 +432,8 @@ class TestBenchServing(CustomTestCase):
                 "fp8",
                 "--pp-size",
                 "2",
-            ],
+            ]
+            + (["--mem-fraction-static", "0.7"] if is_in_amd_ci() else []),
             need_warmup=False,
             seed=42,
         )
@@ -442,7 +443,10 @@ class TestBenchServing(CustomTestCase):
                 f"### test_pp_long_context_latency_prefill\n"
                 f"input_throughput: {res['input_throughput']:.2f} ms\n"
             )
-            self.assertGreater(res["input_throughput"], 4000)
+            if is_in_amd_ci():
+                self.assertGreater(res["input_throughput"], 3000)
+            else:
+                self.assertGreater(res["input_throughput"], 4000)
 
     def test_score_api_latency_throughput(self):
         """Test score API latency and throughput performance"""