Update audio tensor access in offline inference tests and increase image fetch timeout for CI

Signed-off-by: tzhouam <tzhouam@connect.ust.hk>

tests/e2e/offline_inference/test_qwen2_5_omni.py

@@ -88,7 +88,7 @@ def test_mixed_modalities_to_audio(omni_runner: type[OmniRunner], test_config: t
         assert len(audio_output.request_output) > 0
 
         # Verify audio tensor exists and has content
-        audio_tensor = audio_output.request_output[0].multimodal_output["audio"]
+        audio_tensor = audio_output.request_output[0].outputs[0].multimodal_output["audio"]
         assert audio_tensor is not None
         assert audio_tensor.numel() > 0
 

tests/e2e/offline_inference/test_qwen3_omni.py

@@ -75,6 +75,6 @@ def test_video_to_audio(omni_runner: type[OmniRunner], test_config) -> None:
         assert len(audio_output.request_output) > 0
 
         # Verify audio tensor exists and has content
-        audio_tensor = audio_output.request_output[0].multimodal_output["audio"]
+        audio_tensor = audio_output.request_output[0].outputs[0].multimodal_output["audio"]
         assert audio_tensor is not None
         assert audio_tensor.numel() > 0

tests/e2e/online_serving/test_image_gen_edit.py

@@ -23,6 +23,8 @@ from vllm.assets.image import ImageAsset
 from vllm.utils.network_utils import get_open_port
 
 os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn"
+# Increase timeout for downloading assets from S3 (default 5s is too short for CI)
+os.environ.setdefault("VLLM_IMAGE_FETCH_TIMEOUT", "60")
 
 models = ["Qwen/Qwen-Image-Edit-2509"]
 test_params = models