From ffbc1305e63f399e0db0677b7140d6c97f664e6d Mon Sep 17 00:00:00 2001 From: Yan Ma Date: Thu, 5 Feb 2026 23:04:57 +0800 Subject: [PATCH] [XPU] Fix XPU UTs for basic coverage (#1164) Signed-off-by: Yan Ma --- tests/e2e/offline_inference/test_cache_dit.py | 3 ++- tests/e2e/offline_inference/test_t2i_model.py | 2 +- tests/e2e/offline_inference/test_t2v_model.py | 3 ++- 3 files changed, 5 insertions(+), 3 deletions(-) diff --git a/tests/e2e/offline_inference/test_cache_dit.py b/tests/e2e/offline_inference/test_cache_dit.py index 18d3988e0..281ac48dc 100644 --- a/tests/e2e/offline_inference/test_cache_dit.py +++ b/tests/e2e/offline_inference/test_cache_dit.py @@ -24,6 +24,7 @@ if str(REPO_ROOT) not in sys.path: from vllm_omni import Omni from vllm_omni.outputs import OmniRequestOutput +from vllm_omni.platforms import current_omni_platform os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "1" @@ -62,7 +63,7 @@ def test_cache_dit(model_name: str): width=width, num_inference_steps=num_inference_steps, guidance_scale=0.0, - generator=torch.Generator("cuda").manual_seed(42), + generator=torch.Generator(current_omni_platform.device_type).manual_seed(42), num_outputs_per_prompt=1, # Single output for speed ), ) diff --git a/tests/e2e/offline_inference/test_t2i_model.py b/tests/e2e/offline_inference/test_t2i_model.py index e7351f157..74c1bd1ce 100644 --- a/tests/e2e/offline_inference/test_t2i_model.py +++ b/tests/e2e/offline_inference/test_t2i_model.py @@ -47,7 +47,7 @@ def test_diffusion_model(model_name: str): width=width, num_inference_steps=2, guidance_scale=0.0, - generator=torch.Generator("cuda").manual_seed(42), + generator=torch.Generator(current_omni_platform.device_type).manual_seed(42), num_outputs_per_prompt=2, ), ) diff --git a/tests/e2e/offline_inference/test_t2v_model.py b/tests/e2e/offline_inference/test_t2v_model.py index a378291ac..9565534bc 100644 --- a/tests/e2e/offline_inference/test_t2v_model.py +++ b/tests/e2e/offline_inference/test_t2v_model.py @@ -14,6 +14,7 @@ if str(REPO_ROOT) not in sys.path: from vllm_omni import Omni from vllm_omni.outputs import OmniRequestOutput +from vllm_omni.platforms import current_omni_platform # os.environ["VLLM_TEST_CLEAN_GPU_MEMORY"] = "1" os.environ["VLLM_WORKER_MULTIPROC_METHOD"] = "spawn" @@ -42,7 +43,7 @@ def test_video_diffusion_model(model_name: str): num_frames=num_frames, num_inference_steps=2, guidance_scale=1.0, - generator=torch.Generator("cuda").manual_seed(42), + generator=torch.Generator(current_omni_platform.device_type).manual_seed(42), ), ) first_output = outputs[0]