Yes, this can work. I’ve done that in a few cases. In fact, if you split data preprocessing in small enough steps, they could also be run on weaker LLMs. It wo

Yes, this can work. I’ve done that in a few cases. In fact, if you split data preprocessing in small enough steps, they could also be run on weaker LLMs. It would take a lot more time, but that is doable.