perf: eliminate CUDA syncs in VLM embed path (#26082)

2026-06-12 13:15:56 +08:00
parent cce35ee2e5
commit ca17bd8347
2 changed files with 31 additions and 13 deletions
@@ -146,6 +146,7 @@ class TestKimiVLServer(ImageOpenAITestMixin):
    extra_args = [
        "--context-length=8192",
        "--dtype=bfloat16",
+        "--mem-fraction-static=0.40",
    ]

    def test_video_images_chat_completion(self):