Canlin Guo

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED. ## Purpose Related to #348. When the discussion is sufficient on that issue,...

[NPU][Upgrade] Upgrade to v0.12.0

3

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED. ## Purpose Upgrade vllm-omni on NPU with vllm v0.12.0 and vllm-ascend v0.12.0rc1. ##...

[Feature] Support torch profiler across omni stages

3

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED. ## Purpose Close #481. vLLM-Omni supports profiling via PyTorch's built-in profiler, allowing users...

[Perf] Use vLLM's SharedFusedMoE in Qwen3-Omni

2

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED. ## Purpose ## Test Plan ## Test Result execute_model: 177 ms ---> 123...

Here is a minor change in `ci.yml`, which doesn't make some real effect but for more secure usage of actions. [GitHub official docs](https://docs.github.com/en/actions/security-for-github-actions/security-guides/security-hardening-for-github-actions) introduced it. ``` Make sure the credentials...