What is the process to create and publish a custom pipeline?


I have created (locally) a custom pipeline building on the translation pipeline, but with my own pre-processing and post-processing steps, which I want to be integrated in the pipeline. However I cannot see how to easily publish the pipeline, without adding it as a new task in the pipeline registry. My pipeline is not very generic and is very specific to a particular model, so I would ideally like to provide it alongside the model. What are the recommendations for providing such a pipeline, or is there another way to add custom preprocessing and post-processing that cannot be integrated within a tokeniser.

Thanks in advance for your help!