[AMD] Support Wave attention backend with AMD GPU optimizations (#8660)

Signed-off-by: Stanley Winata <stanley.winata@amd.com> Signed-off-by: Harsh Menon <harsh@nod-labs.com> Signed-off-by: nithinsubbiah <nithinsubbiah@gmail.com> Signed-off-by: Ivan Butygin <ivan.butygin@gmail.com> Signed-off-by: xintin <gaurav.verma@amd.com> Co-authored-by: Harsh Menon <harsh@nod-labs.com> Co-authored-by: Stanley Winata <stanley.winata@amd.com> Co-authored-by: Stanley Winata <68087699+raikonenfnu@users.noreply.github.com> Co-authored-by: Stanley Winata <stanley@nod-labs.com> Co-authored-by: Ivan Butygin <ivan.butygin@gmail.com> Co-authored-by: nithinsubbiah <nithinsubbiah@gmail.com> Co-authored-by: Nithin Meganathan <18070964+nithinsubbiah@users.noreply.github.com> Co-authored-by: Ivan Butygin <ibutygin@amd.com>
2025-08-13 04:49:11 +08:00
parent 03d114496f
commit 25caa7a8a9
11 changed files with 1437 additions and 0 deletions
--- a/test/srt/test_wave_attention_backend.py
+++ b/test/srt/test_wave_attention_backend.py
@@ -0,0 +1,61 @@
+"""
+Usage:
+python3 -m unittest test_wave_attention_backend.TestWaveAttnBackend.test_mmlu
+"""
+
+import unittest
+from types import SimpleNamespace
+
+from sglang.srt.utils import kill_process_tree
+from sglang.test.run_eval import run_eval
+from sglang.test.test_utils import (
+    DEFAULT_MODEL_NAME_FOR_TEST,
+    DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+    DEFAULT_URL_FOR_TEST,
+    is_in_ci,
+    popen_launch_server,
+    run_bench_one_batch,
+)
+
+
+class TestWaveAttnBackend(unittest.TestCase):
+    def test_latency(self):
+        _, output_throughput, _ = run_bench_one_batch(
+            DEFAULT_MODEL_NAME_FOR_TEST,
+            [
+                "--attention-backend",
+                "wave",
+                "--enable-torch-compile",
+            ],
+        )
+
+        if is_in_ci():
+            self.assertGreater(output_throughput, 153)
+
+    def _test_mmlu(self):
+        model = DEFAULT_MODEL_NAME_FOR_TEST
+        base_url = DEFAULT_URL_FOR_TEST
+        process = popen_launch_server(
+            model,
+            base_url,
+            timeout=DEFAULT_TIMEOUT_FOR_SERVER_LAUNCH,
+            other_args=["--attention-backend", "wave"],
+        )
+
+        try:
+            args = SimpleNamespace(
+                base_url=base_url,
+                model=model,
+                eval_name="mmlu",
+                num_examples=64,
+                num_threads=32,
+            )
+
+            metrics = run_eval(args)
+            self.assertGreaterEqual(metrics["score"], 0.65)
+        finally:
+            kill_process_tree(process.pid)
+
+
+if __name__ == "__main__":
+    unittest.main()