[AMD] CI - enable deepseekv3.2 on MI325-8gpu and merge perf/accuracy test suites into stage-b suites (#17633)

Co-authored-by: Bingxu Chen <Bingxu.Chen@amd.com>

test/registered/attention/test_triton_sliding_window.py

@@ -10,6 +10,7 @@ from sglang.test.test_utils import (
     DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
     DEFAULT_URL_FOR_TEST,
     CustomTestCase,
+    is_in_amd_ci,
     is_in_ci,
     popen_launch_server,
 )
@@ -62,7 +63,10 @@ class TestSlidingWindowAttentionTriton(CustomTestCase):
         metrics = run_eval(args)
         print(f"MMLU metrics with sliding window: {metrics}")
 
-        self.assertGreaterEqual(metrics["score"], 0.60)
+        if is_in_amd_ci():
+            self.assertGreaterEqual(metrics["score"], 0.55)
+        else:
+            self.assertGreaterEqual(metrics["score"], 0.60)
 
     def _test_short_context_generation(self):
         response = requests.post(