Batch Inference Support

[AnandSingh-0619]

Hey,
I am currently working on integrating the YOLO-World model inside a training loop. However, I noticed that the demo code provided only supports single-image inference. I would like to know if the current codebase supports batch inference, as it would be beneficial for efficiently processing multiple images during training.