Fine-tune LLaVA to generate Elasticsearch KQL queries as responses to user prompts

[Sanujen]

I need the LLaVA model to generate KQL queries as responses to user prompts for the specific user prompts and I also need LLaVA to handle the visual question answering, in other words, I need to do the multi-task fine-tuning or PEFT(Parameter Efficient Fine-Tuning)
I am a beginner and I would appreciate your support, thanks.