[Bugfix] Fix ci bug for qwen2.5-omni (#230)

Signed-off-by: Gaohan123 <hgaoaf@connect.ust.hk>

.buildkite/pipeline.yml

@@ -40,7 +40,7 @@ steps:
       - export VLLM_WORKER_MULTIPROC_METHOD=spawn
       - pytest -s -v tests/multi_stages/
     agents:
-      queue: "gpu_1_queue" # g6.4xlarge instance on AWS, has 1 L4 GPU
+      queue: "gpu_4_queue" # g6.4xlarge instance on AWS, has 1 L4 GPU
     plugins:
       - docker#v5.2.0:
           image: public.ecr.aws/q9t5s3a7/vllm-ci-test-repo:$BUILDKITE_COMMIT

docs/api/README.md

@@ -49,10 +49,9 @@ Engine classes for offline and online inference.
 Core scheduling and caching components.
 
 - [vllm_omni.core.dit_cache_manager.DiTCacheManager][]
-- [vllm_omni.core.sched.diffusion_scheduler.DiffusionScheduler][]
-- [vllm_omni.core.sched.generation_scheduler.GenerationScheduler][]
+- [vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler][]
+- [vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler][]
 - [vllm_omni.core.sched.output.OmniNewRequestData][]
-- [vllm_omni.core.sched.scheduler.OmniScheduler][]
 
 ## Model Executor
 
@@ -108,13 +107,11 @@ Worker classes and model runners for distributed inference.
 - [vllm_omni.diffusion.worker.npu.npu_worker.NPUWorkerProc][]
 - [vllm_omni.worker.gpu_ar_model_runner.GPUARModelRunner][]
 - [vllm_omni.worker.gpu_ar_worker.GPUARWorker][]
-- [vllm_omni.worker.gpu_diffusion_model_runner.GPUDiffusionModelRunner][]
-- [vllm_omni.worker.gpu_diffusion_worker.GPUDiffusionWorker][]
 - [vllm_omni.worker.gpu_generation_model_runner.GPUGenerationModelRunner][]
 - [vllm_omni.worker.gpu_generation_worker.GPUGenerationWorker][]
 - [vllm_omni.worker.gpu_model_runner.OmniGPUModelRunner][]
 - [vllm_omni.worker.npu.npu_ar_model_runner.NPUARModelRunner][]
 - [vllm_omni.worker.npu.npu_ar_worker.NPUARWorker][]
-- [vllm_omni.worker.npu.npu_diffusion_model_runner.NPUDiffusionModelRunner][]
-- [vllm_omni.worker.npu.npu_diffusion_worker.NPUDiffusionWorker][]
+- [vllm_omni.worker.npu.npu_generation_model_runner.NPUGenerationModelRunner][]
+- [vllm_omni.worker.npu.npu_generation_worker.NPUGenerationWorker][]
 - [vllm_omni.worker.npu.npu_model_runner.OmniNPUModelRunner][]

docs/user_guide/examples/online_serving/qwen3_omni.md

@@ -114,6 +114,10 @@ The gradio script supports the following arguments:
     ``````py
     --8<-- "examples/online_serving/qwen3_omni/openai_chat_completion_client_for_multimodal_generation.py"
     ``````
+??? abstract "qwen3_omni_moe_thinking.yaml"
+    ``````yaml
+    --8<-- "examples/online_serving/qwen3_omni/qwen3_omni_moe_thinking.yaml"
+    ``````
 ??? abstract "run_curl_multimodal_generation.sh"
     ``````sh
     --8<-- "examples/online_serving/qwen3_omni/run_curl_multimodal_generation.sh"

tests/multi_stages/stage_configs/qwen2_5_omni_ci.yaml

@@ -16,7 +16,7 @@ stage_args:
       max_model_len: 896
       max_num_batched_tokens: 896
       max_num_seqs: 1
-      gpu_memory_utilization: 0.64
+      gpu_memory_utilization: 0.8
       skip_mm_profiling: true
       enforce_eager: true  # Now we only support eager mode
       trust_remote_code: true
@@ -36,7 +36,7 @@ stage_args:
   - stage_id: 1
     runtime:
       process: true
-      devices: "0"
+      devices: "1"
       max_batch_size: 1
     engine_args:
       model_stage: talker
@@ -46,7 +46,7 @@ stage_args:
       max_model_len: 896
       max_num_batched_tokens: 896
       max_num_seqs: 1
-      gpu_memory_utilization: 0.28
+      gpu_memory_utilization: 0.8
       skip_mm_profiling: true
       enforce_eager: true
       trust_remote_code: true
@@ -73,7 +73,7 @@ stage_args:
       model_arch: Qwen2_5OmniForConditionalGeneration
       worker_cls: vllm_omni.worker.gpu_generation_worker.GPUGenerationWorker
       scheduler_cls: vllm_omni.core.sched.omni_generation_scheduler.OmniGenerationScheduler
-      gpu_memory_utilization: 0.07
+      gpu_memory_utilization: 0.15
       enforce_eager: true
       trust_remote_code: true
       enable_prefix_caching: false