Support Qwen3 Omni thinker cuda graph (#523)

Signed-off-by: ZeldaHuang <hzm414167@alibaba-inc.com>
Co-authored-by: Hongsheng Liu <liuhongsheng4@huawei.com>
This commit is contained in:
Ziming HuangandHongsheng Liu authored and GitHub committed 2025-12-30 23:46:13 +13:00
1 parent 1753a878e0
commit 1ae0c64fe0
5 files changed
+38 -30

No files matched your search

@@ -15,7 +15,7 @@ stage_args:
worker_cls: vllm_omni.worker.gpu_ar_worker.GPUARWorker
scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
gpu_memory_utilization: 0.6
enforce_eager: true
enforce_eager: false
trust_remote_code: true
engine_output_type: latent # Output hidden states for talker
distributed_executor_backend: "mp"
@@ -15,7 +15,7 @@ stage_args:
worker_cls: vllm_omni.worker.gpu_ar_worker.GPUARWorker
scheduler_cls: vllm_omni.core.sched.omni_ar_scheduler.OmniARScheduler
gpu_memory_utilization: 0.6
enforce_eager: true
enforce_eager: false
trust_remote_code: true
engine_output_type: latent # Output hidden states for talker
distributed_executor_backend: "mp"