Support Eagle cuda graph for Triton backend (#3500)

2025-02-12 02:27:45 +08:00
parent cadd5dbe6a
commit 7e6d5fc694
2 changed files with 142 additions and 57 deletions
--- a/test/srt/test_eagle_infer.py
+++ b/test/srt/test_eagle_infer.py
@@ -216,8 +216,6 @@ class TestEAGLEServerTriton(TestEAGLEServer):
                "0.7",
                "--attention-backend",
                "triton",
-                # TODO: Support cuda graph
-                "--disable-cuda-graph",
            ],
        )