Batched inference

[Deep-unlearning]

Hi, thanks for the awesome work on liquid-audio!

Is there any support or plan for batched ASR inference (processing multiple audio clips at once) in the current API? If not, could you point to the parts of the code where batching could be added for the ASR pipeline (generate_sequential, etc.)?

Thanks!