Signed-off-by: Gaohan123 <gaohan19@huawei.com>
vLLM-Omni introduces new features to support multimodal comprehension and generation models inference and serving.