vllm_omni.worker.mixins ¶ OmniWorkerMixin ¶ Shared Omni plugin and native KV connector setup for workers. initialize_from_config ¶ initialize_from_config(kv_cache_config) -> None maybe_unpad_input_ids ¶ maybe_unpad_input_ids( model, input_ids, num_tokens_unpadded ) Trim cudagraph-bucket padding for models that split a flat input_ids by per-request seq_token_counts. See #6712.